Monitoring river water quality through predictive modeling using artificial neural networks backpropagation

Muhammad Andang Novianta; Syafrudin; Budi Warsito; Siti Rachmawati; Muhammad Andang Novianta; Syafrudin; Budi Warsito; Siti Rachmawati

doi:10.3934/environsci.2024032

AIMS Environmental Science

2024, Volume 11, Issue 4: 649-664. doi: 10.3934/environsci.2024032

Previous Article Next Article

Research article Special Issues

Monitoring river water quality through predictive modeling using artificial neural networks backpropagation

1.
Department of Doctoral Environmental Science, Faculty of Postgraduate, Universitas Diponegoro, Semarang 50275, Indonesia
2.
Department of Electrical Engineering, Faculty of Engineering, Universitas AKPRIND Indonesia, Yogyakarta 55222, Indonesia
3.
Department of Environmental Engineering, Faculty of Engineering, Universitas Diponegoro, Semarang 50275, Indonesia
4.
Department of Statistics, Faculty of Science and Mathematics, Universitas Diponegoro, Semarang 50275, Indonesia
5.
Department of Environmental Science, Faculty of Mathematics and Natural Sciences, Universitas Sebelas Maret, Surakarta 57126, Indonesia

Received: 02 July 2024 Revised: 05 August 2024 Accepted: 07 August 2024 Published: 19 August 2024

Predicting river water quality in the Special Region of Yogyakarta (DIY) is crucial. In this research, we modeled a river water quality prediction system using the artificial neural network (ANN) backpropagation method. Backpropagation is one of the developments of the multilayer perceptron (MLP) network, which can reduce the level of prediction error by adjusting the weights based on the difference in output and the desired target. Water quality parameters included biochemical oxygen demand (BOD), chemical oxygen demand (COD), total suspended solids (TSS), dissolved oxygen (DO), total phosphate, fecal coliforms, and total coliforms. The research object was the upstream, downstream, and middle parts of the Oya River. The data source was secondary data from the DIY Environment and Forestry Service. Data were in the form of time series data for 2013–2023. Descriptive data results showed that the water quality of the Oya River in 2020–2023 was better than in previous years. However, increasing community and industrial activities can reduce water quality. This was concluded based on the prediction results of the ANN backpropagation method with a hidden layer number of 4. The prediction results for period 3 in 2023 and period 1 in 2024 are that 1) the concentrations of BOD, fecal coli, and total coli will increase and exceed quality standards, 2) COD and TSS concentrations will increase but will still be below quality standards, 3) DO and total phosphate concentrations will remain constant and still on the threshold of quality standards. The possibility of several water quality parameters increasing above the quality standards remains, so the potential for contamination of the Oya River is still high. Therefore, early prevention of river water pollution is necessary.

Keywords:

artificial neural network (ANN) backpropagation,
system modeling,
prediction

Citation: Muhammad Andang Novianta, Syafrudin, Budi Warsito, Siti Rachmawati. Monitoring river water quality through predictive modeling using artificial neural networks backpropagation[J]. AIMS Environmental Science, 2024, 11(4): 649-664. doi: 10.3934/environsci.2024032

Related Papers:

[1]	Sohaib Ahmad, Sardar Hussain, Javid Shabbir, Muhammad Aamir, M. El-Morshedy, Zubair Ahmad, Sharifah Alrajhi . Improved generalized class of estimators in estimating the finite population mean using two auxiliary variables under two-stage sampling. AIMS Mathematics, 2022, 7(6): 10609-10624. doi: 10.3934/math.2022592
[2]	Khazan Sher, Muhammad Ameeq, Sidra Naz, Basem A. Alkhaleel, Muhammad Muneeb Hassan, Olayan Albalawi . Developing and evaluating efficient estimators for finite population mean in two-phase sampling. AIMS Mathematics, 2025, 10(4): 8907-8925. doi: 10.3934/math.2025408
[3]	Tolga Zaman, Cem Kadilar . Exponential ratio and product type estimators of the mean in stratified two-phase sampling. AIMS Mathematics, 2021, 6(5): 4265-4279. doi: 10.3934/math.2021252
[4]	Xuechen Liu, Muhammad Arslan . A general class of estimators on estimating population mean using the auxiliary proportions under simple and two phase sampling. AIMS Mathematics, 2021, 6(12): 13592-13607. doi: 10.3934/math.2021790
[5]	Olayan Albalawi . Estimation techniques utilizing dual auxiliary variables in stratified two-phase sampling. AIMS Mathematics, 2024, 9(11): 33139-33160. doi: 10.3934/math.20241582
[6]	Amber Yousaf Dar, Nadia Saeed, Moustafa Omar Ahmed Abu-Shawiesh, Saman Hanif Shahbaz, Muhammad Qaiser Shahbaz . A new class of ratio type estimators in single- and two-phase sampling. AIMS Mathematics, 2022, 7(8): 14208-14226. doi: 10.3934/math.2022783
[7]	Yasir Hassan, Muhammad Ismai, Will Murray, Muhammad Qaiser Shahbaz . Efficient estimation combining exponential and ln functions under two phase sampling. AIMS Mathematics, 2020, 5(6): 7605-7623. doi: 10.3934/math.2020486
[8]	Hleil Alrweili, Fatimah A. Almulhim . Estimation of the finite population mean using extreme values and ranks of the auxiliary variable in two-phase sampling. AIMS Mathematics, 2025, 10(4): 8794-8817. doi: 10.3934/math.2025403
[9]	Sohaib Ahmad, Sardar Hussain, Muhammad Aamir, Faridoon Khan, Mohammed N Alshahrani, Mohammed Alqawba . Estimation of finite population mean using dual auxiliary variable for non-response using simple random sampling. AIMS Mathematics, 2022, 7(3): 4592-4613. doi: 10.3934/math.2022256
[10]	Saman Hanif Shahbaz, Aisha Fayomi, Muhammad Qaiser Shahbaz . Estimation of the general population parameter in single- and two-phase sampling. AIMS Mathematics, 2023, 8(7): 14951-14977. doi: 10.3934/math.2023763

Abstract

1. Introduction

In large scale surveys, it is a usual practice to prefer multi-stage sampling to estimate the population characteristics over single-stage sampling. The main purpose to use multi-stage sampling is the clear reduction in the cost of survey operations even if estimates derived from multi-stage sampling are likely to be less efficient than those of the single-stage sampling. Sukhatme et al. ^[1] advised some ratio and regression type estimators in two-stage sampling using single auxiliary variable when first stage units are of unequal or equal sizes. The suitable use of auxiliary information in estimation stage results as a considerable reduction in the mean square error of the estimator. By making use of auxiliary information, Srivastava and Garg ^[2] proposed separate-type estimator for the estimation of population mean in two-stage sampling design. Taking inspiration from Srivastava and Garg ^[2], Koyuncu and Kadilar ^[3] and Jabeen et al. ^[4] proposed separate type estimator under two-stage sampling. In the literature, the use of two-phase sampling under two-stage sampling design is not well documented. Saini and Bahl ^[5] proposed estimator under two-stage sampling design using double sampling for stratification and multi-auxiliary information. A generalized ratio cum product estimator for population mean in simple random sampling was developed by Singh et al. ^[6]. Shabbir ^[7] produces the estimators of population mean under stratified two phase sampling.

In the literature (^[8,9,10,11]), the use of two phase sampling under two stage sampling design is not well documented. Also the shrinkage estimators have been discussed several times in literature by considering the unbiasedness of the estimators but no one has discussed the situation when the property of unbiasedness is not fulfill that is very common in real life applications. In order to fill this gap, we are motivated to produce two-exponential estimators under two stage two phase sampling that is discussed in section 2. Also we will discuss general shrinkage estimator in section 3. We will compare both estimators mathematically and by using real population data in section 4 and 5. Finally the conclusion will be discussed in section 6.

Let a population consists of N first stage units, each containing ${M_i}$ second-stage units where i = 1, 2, …, N. Let a first-stage sample of size n ( $\subset$ N) is selected and subsequently a second-stage sample of ${m_{i(1)}}$ ( $\subset$ ${M_i}$ ) units is selected and information on some auxiliary variables say ${x_{ij(1)}}$ is taken. Here it is assumed that each first stage unit/cluster is of different size so each cluster is assigned a weight ${\eta _i} = \frac{{{M_i}}}{{\bar M}}$ to it. Let a sub-sample (second-phase sample) of ${m_{i(2)}}$ units is selected from ${m_{i(1)}}$ (first-phase sample) such that ${m_{i(2)}} \subset {M_i}$ . Let ${m_{i(2)}}$ units are observed so as to collect information regarding study variable ${y_{ij(2)}}$ and auxiliary variables ${x_{ij(2)}}$ . Let ${\bar X_{ts}} = \frac{1}{N}\sum\limits_{i = 1}^N {\frac{1}{{\bar M}}{M_i}{{\bar X}_{i.}}}$ be the mean in the population and ${\bar X_i} = \frac{1}{{{M_i}}}\sum\limits_{j = 1}^{{M_i}} {{x_{ij}}}$ be the mean of i^th first stage unit in the population. Let ${\bar x_{ts(1)}} = \frac{1}{n}\sum\limits_{i = 1}^n {{\eta _i}{{\bar x}_{i(1)}}}$ and ${\bar x_{i(2)}} = \frac{1}{{{m_{i(2)}}}}\sum\limits_{j = 1}^{{m_{i(2)}}} {{x_{ij(2)}}}$ respectively are the means of first-phase and second-phase sample in two-stage sampling where ${\bar x_{i(1)}}$ and ${\bar x_{i(2)}}$ be the sample means of first phase and second phase in ith stage. Let $S_{_{xb}}^2 = \frac{1}{{N - 1}}\sum\limits_{i = 1}^N {{{({\eta _i}{{\bar X}_i} - {{\bar X}_{ts}})}^2}}$ be the population variance between fsu’s and $S_{_{xwi}}^2 = \frac{1}{{{M_i} - 1}}\sum\limits_{i = 1}^n {{{({x_{ij}} - {{\bar X}_{i.}})}^2}}$ be the population variance within first stage units. Similarly these notations can be defined for other variables. Further we consider that the selection of units at each stage (or phase) has been made by simple random sampling without replacement.

2. Materials and method

2.1. Proposed generalized estimator

We propose a generalized estimator by considering the exponential relationship as:

${t^G} = {\bar y_{ts}}{\mkern 1mu} {\rm{exp}}\left( {\alpha \left( {1 - \frac{{a{\mkern 1mu} {{\bar x}_{ts(1)}}}}{{\left( {\bar X + (a - 1){\mkern 1mu} {{\bar x}_{ts(1)}}\;\;} \right)}}\;} \right)} \right){\rm{exp}}\left( {\beta \left( {1 - \frac{{b{\mkern 1mu} {{\bar z}_{ts(2)}}}}{{\left( {\bar Z + (b - 1){\mkern 1mu} {{\bar z}_{ts(2)}}\;\;} \right)}}\;} \right)\;} \right),{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu}$

(1)

Where ( $a$ , $b$ ) are constants to be determined such that the mean square error is minimum, ( $\alpha$ , $\beta$ ) are known constants takes the value (0, 1, -1) to produce different ratio-type and product-type estimators as presented in Table B1 (Appendix B).

2.2. Bias and mean square error of generalized estimator

To derive the bias and mean square error, we proceed as follows:

$\frac{{\bar y_{ts(2)}^{} - {{\bar Y}_{ts}}}}{{{{\bar Y}_{ts}}}} = {e_{0(2)}}\, , \, \, \frac{{\bar x_{ts(1)}^{} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}}}} = {e_{1(1)}}, \, \, \, \, \frac{{\bar z_{ts(2)}^{} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}}}} = {e_{2(2)}}, \, \, \frac{{\bar x_{ts(2)}^{} - {{\bar X}^{}}_{ts}}}{{{{\bar X}^{}}_{ts}}} = {e_{1(2)}}, \, \, \, \frac{{\, \bar z_{ts(1)}^{} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}}}} = {e_{2(1)}}$

Further we assume that $E\left({{e_{0(2)}}} \right) = E\left({{e_{1(2)}}} \right) = E\left({{e_{2(2)}}} \right) = E\left({{e_{2(1)}}} \right) = E\left({{e_{1(1)}}} \right) = 0$ , and some expectations under two-stage sampling design are obtained in order to obtain the bias and mean square error as,

$\left.\begin{array}{l} E({e_{0(2)}}^2) = {C_{020(2)}}, E({e_{1(2)}}^2) = {C_{200(2)}}, \\ E({e_{2(2)}}^2) = {C_{002(2)}}, E({e_{1(1)}}^2) = {C_{200(1)}}, \\ E({e_{2(1)}}^2) = {C_{002(1)}}, E({e_{1(2)}}{e_{1(1)}}) = {C_{200(1)}}\, \\ E({e_{0(2)}}{e_{1(2)}}) = {C_{110(2)}}, \, E({e_{1(2)}}{e_{2(2)}}) = {C_{101(2)}}\, , \\ E({e_{0(2)}}{e_{2(2)}}) = {C_{011(2)}}, E({e_{0(2)}}{e_{1(1)}}) = {C_{110(1)}}, \, \\ E({e_{1(2)}}{e_{2(1)}}) = {C_{101(1)}}\, , E({e_{0(2)}}{e_{2(1)}}) = {C_{011(1)}}\, \, \, \, \, \\ {\rm{Where}} \\ {C_{020(2)}} = \frac{1}{{{{\bar Y}^2}_{ts}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right)\, S_{yb}^2 + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\eta _i^2\, \, \left( {\frac{1}{{{m_{i(2)}}}} - \frac{1}{{{M_i}}}} \right)} \, S_{ywi}^2, } \right\}\, , \\ {C_{200(1)}} = \frac{1}{{{{\bar X}^2}_{ts}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right)\, \, S_{xb}^2 + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\eta _i^2\, \left( {\frac{1}{{{m_{i(1)}}}} - \frac{1}{{{M_i}}}} \right)} \, S_{xwi}^2} \right\} \\ {C_{200(2)}} = \frac{1}{{{{\bar X}^2}_{ts}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right)\, S_{xb}^2 + \frac{1}{{nN}}\sum\limits_{i = 1}^N {\eta _i^2\, \left( {\frac{1}{{{m_{i(2)}}}} - \frac{1}{{{M_i}}}} \right)} \, S_{xwi}^2} \right\}, \\ {C_{002(1)}} = \frac{1}{{{{\bar Z}^2}_{ts}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right)\, \, S_{zb}^2 + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\eta _i^2\left( {\frac{1}{{{m_{i(1)}}}} - \frac{1}{{{M_i}}}} \right)} \, \, S_{zwi}^2} \right\}, \\ {C_{002(2)}} = \frac{1}{{{{\bar Z}^2}_{ts}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right)S_{zb}^2 + \frac{1}{{nN}}\sum\limits_{i = 1}^N {\eta _i^2\left( {\frac{1}{{{m_{i(2)}}}} - \frac{1}{{{M_i}}}} \right)} S_{zwi}^2} \right\}, \\ {C_{110(1)}} = \frac{1}{{{{\bar Y}_{ts}}\, {{\bar X}_{ts}}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right){\rho _{xyb}}\, {S_{yb}}\, {S_{xb}} + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\eta _i^2\left( {\frac{1}{{{m_{i(1)}}}} - \frac{1}{{{M_i}}}} \right)} \, {\rho _{xywi}}\, {S_{ywi}}\, {S_{xwi}}} \right\}, \\ {C_{110(2)}} = \frac{1}{{{{\bar Y}_s}\, {{\bar X}_s}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right){\rho _{xyb}}\, {S_{yb}}\, {S_{xb}} + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\, \eta _i^2\, \left( {\frac{1}{{{m_{i(2)}}}} - \frac{1}{{{M_i}}}} \right)} \, {\rho _{xywi}}\, {S_{ywi}}\, {S_{xwi}}} \right\}, \, \\ {C_{101(1)}} = \frac{1}{{{{\bar Z}_{ts}}\, {{\bar X}_{ts}}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right){\rho _{xzb}}\, {S_{zb}}\, {S_{xb}} + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\, \eta _i^2\, \left( {\frac{1}{{{m_{i(1)}}}} - \frac{1}{{{M_i}}}} \right)} \, {\rho _{xzwi}}\, {S_{zwi}}\, {S_{xwi}}} \right\}, \\ {C_{101(2)}} = \frac{1}{{{{\bar Z}_{ts}}\, {{\bar X}_{ts}}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right){\rho _{xzb}}\, {S_{zb}}\, {S_{xb}} + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\eta _i^2\left( {\frac{1}{{{m_{i(2)}}}} - \frac{1}{{{M_i}}}} \right)} \, {\rho _{xzwi}}{S_{zwi}}{S_{xwi}}} \right\}, \,\\ {C_{200(1)}} = \frac{1}{{{{\bar Z}_{ts\, }}{{\bar X}_{ts}}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right)\, {S_{xb}}\, {S_{xb}} + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\eta _i^2\, \left( {\frac{1}{{{m_{i(1)}}}} - \frac{1}{{{M_i}}}} \right)} \, {S_{xwi}}\, {S_{xwi}}} \right\} \\ {C_{011(2)}} = \frac{1}{{{{\bar Z}_{ts}}{{\bar Y}_{ts}}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right)\, {\rho _{yzb}}\, {S_{zb}}\, {S_{yb}} + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\eta _i^2\left( {\frac{1}{{{m_{i(2)}}}} - \frac{1}{{{M_i}}}} \right)} \, {\rho _{yzwi}}{S_{zwi}}{S_{ywi}}} \right\}, \,\\ {C_{011(1)}} = \frac{1}{{{{\bar Z}_{ts}}\, {{\bar Y}_{ts}}}}\left\{ {\left( {\frac{1}{n} - \frac{1}{N}} \right)\, {\rho _{yzb}}\, {S_{zb}}\, {S_{yb}} + \frac{1}{{n\, N}}\sum\limits_{i = 1}^N {\eta _i^2\left( {\frac{1}{{{m_{i(1)}}}} - \frac{1}{{{M_i}}}} \right)} \, {\rho _{yzwi}}{S_{zwi}}{S_{ywi}}} \right\} \\ \\ \end{array} \right\}\, \,$

(2)

Using (2) we can express (1) to derive the bias and mean square error as,

${t^G} = {\bar Y_{ts}}\left( {1 + {e_{0(2)}}} \right){\rm{exp}}\left[ { - \frac{\alpha }{a}{e_{1(1)}}{{\left( {1 + \frac{{(a - 1)}}{a}{e_{1(1)}}} \right)}^{ - 1}}} \right]{\rm{exp}}\left[ { - \frac{\beta }{b}{e_{2(2)}}{{\left( {1 + \frac{{(b - 1)}}{b}{e_{2(2)}}} \right)}^{ - 1}}} \right],$

(3)

If $\left| {{e_{1(1)}}} \right| < 1$ , we expand the series, ${\left({1 + \frac{{(a - 1)}}{a}{e_{1(1)}}} \right)^{ - 1}}$ and ${\left({1 + \frac{{(b - 1)}}{b}{e_{2(2)}}} \right)^{ - 1}}$ up to the order n^-1, we get,

$\begin{gathered} {t^G} = {{\bar Y}_{ts}}\left( {1 + {e_{0(2)}}} \right){\rm{exp}}\left[ { - \frac{\alpha }{a}{e_{1(1)}}\left( {1 - \frac{{(a - 1)}}{a}{e_{1(1)}} + \frac{{{{(a - 1)}^2}}}{{{a^2}}}{e_{1(1)}}^2 + ...} \right)} \right]\, \\ \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, {\rm{exp}}\left[ { - \frac{\beta }{b}{e_{2(2)}}\left( {1 - \frac{{(b - 1)}}{b}{e_{2(2)}} + \frac{{{{(b - 1)}^2}}}{{{b^2}}}{e_{2(2)}}^2 + ...} \right)} \right], \, \\ \end{gathered}$

(4)

If the contribution of terms involving powers in ${e_{0(2)}}$ , ${e_{1(1)}}$ and ${e_{2(2)}}$ higher than two is negligible, we have;

${t^G} = {\bar Y_{ts}}\left( {1 + {e_{0(2)}}} \right)\left[ {1 - \frac{\alpha }{a}{e_{1(1)}} + \frac{{\alpha {{(a - 1)}^2}}}{{{a^2}}}{e^2}_{1(1)} + \frac{1}{2}\frac{{{\alpha ^2}}}{{{a^2}}}{e^2}_{1(1)}} \right]\, \left[ {1 - \frac{\beta }{b}{e_{2(2)}} + \frac{{\beta {{(b - 1)}^2}}}{{{b^2}}}{e^2}_{2(2)} + \frac{1}{2}\frac{{{\beta ^2}}}{{{b^2}}}{e^2}_{2(2)}} \right], \,$

(5)

${t^G} - {\bar Y_{ts}} = \, \, {\bar Y_{ts}}\left[ \begin{array}{l} {e_{0(2)}} - \frac{\alpha }{a}{e_{1(1)}} + \frac{{\alpha (a - 1)}}{{{a^2}}}{e^2}_{1(1)} + \frac{{\alpha \, \beta }}{{a\, b}}\, {e_{1(1)}}\, {e_{2(2)}} + \frac{{{\alpha ^2}}}{{2\, {a^2}}}\, {e^2}_{1(1)} - \frac{\beta }{b}{e_{2(2)}} + \frac{{{\beta ^2}}}{{2\, {b^2}}}\, {e_{2(2)}}^2 + \frac{{\beta \, (b - 1)}}{{{b^2}}}\, {e_{2(2)}}^2 \\ - \frac{\beta }{b}\, {e_{2(2)}}\, {e_{0(2)}} - \frac{\alpha }{a}\, {e_{1(1)}}\, {e_{0(2)}} \\ \end{array} \right],$

(6)

In order to get the bias, we take expectation on (6) and get,

$Bias\, (\, {t^G}) = {\bar Y_{ts}}\, \, \left[ {\frac{{{\alpha ^2}}}{{2\, {a^2}}}{C_{200(1)}} + \frac{{\alpha (a - 1)}}{{{a^2}}}{C_{200(1)}} + \frac{{{\beta ^2}}}{{2\, {b^2}}}{C_{002(2)}} + \frac{{\beta \, (b - 1)}}{{{b^2}}}{C_{002(2)}} - \frac{\beta }{b}{C_{011(2)}} - \frac{\alpha }{a}{C_{110(1)}} + \frac{\alpha }{a}\frac{\beta }{b}{C_{101(1)}}} \right]\, ,$

(7)

To get the mean square error of the estimator, we take square and retain terms up to first order of e’s then we take expectation of (6) and we obtain,

${\rm{MSE}}\left( {\, {t^G}} \right) = {\bar Y^2}_{ts}\, \left( {{C_{020(2)}} + {{\left( {\frac{\alpha }{a}} \right)}^2}{C_{200(1)}} + {{\left( {\frac{\beta }{b}} \right)}^2}{C_{002(2)}} - 2\, \frac{\alpha }{a}\, {C_{110(1)}} - 2\frac{\beta }{b}{C_{011(2)}} + 2\frac{\alpha }{a}\frac{\beta }{b}{C_{101(1)}}} \right)\, , \,$

(8)

For the following optimal value of the constants $\, a$ and $\, b$ , we achieve the minimum Variance among the class of proposed generalized estimator,

$\begin{gathered} a = \frac{{\alpha \left( {{C_{200(1)}}{C_{002(2)}} - {C_{101(2)}}^2} \right)}}{{{C_{110(1)}}{C_{002(2)}} - {C_{011(1)}}{C_{101(2)}}}} , \, \, \, \, \, \, \, {\rm{and }}\, b = \frac{{\beta \left( {{C_{200(1)}}\, {C_{002(2)}} - {C_{101(2)}}^2} \right)}}{{{C_{200(1)}}\, {C_{011(1)}} - {C_{110(1)}}{C_{101(2)}}}}, \end{gathered}$

(9)

We obtain minimum mean square error as,

$\min MSE\left( {{t^G}} \right) = {\bar Y^2}_{ts}\, \left( {{C_{020(2)}} - \frac{{{C_{110(1)}}^2{C_{002(2)}} + {C_{011(1)}}^2{C_{200(1)}} - 2{C_{110(1)}}{C_{101(2)}}{C_{011(1)}}}}{{{C_{200(1)}}{C_{002(2)}} - {C_{101(2)}}^2}}} \right),$

(10)

We observe from (10) that proposed generalized estimator gives us more precise results under the optimal conditions, as compare to its class of the estimators.

From (7)–(10), we get expressions of the bias, mean square error, optimal values and minimum mean square error for the exponential-type estimators presented in Table B1 as class of estimators. Some of the results are discussed as,

ⅰ). For $\alpha = 1, \beta = 1$ , we get a class of exponential-type ratio-cum-ratio estimators given in Table B1 and expressions of the mean square error and bias for these estimators are given as,

$\left. \begin{array}{l} {\rm{MSE}}{\mkern 1mu} (t_{}^{j - err}) = \left\{ {\bar Y_{ts}^2\left( {{C_{020(2)}} + \frac{1}{{{a^2}}}{\mkern 1mu} {C_{200(1)}} + \frac{1}{{{b^2}}}{\mkern 1mu} {C_{002(2)}} - 2\frac{1}{a}{C_{110(1)}} - 2\frac{1}{b}{C_{011(2)}} + 2\frac{1}{{ab}}{C_{101(1)}}} \right)\;\;j( \in G) = 1,3,5} \right\},\\ {\rm{and}}\\ {\rm{Bias}}{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} (t_{}^{j - err}) = \left\{ {\bar Y_{ts}^{}\left( {\frac{{a - 1}}{{{a^2}}}{C_{200(1)}} + \frac{1}{{2{\mkern 1mu} {a^2}}}{C_{200(1)}} + \frac{{b - 1}}{{{b^2}}}{C_{002(2)}} + \frac{1}{{2{\mkern 1mu} {b^2}}}{C_{002(2)}} - \frac{1}{b}{C_{101(2)}} - \frac{1}{a}{C_{110(1)}}} \right)\;\;j( \in G) = 1,3,5} \right\}, \end{array} \right\}$

(11)

Substitution of the different values of a and b yield the mean square error and bias for the estimators belongs to the class of exponential-type ratio-cum-ratio estimators. The optimal values which lead to minimum mean square error for the class of exponential ratio-cum-ratio estimators is obtained as,

$a = \frac{{{C_{200(1)}}\;\;{C_{002(2)}}\; - \;\;{C_{101(2)}}^2}}{{{C_{110(1)}}\;\;{C_{002(2)}}\;\; - \;{C_{011(1)}}\;\;{C_{101(2)}}}}{\mkern 1mu} \;\;{\mkern 1mu} {\rm{and}}{\mkern 1mu} {\mkern 1mu} b = \frac{{{C_{200(1)}}\;\;{C_{002(2)}}\; - \;\;{C_{101(2)}}^2}}{{{C_{200(1)}}\;\;{C_{011(1)}}\;\; - \;{C_{110(1)}}\;\;{C_{101(2)}}}}\;\;\;.$

ⅱ). For $\alpha = - 1, \beta = - 1$ , we get exponential-type product-cum-product estimators given in Table B1.The mean square error of $t_{}^G$ are expressed using as,

$\left. \begin{array}{l} {\rm{MSE}}(t_{}^{j - epp}) = \left\{ {\bar Y_{ts}^2\left( {{C_{020(2)}} + \frac{1}{{{a^2}}}{C_{200(1)}} + \frac{1}{{{b^2}}}{C_{002(2)}} + 2\frac{1}{a}{C_{110(1)}} + 2\frac{1}{b}{C_{011(2)}} + 2\frac{1}{{ab}}{C_{101(1)}}} \right)\;,\;\;\;\;\;\;\;\;\;\;\;j( \in G) = 2,4,6} \right\}\\ {\rm{and}}\\ {\rm{Biast}}_{}^{j - epp}) = \left\{ {\bar Y_{ts}^{}\left( {\frac{{a - 1}}{{{a^2}}}{C_{200(1)}} + \frac{1}{{2{a^2}}}{C_{200(1)}} + \frac{{b - 1}}{{{b^2}}}{C_{002(2)}} + \frac{1}{{2{b^2}}}{C_{002(2)}} - \frac{1}{b}{C_{101(2)}} + \frac{1}{a}{C_{110(1)}}} \right),\;\;\;\;\;\;\;\;\;\;j( \in G) = 2,4,6} \right\} \end{array} \right)$

(12)

Substitution of the different values of “a” and “b” yield the Mean Square Error and bias for the estimators belongs to the class of exponential-type ratio-cum-ratio estimators. The optimal values which lead to minimum mean square error for the class of exponential product-cum-product estimators is obtained as,

$a = \frac{{{C_{101(2)}}^2\;\; - {C_{200(1)}}\;\;{C_{002(2)}}}}{{{C_{110(1)}}\;\;{C_{002(2)}} - {C_{011(1)}}\;\;{C_{101(2)}}}}{\mkern 1mu} \;\;{\mkern 1mu} {\rm{and}}{\mkern 1mu} {\mkern 1mu} b = \frac{{{C_{101(2)}}^2\;\; - {C_{200(1)}}\;\;{C_{002(2)}}}}{{{C_{200(1)}}\;\;{C_{011(1)}}\;\; - {C_{110(1)}}\;\;{C_{101(2)}}}}\;\;.$

ⅲ). For $\alpha = - 1, \beta = 1$ , we get exponential-type product-cum-ratio estimators given in Table B1.The mean square error of $t_{}^G$ is expressed as,

$\left. \begin{array}{l} {\rm{MSE}}\, (t_{}^{j - erp}) = \left\{ {\bar Y_{ts}^2\left( {{C_{020(2)}} + \frac{1}{a}{C_{200(1)}} + \frac{1}{b}{C_{002(2)}} + 2\frac{1}{a}{C_{110(1)}} - 2\frac{1}{b}{C_{011(2)}} - 2\frac{1}{{ab}}{C_{101(1)}}} \right), \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, j( \in G) = 7, 9, 11} \right\} \\ {\rm{and}} \\ {\rm{Bias}} \, (t_{}^{j - erp}) = \left\{ {\bar Y_{ts}^{}\left( {\frac{{a - 1}}{{{a^2}}}{C_{200(1)}} + \frac{1}{{2\, {a^2}}}{C_{200(1)}} + \frac{{b - 1}}{{{b^2}}}{C_{002(2)}} + \frac{1}{{2\, {b^2}}}{C_{002(2)}} - \frac{1}{b}{C_{101(2)}} + \frac{1}{a}{C_{110(1)}}} \right) \, , \, \, \, j( \in G) = 7, 9, 11} \right\} \\ \end{array} \right)\, \,$

(13)

Substitution of the different values of “a” and “b” yield the mean square error and bias for the estimators belongs to the class of exponential-type ratio-cum-ratio estimators. The optimal values which lead to minimum mean square error for the class of exponential product-cum-ratio estimators are obtained as,

$a = \frac{{{C_{101(2)}}^2\; - {C_{200(1)}}\;\;{C_{002(2)}}}}{{{C_{110(1)}}\;\;{C_{002(2)}}\;\; - \;{C_{011(1)}}\;\;{C_{101(2)}}}}{\mkern 1mu} {\mkern 1mu} \;\;{\rm{and}}{\mkern 1mu} {\mkern 1mu} b = \frac{{{C_{200(1)}}\;\;{C_{011(1)}}\;\; - \;{C_{110(1)}}{C_{101(2)}}}}{{{C_{200(1)}}\;\;{C_{002(2)}}\;\; - {C_{101(2)}}^2}}\;\;.$

ⅳ). For $\alpha = 1, \beta = - 1$ , we get exponential-type ratio-cum-product estimators given in Table A1. The mean square error of ${t^G}$ is expressed as,

$\left. \begin{array}{l} {\rm{MSE}}\, \, (t_{}^{j - epr}) = \left\{ {\bar Y_{ts}^2\left( {{C_{020(2)}} + \frac{1}{a}{C_{200(1)}} + \frac{1}{b}{C_{002(2)}} - 2\frac{1}{a}{C_{110(1)}} + 2\frac{1}{b}{C_{011(2)}} - 2\frac{1}{a}\frac{1}{b}{C_{101(1)}}} \right),\, \, \, \, \, \, \, j( \in G) = 8, 10, 12} \right\} \\ {\rm{and}} \\ {\rm{Bias}}\, \, (t_{}^{j - epr}) = \left\{ {\bar Y_{ts}^{}\left( {\frac{{a - 1}}{{{a^2}}}{C_{200(1)}} + \frac{1}{{2\, {a^2}}}{C_{200(1)}} + \frac{{b - 1}}{{{b^2}}}{C_{002(2)}} + \frac{1}{{2\, {b^2}}}{C_{002(2)}} + \frac{1}{b}{C_{101(2)}} - \frac{1}{a}{C_{110(1)}}} \right) \, , \, \, \, \, \, \, \, \, \, \, j( \in G) = 8, 10, 12} \right\} \\ \end{array} \right)\, \,$

(14)

$a = \frac{{{C_{200(1)}}\;\;{C_{002(2)}}\;\; - {C_{101(2)}}^2}}{{{C_{110(1)}}\;\;{C_{002(2)}}\;\; - {C_{011(1)}}{\;\;C_{101(2)}}}}\;\;{\mkern 1mu} {\mkern 1mu} {\rm{and}}{\mkern 1mu} {\mkern 1mu} b = \frac{{{C_{101}}^2 - {C_{200(1)\;\;}}{C_{002(2)}}}}{{{C_{200(1)}}\;\;{C_{011(1)}} - {C_{110(1)}}\;\;{C_{101(2)}}}}\;\;.$

It is to mention that minimum mean square error of $t_{}^{j - epr}$ , $t_{}^{j - epr}$ , $t_{}^{j - epr}$ , and $t_{}^{j - epr}$ in (11)–(14) will be same but the conditions under which these attain minimum value are not unique.

3. Proposed two-exponential general Shrinkage estimator

In literature, the shrinkage estimators have been discussed several times considering the property that the estimators are unbiased. In many practical situations the property of unbiasedness is not met so consequently the existing literature will not be helpful. Such deviations from the said property have motivated us to consider the shrinkage estimators following the property of biasedness. Important results are presented in theorem and subsequently proof is provided in Appendix A.

3.1. Theorem for estimating any population characteristic

For any general estimator $t$ , let ${t_s}$ be a general shrinkage estimator for estimating any population characteristic T (e.g., population mean) which is unknown to us,

$t_s^{} = \lambda t$

(15)

A general form for the bias and mean square error of a general shrinkage estimator up to the first order of approximation may be given as,

$Bias\left( {t_s^{}} \right) = \left( {\lambda - 1} \right)T + Bias\left( t \right)$

(16)

and

$MSE\left( {{t_s}} \right) = {T^2}{\left( {\lambda - 1} \right)^2} + {\lambda ^2}MSE\left( t \right) + 2T\lambda (\lambda - 1)Bias(t)$

(17)

And form of the optimal shrinkage estimator in (15) along with minimum bias and minimum mean square error will be as,

${t_s} = \frac{{\left( {{T^2} + MSE(t) + 2TBias(t)} \right)}}{{T(T + Bias(t)}}\;\;\;t$

(18)

$Bias\left( {t_{s}^{opt}} \right) = \left( {\frac{{\left( {{T^2} + MSE(t) + 2TBias(t)} \right)}}{{T(T + Bias(t)}} - 1} \right)T + Bias\left( t \right)$

(19)

and

$\min \, MSE\left( {t_s^{}} \right) = {T^2} - \frac{{{T^2}(T + Bias(t)}}{{{T^2} + MSE(t) + 2TBias(t)}}$

(20)

Proof: For details see Appendix A.

By using (15), generalized form of a shrinkage class of estimator is,

$t_s^G = \lambda \, {t_G}$

where

${t_G} = {\bar y_{ts}}{\mkern 1mu} {\rm{exp}}\left( {\alpha \left( {1 - \frac{{a{\mkern 1mu} {{\bar x}_{ts(1)}}}}{{\left( {{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts(1)}}\;\;} \right)}}} \right)} \right){\rm{exp}}\left( {\beta \left( {1 - \frac{{b{\mkern 1mu} {{\bar z}_{ts(2)}}}}{{\left( {{{\bar Z}_{ts}} + (b - 1){{\bar z}_{ts(2)}}\;\;} \right)}}} \right)} \right)$

(21)

For different choices of the values of the constants $\alpha, \beta, a, b$ , we get different family of shrinkage estimators as given in Table B2 (See Appendix).

Following the Theorem, we can directly write the mean square error and bias expressions of $t_s^G$ using (7) and (8) respectively as,

$Bias\, (t_s^G) = \lambda \, {\bar Y_{ts}}\left[ {\frac{{\alpha \, (a - 1)}}{{{a^2}}}{C_{200(1)}} + \frac{{{\alpha ^2}}}{{2{a^2}}}\, {C_{200(1)}} + \frac{{\beta \, (b - 1)}}{{{b^2}}}{C_{002(2)}} + \frac{{{\beta ^2}}}{{2{b^2}}}{C_{002(2)}} - \frac{\beta }{b}{C_{101(2)}} - \frac{\alpha }{a}{C_{110(1)}}} \right] + {\bar Y_{ts}}(\lambda - 1)\, , {\rm{or}}$

$Bias\, (t_s^G) = \lambda \, Bias(t_{}^{G - e}) + {\bar Y_{ts}}\, (\lambda - 1)\, ,$

(22)

and

$\begin{gathered} MSE\, \left( {t_s^G} \right) = {\lambda ^2}\, {{\bar Y}^2}_{ts}\left( {{C_{020(2)}} + {{\left( {\frac{\alpha }{a}} \right)}^2}{C_{200(1)}} + {{\left( {\frac{\beta }{b}} \right)}^2}{C_{002(2)}} - 2\, \frac{\alpha }{a}{C_{110(1)}} - 2\frac{\beta }{b}{C_{011(2)}} + 2\frac{\alpha }{a}\frac{\beta }{b}{C_{101(1)}}} \right) + {\left( {\lambda - 1} \right)^2}{{\bar Y}_{ts}}^2 \\ \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, + 2\lambda \left( {\lambda - 1} \right){{\bar Y}_{ts}}\left[ {\frac{{\alpha \, (a - 1)}}{{{a^2}}}{C_{200(1)}} + \frac{{{\alpha ^2}}}{{2\, {a^2}}}{C_{200(1)}} + \frac{{\beta \, (b - 1)}}{{{b^2}}}{C_{002(2)}} + \frac{{{\beta ^2}}}{{2\, {b^2}}}{C_{002(2)}} - \frac{\beta }{b}{C_{101(2)}} - \frac{\alpha }{a}{C_{110(1)}}} \right] \\ \end{gathered}$

$MSE\left( {t_s^G} \right) = {\lambda ^2}MSE\left( {t_{}^{G - e}} \right) + {\left( {\lambda - 1} \right)^2}{\bar Y_{ts}}^2 + 2{\bar Y_{ts}}\, \lambda \left( {\lambda - 1} \right)MSE\left( {t_{}^{G - e}} \right)$

(23)

The optimal value of $\lambda$ , which minimize the expression (23) is obtained as,

$\lambda = \frac{{\left( {{T^2} + MSE{\mkern 1mu} ({t^{G - e}}\;\;) + 2{\mkern 1mu} T{\mkern 1mu} Bias({t^{G - e\;\;\;}})} \right)}}{{T{\mkern 1mu} (T + Bias({t^{G - e}}\;\;)}}$

(24)

Using (23), the optimal estimator for the shrinkage family of estimator ( $t_s^G$ ) is written as

$\left. {t_s^G = {{\bar y}_{ts}}\frac{{\left( {{{\bar Y}_{ts}}^2 + MSE({t^{G - e}}\;\;) + 2{{\bar Y}_{ts}}Bias({t^{G - e}}\;)} \right)}}{{{{\bar Y}_{ts}}({{\bar Y}_{ts}} + Bias({t^{G - e}}\;)}}{\mkern 1mu} {\rm{exp}}\left( {\alpha {\mkern 1mu} \left( {1 - \frac{{a{\mkern 1mu} {{\bar x}_{ts(1)}}}}{{\left( {{{\bar X}_{ts}} + (a - 1){\mkern 1mu} {{\bar x}_{ts(1)}}} \right)}}\;} \right)} \right){\rm{exp}}\left( {\beta \left( {1 - \frac{{b{\mkern 1mu} {{\bar z}_{ts(2)}}}}{{\left( {{{\bar Z}_{ts}} + (b - 1){\mkern 1mu} {{\bar z}_{ts(2)}}\;} \right)}}} \right)} \right)} \right\}$

(25)

The expression for minimum value of min $MSE\left({t_s^G} \right)$ is obtained as,

$\begin{gathered} \left. {\min MSE\left( {t_s^G} \right) = {{\bar Y}^2}_{ts} - \frac{{{{\bar Y}^2}_{ts}{{\left( {{{\bar Y}^{}}_{ts} + Bias\left( {{t^{G - e}}} \right)} \right)}^2}}}{{\left[ {{{\bar Y}^2}_{ts} + MSE\left( {{t^{G - e}}} \right) + 2{{\bar Y}^{}}_{ts}Bias\left( {{t^{G - e}}} \right)} \right]}}} \right\}\, \, \, \, \, \, \\ \min MSE\left( {t_s^G} \right) = {{\bar Y}^2}_{ts}\left[ {1 - \frac{{{{\left( {1 + \frac{{Bias\left( {{t_a}} \right)}}{{{{\bar Y}^{}}_{ts}}}} \right)}^2}}}{{\left[ {1 + \frac{{MSE\left( {{t_a}} \right)}}{{{{\bar Y}^2}_{ts}}} + 2\frac{{Bias\left( {{t_a}} \right)}}{{{{\bar Y}^{}}_{ts}}}} \right]}}} \right]\, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \\ \end{gathered}$

(26)

From (22)–(26), we get expressions of the bias, mean square error, optimal values and minimum mean square error for the exponential-type estimators presented in Table B2 (See Appendix B) as class of estimators. Some of the results are discussed as,

ⅰ). For $\alpha = 1, \beta = 1$ , we get a class of exponential-type ratio-cum-ratio estimators given in Table B2 (See Appendix) and expressions of the mean square error and bias for these estimators are given as,

$\begin{array}{*{20}{l}} {{\rm{MSE}}{\mkern 1mu} (t_s^{j - err\;\;}) = \left\{ {{{\bar Y}^2}_{ts}\left[ {1 - \frac{{{{\left( {1 + \frac{{Bias\left( {t_{}^{i - err\;\;}} \right)}}{{{{\bar Y}_{ts}}}}\;\;} \right)}^2}}}{{\left[ {1 + \frac{{MSE{\mkern 1mu} \left( {t_{}^{i - err\;\;}\;\;} \right)}}{{{{\bar Y}^2}_{ts}}} + 2\frac{{Bias\left( {t_{}^{i - err\;\;}} \right)}}{{{{\bar Y}^2}_{ts}}}} \right]\;\;}}\;} \right]{\mkern 1mu} \;j( \in G) = 1,3,5} \right\}}\\ {{\rm{and}}}\\ {{\rm{Bias}}{\mkern 1mu} (t_s^{j - err\;\;}) = \left\{ {\bar Y_{ts}^{}\left( {\lambda \left( {\frac{{{\rm{Bias}}{\mkern 1mu} (t_{}^{j - err\;\;}\;\;)}}{{\bar Y_{ts}^{}}}} \right) + (\lambda - 1)} \right){\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} j( \in G) = 1,3,5} \right\}{\mkern 1mu} {\mkern 1mu} } \end{array}$

(27)

Substitution of the different values of a and b produce the mean square error and bias for the estimators belongs to the class of exponential-type ratio-cum-ratio estimators. The optimal values which lead to minimum mean square error for the class of exponential product-cum-ratio estimators are obtained as,

$\begin{gathered} a = \frac{{{C_{200(1)}}{C_{002(2)}} - {C_{101(2)}}^2}}{{{C_{110(1)}}{C_{002(2)}} - {C_{011(1)}}{C_{101(2)}}}}\, \, and\, \, b = \frac{{{C_{200(1)}}{C_{002(2)}} - {C_{101(2)}}^2}}{{{C_{200(1)}}{C_{011(1)}} - {C_{110(1)}}{C_{101(2)}}}}, \, \, \lambda = \frac{{\left( {{T^2} + MSE\, ({t^{G - e}}) + 2\, T\, Bias\, ({t^{G - e}})} \right)}}{{T(T + Bias\, ({t^{G - e}})}}, \\ \, \, \, \, \, \, \, where\, \, MSE\left( {{t^G}} \right) = {{\bar Y}_{ts}}^2\left[ {{C_{020(2)}} + \frac{1}{{{a^2}}}{C_{200(1)}} + \frac{1}{{{b^2}}}{C_{002(2)}} - 2\frac{1}{a}{C_{110(1)}} - 2\frac{1}{b}{C_{011(2)}} + 2\frac{1}{a}\frac{1}{b}{C_{101(1)}}} \right]. \\ \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, {\rm{Bias}}\, (t_{}^G) = \bar Y_{ts}^{}\left( {\frac{{a - 1}}{{{a^2}}}{C_{200(1)}} + \frac{1}{{2\, {a^2}}}{C_{200(1)}} + \frac{{b - 1}}{{{b^2}}}{C_{002(2)}} + \frac{1}{{2\, {b^2}}}{C_{002(2)}} - \frac{1}{b}{C_{101(2)}} - \frac{1}{a}{C_{110(1)}}} \right) \\ \end{gathered}$

ⅱ). For $\alpha = - 1, \beta = - 1$ , we get exponential-type product-cum-product estimators given in Table B2 (See Appendix B).The mean square error of $t_s^G$ is expressed using as,

$\begin{array}{*{20}{l}} {{\rm{MSE}}{\mkern 1mu} (t_s^{j - epp\;}) = \left\{ {{{\bar Y}^2}_{ts}\left[ {1 - \frac{{{{\left( {1 + \frac{{Bias\left( {t_{}^{i - err}\;\;} \right)}}{{{{\bar Y}_{ts}}}}\;} \right)}^2}}}{{\left[ {1 + \frac{{MSE\left( {t_{}^{i - err}} \right)}}{{{{\bar Y}^2}_{ts}}} + 2\frac{{Bias\left( {t_{}^{i - err}} \right)}}{{{{\bar Y}^2}_{ts}}}\;} \right]}}\;} \right]{\mkern 1mu} {\mkern 1mu} j( \in G) = 2,4,6} \right\}}\\ {{\rm{and}}}\\ {{\rm{Bias}}{\mkern 1mu} (t_s^{j - epp}\;) = \left\{ {\bar Y_{ts}^{}\left( {\lambda \frac{{{\rm{Bias}}{\mkern 1mu} (t_s^{j - epp}\;\;)}}{{\bar Y_{ts}^{}}} + (\lambda - 1)} \right){\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} j( \in G) = 2,4,6} \right\}} \end{array}$

(28)

Substitution of the different values of a and b yield the mean square error and bias for the estimators belongs to the class of exponential-type product-cum-product estimators. The optimal values which lead to minimum mean square error for the class of exponential product-cum-ratio estimators are obtained as,

$\begin{array}{*{20}{l}} {a = \frac{{{C_{200(1)}}\;\;{C_{002(2)}}\;\; - {C_{101(2)}}^2}}{{{C_{110(1)}}\;\;{C_{002(2)}}\;\; - {C_{011(1)\;\;}}{C_{101(2)}}}}{\mkern 1mu} {\mkern 1mu} \;\;and{\mkern 1mu} {\mkern 1mu} b = \frac{{{C_{200(1)}}\;\;{C_{002(2)}}\;\; - {C_{101(2)}}^2}}{{{C_{200(1)}}\;\;{C_{011(1)}}\;\; - {C_{110(1)}}\;\;{C_{101(2)}}}},\lambda = \frac{{\left( {{T^2} + MSE({t^{G - e}}\;\;) + 2TBias({t^{G - e}}\;\;)} \right)}}{{T(T + Bias({t^{G - e}}\;)}},}\\ {{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} where{\mkern 1mu} {\mkern 1mu} MSE\left( {{t^G}} \right) = {{\bar Y}_{ts}}^2\left[ {{C_{020(2)}} + \frac{1}{{{a^2}}}{C_{200(1)}} + \frac{1}{{{b^2}}}{C_{002(2)}} - 2\frac{1}{a}{C_{110(1)}} - 2\frac{1}{b}{C_{011(2)}} + 2\frac{1}{a}\frac{1}{b}{C_{101(1)}}} \right].}\\ {{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\rm{Bias}}t_{}^G) = \bar Y_{ts}^{}\left( {\frac{{a - 1}}{{{a^2}}}{C_{200(1)}} + \frac{1}{{2{a^2}}}{C_{200(1)}} + \frac{{b - 1}}{{{b^2}}}{C_{002(2)}} + \frac{1}{{2{b^2}}}{C_{002(2)}} - \frac{1}{b}{C_{101(2)}} - \frac{1}{a}{C_{110(1)}}} \right)} \end{array}$

ⅲ). For $\alpha = - 1, \beta = 1$ , we get exponential-type product-cum-ratio estimators given in Table B2. The mean square error of $t_s^G$ is expressed as,

$\begin{array}{*{20}{l}} {{\rm{MSE}}{\mkern 1mu} (t_s^{j - erp\;\;}) = \left\{ {{{\bar Y}^2}_{ts}\left[ {1 - \frac{{{{\left( {1 + \frac{{Bias\left( {t_{}^{i - err}} \right)}}{{{{\bar Y}_{ts}}}}\;\;} \right)}^2}}}{{\left[ {1 + \frac{{MSE\left( {t_{}^{i - err}\;} \right)}}{{{{\bar Y}^2}_{ts}}} + 2\frac{{Bias\left( {t_{}^{i - err}} \right)}}{{{{\bar Y}^2}_{ts}}}} \right]}}} \right]{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} j( \in G) = 7,9,11} \right\}}\\ {{\rm{and}}}\\ {{\rm{Bias}}{\mkern 1mu} (t_s^{j - erp\;\;}\;) = \left\{ {\bar Y_{ts}^{}\left( {\lambda \frac{{{\rm{Bias}}{\mkern 1mu} (t_s^{j - erp\;\;}\;\;)}}{{\bar Y_{ts}^{}}} + (\lambda - 1)} \right)j( \in G) = 7,9,11} \right\}} \end{array}$

(29)

Substitution of the different values of a and b produce the mean square error and bias for the estimators belongs to the class of exponential-type ratio-cum-product estimators. The optimal values which lead to minimum mean square error for the class of exponential ratio-cum-product estimators, are obtained as,

$\begin{array}{*{20}{l}} {a = \frac{{{C_{200(1)}}\;\;{C_{002(2)}} - {C_{101(2)}}^2}}{{{C_{110(1)}}{C_{002(2)}} - {C_{011(1)}}{C_{101(2)}}}}\;\;\;{\mkern 1mu} {\mkern 1mu} and{\mkern 1mu} {\mkern 1mu} b = \frac{{{C_{200(1)}}\;\;{C_{002(2)}} - {C_{101(2)}}^2}}{{{C_{200(1)}}\;\;{C_{011(1)}} - \;\;{C_{110(1)}}{C_{101(2)}}}}\;\;,\lambda = \frac{{\left( {{T^2} + MSE({t^{G - e}}\;) + 2TBias({t^{G - e}}\;\;)} \right)}}{{T(T + Bias({t^{G - e}})}},}\\ {{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} where{\mkern 1mu} {\mkern 1mu} MSE\left( {{t^{G - e}}} \right) = {{\bar Y}_{ts}}^2\left[ {{C_{020(2)}} + \frac{1}{{{a^2}}}{C_{200(1)}} + \frac{1}{{{b^2}}}{C_{002(2)}} - 2\frac{1}{a}{C_{110(1)}} - 2\frac{1}{b}{C_{011(2)}} + 2\frac{1}{a}\frac{1}{b}{C_{101(1)}}} \right].}\\ {{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} Biast_{}^{G - e}) = \bar Y_{ts}^{}\left( {\frac{{a - 1}}{{{a^2}}}{C_{200(1)}} + \frac{1}{{2{a^2}}}{C_{200(1)}} + \frac{{b - 1}}{{{b^2}}}{C_{002(2)}} + \frac{1}{{2{b^2}}}{C_{002(2)}} - \frac{1}{b}{C_{101(2)}} - \frac{1}{a}{C_{110(1)}}} \right)} \end{array}$

ⅳ). For $\alpha = 1, \beta = - 1$ , we get exponential-type ratio-cum-product estimators given in Table B2. The mean square error of $t_s^G$ is expressed as,

$\begin{array}{*{20}{l}} {{\rm{MSE}}{\mkern 1mu} (t_s^{j - epr}) = \left\{ {{{\bar Y}^2}_{ts}\left[ {1 - \frac{{{{\left( {1 + \frac{{Bias\left( {t_{}^{i - err}} \right)}}{{{{\bar Y}_{ts}}}}\;\;} \right)}^2}}}{{\left[ {1 + \frac{{MSE\left( {t_{}^{i - err}\;} \right)}}{{{{\bar Y}^2}_{ts}}} + 2\frac{{Bias\left( {t_{}^{i - err}} \right)}}{{{{\bar Y}^2}_{ts}}}\;\;\;} \right]}}} \right]{\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} j( \in G) = 8,10,12} \right\}}\\ {{\rm{and}}}\\ {{\rm{Bias}}{\mkern 1mu} (t_s^{j - epr}) = \left\{ {\bar Y_{ts}^{}\left( {\lambda \frac{{{\rm{Bias}}{\mkern 1mu} (t_s^{j - epr\;})}}{{\bar Y_{ts}^{}}} + (\lambda - 1)} \right){\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} {\mkern 1mu} j( \in G) = 8,10,12} \right\}} \end{array}$

(30)

Substitution of the different values of a and b yield the mean square error and bias for the estimators belongs to the class of exponential-type product-cum-ratio estimators. The optimal values which lead to minimum mean square error for the class of exponential product-cum-ratio estimators, are obtained as,

$\begin{gathered} a = \frac{{{C_{200(1)}}{C_{002(2)}} - {C_{101(2)}}^2}}{{{C_{110(1)}}{C_{002(2)}} - {C_{011(1)}}{C_{101(2)}}}}\, \, and\, \, b = \frac{{{C_{200(1)}}{C_{002(2)}} - {C_{101(2)}}^2}}{{{C_{200(1)}}{C_{011(1)}} - {C_{110(1)}}{C_{101(2)}}}}, \lambda = \frac{{\left( {{T^2} + MSE\, ({t^{G - e}}) + 2\, T\, Bias({t^{G - e}})} \right)}}{{T(T + Bias({t^{G - e}})}}, \\ \, \, \, \, \, \, \, where\, \, MSE\, \left( {{t^G}} \right) = {{\bar Y}_{ts}}^2\left[ {{C_{020(2)}} + \frac{1}{{{a^2}}}{C_{200(1)}} + \frac{1}{{{b^2}}}{C_{002(2)}} - 2\frac{1}{a}{C_{110(1)}} - 2\frac{1}{b}{C_{011(2)}} + 2\frac{1}{a}\frac{1}{b}{C_{101(1)}}} \right]. \\ \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, \, {\rm{Bias}}\, \, (t_{}^G) = \bar Y_{ts}^{}\left( {\frac{{a - 1}}{{{a^2}}}{C_{200(1)}} + \frac{1}{{2\, {a^2}}}{C_{200(1)}} + \frac{{b - 1}}{{{b^2}}}{C_{002(2)}} + \frac{1}{{2\, {b^2}}}{C_{002(2)}} - \frac{1}{b}{C_{101(2)}} - \frac{1}{a}{C_{110(1)}}} \right) \\ \end{gathered}$

It is to mention that minimum mean square error of $t_s^{j - epr}$ , $t_s^{j - epr}$ , $t_s^{j - epr}$ , and $t_s^{j - epr}$ in (27)–(30) will be same but the conditions under which these attain minimum value are not unique.

4. Efficiency comparison

In this section, some efficiency conditions have been derived in terms of mean square error and bias. The efficiency comparison of proposed two-exponential general shrinkage estimator (see section 3) and two-exponential estimator (see section 2) with unbiased estimator in two-stage sampling. In order to derive the efficiency conditions, consider the following notations,

${A_1} = {C_{020(2)}} + {\left( {\frac{\alpha }{a}} \right)^2}{C_{200(1)}} + {\left( {\frac{\beta }{b}} \right)^2}{C_{002(2)}} - 2\left( {\frac{\alpha }{a}} \right){C_{110(1)}} - 2\left( {\frac{\beta }{b}} \right){C_{011(2)}} + 2\left( {\frac{\alpha }{a}} \right)\left( {\frac{\beta }{b}} \right){C_{101(1)}}$

${A_2} = \left[ {\frac{{\alpha (a - 1)}}{{{a^2}}}{C_{200(1)}} + \frac{{{\alpha ^2}}}{{2{a^2}}}{C_{200(1)}} + \frac{{\beta (b - 1)}}{{{b^2}}}{C_{002(2)}} + \frac{{{\beta ^2}}}{{2{b^2}}}{C_{002(2)}} - \frac{\beta }{b}{C_{101(2)}} - \frac{\alpha }{a}{C_{110(1)}}} \right]$

${A_3} = {C_{020(2)}} + {C_{200(1)}} + {C_{002(2)}} - 2{C_{110(1)}} - 2{C_{011(2)}} + 2{C_{101(1)}}$

${A_4} = \left( {{C_{020(2)}} - \frac{{{C_{110(1)}}^2\;\;{C_{002(2)}} + {C_{011(1)}}^2\;\;{C_{200(1)}}\;\; - 2{C_{110(1)}}\;\;{C_{101(2)}}\;\;{C_{011(1)}}}}{{{C_{200(1)}}\;\;{C_{002(2)}}\;\; - {C_{101(2)}}^2}}} \right)$

${A_5} = \left[ {{\lambda ^2} + 2\lambda (\lambda - 1)\left( {\frac{1}{\alpha }{C_{200(1)}} + \frac{1}{2}{C_{200(1)}}} \right)} \right]$

${A_6} = \left[ {{\lambda ^2}\left( {{C_{110(1)}} - \frac{\beta }{b}{C_{101(!)}}} \right) + \lambda (\lambda - 1)\left( {{C_{200(1)}} - {C_{110(1)}}} \right)} \right]$

${A_7} = \left[ {{\lambda ^2}\left( {{{\left( {\frac{\beta }{b}} \right)}^2}{C_{002(2)}} - 2\frac{\beta }{b}{C_{011(2)}}} \right) + {{(\lambda - 1)}^2} + 2\lambda (\lambda - 1)\left( {\frac{{\beta (b - 1)}}{{{b^2}}}{C_{002(2)}} + {{\left( {\frac{\beta }{b}} \right)}^2}{C_{002(2)}} - \frac{\beta }{b}{C_{101(1)}}} \right)} \right]$

4.1. When is known and is unknown

The efficiency conditions may be written as:

ⅰ. ${MSE}\left(t_{s}^{G}\right)-{MSE}\left(\bar{y}_{t s}\right) \leq 0$ if

$\min \left( {\frac{{\left( {1 - {A_2}} \right) \pm \sqrt {\left( {{{\left( {1 - {A_2}} \right)}^2} - \left( {1 + {A_1} + 2{A_2}} \right)\left( {{C_{020(2)}}} \right)} \right)} }}{{\left( {1 - {A_2}} \right)}}} \right) \leqslant \lambda \leqslant \max \left( {\frac{{\left( {1 - {A_2}} \right) \pm \sqrt {\left( {{{\left( {1 - {A_2}} \right)}^2} - \left( {1 + {A_1} + 2{A_2}} \right)\left( {{C_{020(2)}}} \right)} \right)} }}{{\left( {1 - {A_2}} \;\right)}}} \;\right),$

(31)

ⅱ. $MSE\, (t_s^G) - MSE\, (t_r^{}) \leqslant 0$

$\min \left( {\frac{{\left( {1 - {A_2}} \right) \pm \sqrt {\left( {{{\left( {1 - {A_2}} \right)}^2} - \left( {1 + {A_1} + 2{A_2}} \right)\left( {{A_3}} \right)} \right)} }}{{\left( {1 - {A_2}} \right)}}} \right) \leqslant \lambda \leqslant \max \left( {\frac{{\left( {1 - {A_2}} \right) \pm \sqrt {\left( {{{\left( {1 - {A_2}} \right)}^2} - \left( {1 + {A_1} + 2{A_2}} \right)\left( {{A_3}} \right)} \right)} }}{{\left( {1 - {A_2}} \;\right)}}}\; \right)$

(32)

ⅲ. $MSE\, (t_s^G) - MSE\, (t_{}^G) \leqslant 0$

$\min \left( {\frac{{\left( {1 - {A_2}} \right) \pm \sqrt {\left( {{{\left( {1 - {A_2}} \right)}^2} - \left( {1 + {A_1} + 2{A_2}} \right)\left( {{A_1}} \right)} \right)} }}{{\left( {1 - {A_2}} \right)}}} \right) \leqslant \lambda \leqslant \max \left( {\frac{{\left( {1 - {A_2}} \right) \pm \sqrt {\left( {{{\left( {1 - {A_2}} \right)}^2} - \left( {1 + {A_1} + 2{A_2}} \right)\left( {{A_1}} \right)} \right)} }}{{\left( {1 - {A_2}}\; \right)}}} \right)$

(33)

ⅳ. $MSE\, (t_s^G) - \min MSE\, (t_{}^G) \leqslant 0$

$\min \left( {\frac{{\left( {1 - {A_2}} \right) \pm \sqrt {\left( {{{\left( {1 - {A_2}} \right)}^2} - \left( {1 + {A_1} + 2{A_2}} \right)\left( {{A_4}} \right)} \right)} }}{{\left( {1 - {A_2}} \right)}}} \right) \leqslant \lambda \leqslant \max \left( {\frac{{\left( {1 - {A_2}} \right) \pm \sqrt {\left( {{{\left( {1 - {A_2}} \right)}^2} - \left( {1 + {A_1} + 2{A_2}} \right)\left( {{A_4}} \right)} \right)} }}{{\left( {1 - {A_2}} \right)}}} \;\right)$

(34)

4.2. If a is unknown λ, b is known

ⅰ. ${MSE}\left(t_{s}^{G}\right)-{MSE}\left(\bar{y}_{t s}\right) \leq 0$ if

$\min \left( {\frac{{{A_6} \pm \sqrt {{A_6}^2 - {A_5}\left( {{A_7} - {C_{020(2)}}} \right)} }}{{\alpha {A_5}}}} \right) \leqslant \frac{1}{a} \leqslant \max \left( {\frac{{{A_6} \pm \sqrt {{A_6}^2 - {A_5}\left( {{A_7} - {C_{020(2)}}} \right)} }}{{\alpha {A_5}}}} \right)$

(35)

ⅱ. $MSE\, (t_s^G) - MSE\, (t_r^{}) \leqslant 0$

$\min \left( {\frac{{{A_6} \pm \sqrt {{A_6}^2 - {A_5}\left( {{A_7} - {A_3}} \right)} }}{{\alpha {A_5}}}} \right) \leqslant \frac{1}{a} \leqslant \max \left( {\frac{{{A_6} \pm \sqrt {{A_6}^2 - {A_5}\left( {{A_7} - {A_3}} \right)} }}{{\alpha {A_5}}}} \right)$

(36)

Substituting different values of ( $\alpha, \beta, a, b$ ) in above equations, we get different mathematical comparisons for the estimators given in Table B2 (See Appendix B). Also similar comparison is obtained if we assume $b$ as unknown and $\lambda, a$ is known.

5. Numerical study

For the demonstration of the performance of proposed two-exponential general shrinkage estimator, we take real population consists of four clusters with unequal first stage units. The description about the populations is given in Table B1 in Appendix B. The mean square error and percentage relative efficiency values for each of the estimators are given in Table C2.For this population, correlations within the clusters are positive (see Table C1). It is therefore the population is applicable only for ratio/ratio-type estimators.

The performance of the proposed two-exponential general shrinkage estimator and two-exponential estimator under two stage two phase sampling has been expressed in form of their mean square error and percentage relative efficiency values in Table C2. From Table C2 we see that the minimum mean square error of two-exponential general shrinkage estimator $t_s^G$ that is equal to the mean square error of $t_s^5$ . The percentage relative efficiency demonstrate the same result as the percentage relative efficiency of the proposed two-exponential general shrinkage is high among the family of proposed two-exponential general shrinkage and $t_s^5$ is the same efficient as $t_s^G$ .

We also see that the two-exponential general shrinkage estimator (section 3) is better to be used as compare to two-exponential estimator (section 2). The performance of $t_s^1$ is better than ${t^1}$ , similarly $t_s^3$ performs better than $t_{}^3$ . So it is concluded that for this particular population the use of two-exponential general shrinkage estimator in two stage two phase sampling produce better results.

6. Conclusion

Finally from the above empirical results, it is concluded that the performance of the two-exponential general shrinkage estimator (section 3) is higher as compare to proposed two-exponential estimator in two stage two phase sampling. So the two-exponential general shrinkage estimator (section 3) is acceptable for the real life application in two stage two phase sampling design.

Acknowledgments

We are grateful to the anonymous reviewers for their valuable comments and suggestions in improving the manuscript.

Conflict of interest

The authors declare no conflict of interest.

Appendix A

Proof of Theorem. Let $t$ be any general estimator of some population parameter T in two stage sampling for which a general shrinkage estimator ${t_s}$ is defined as,

$t_s^{} = \lambda t,$

(37)

By definition the bias of $t_s^{}$ is

$Bias\left( {{t_s}} \right) = \lambda E\left( t \right) - T$

$= \lambda E\left( {t - T} \right) + \left( {\lambda - 1} \right)T$

$= \lambda Bias\left( t \right) + \left( {\lambda - 1} \right)T$

(38)

By definition the MSE of ${t_s}$ may be defined as,

$MSE\left( {t_s^{}} \right) = E{\left( {t_s^{} - T} \right)^2}$

$= E{\left( {\lambda t - T} \right)^2}$

consider $t = T(1 + {e_{_T}})$ , with $E({e_{_T}}) = 0$ and $E({e_{_T}}^2) = \frac{{Var(t)}}{{{T^2}}} = \frac{{MSE\left(t \right)}}{{{T^2}}}$ .

$= E{\left\{ {\lambda \left( {T\left( {1 + {e_T}} \right)} \right) - T} \right\}^2}$

$= {T^2}E\left\{ {{\lambda ^2}{{\left( {1 + {e_T}} \right)}^2} + 1 - 2\lambda \left( {1 + {e_T}} \right)} \right\}$

$MSE\left( {t_s^{}} \right) = {T^2}\left\{ {{\lambda ^2}\left( {1 + \frac{{MSE\left( t \right)}}{{{T^2}}}} \right) + 1 - 2\lambda } \right\}$

(39)

We may also write (B-3) as

$MSE\left( {t_s^{}} \right) = {\lambda ^2}MSE\left( t \right) + {\left( {\lambda - 1} \right)^2}{T^2}$

$MSE\left( {t_s^{}} \right) = {\lambda ^2}{T^2}C_t^2 + {\left( {\lambda - 1} \right)^2}{T^2} \;\;{\rm{Where}}\;\;C_t^2 = \frac{{MSE\left( t \right)}}{{{T^2}}}$

(40)

Where $C_t^{}$ is a coefficient of variation of an estimator t.

Partially differentiating (40) w.r.t $\lambda$ and simplifying, we get:

$\lambda = \frac{1}{{1 + {T^{ - 2}}\;MSE\left( t \right)}}\;,$

(41)

$\lambda = \frac{1}{{1 + C_t^2}},$

(42)

Now using (41) in (15), (16) and (17) respectively, we may get,

${t_s} = \frac{t}{{1 + {T^{ - 2}}\;MSE\left( t \right)}}$

(43)

$Bias\left( {t_s^{}} \right) = \frac{1}{{1 + {T^{ - 2}}\;MSE\left( t \right)}}\left[ {Bias\left( t \right) + {T^{ - 2}}MSE\left( t \right)T} \right]$

(44)

$MS{E_{\min }}\left( {t_s^{}} \right) = \frac{{MSE\left( t \right)}}{{1 + {T^{ - 2}}\;MSE\left( t \right)}}$

(45)

Alternatively we may also produce the expressions in (44) to (46) as,

${t_s} = \frac{t}{{1 + C_t^2}}$

(46)

$Bias\left( {t_s^{}} \right) = \frac{1}{{1 + C_t^2}}\left( {Bias\left( t \right) + C_t^2T} \right)$

(47)

$MS{E_{\min }}\left( {t_s^{}} \right) = \frac{{{T^2}C_t^2}}{{1 + C_t^2}}$

(48)

Appendix B

Table B1. Some members of the proposed two exponential estimator (

${t^G}$ ).

Ratio-cum-Ratio estimator $\alpha = 1, \beta = 1$	Product -cum-product estimator $\alpha = - 1, \beta = - 1$	$a$	$b$
${t^{1 - err}} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}} + {{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar Z}_{ts}} - {{\bar z}_{ts(2)}}}}{{{{\bar Z}_{ts}} + {{\bar z}_{ts(2)}}}}} \right)$	$t_{}^{2 - epp} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar x}_{ts(1)}} + {{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar z}_{ts(2)}} + {{\bar Z}_{ts}}}}} \right)$	2	2
$t_{}^{3 - err} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar Z}_{ts}} - {{\bar z}_{ts(2)}}}}{{{{\bar Z}_{ts}}}}} \right)$	$t_{}^{4 - epp} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}}}}} \right)$	1	1
$t_{}^{5 - err} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar Z}_{ts}} - {{\bar z}_{ts(2)}}}}{{{{\bar Z}_{ts}} + \left({b - 1} \right){{\bar z}_{ts(2)}}}}} \right)$	$t_{}^{6 - epp} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + \left({b - 1} \right){{\bar z}_{ts(2)}}}}} \right)$	$a$	$b$
Product -cum-ratio estimator $\alpha = - 1, \beta = 1$	Ratio-cum-product estimator $\alpha = 1, \beta = - 1$	$a$	$b$
$t_{}^{7 - epr} = {\bar y_{ts}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}} + {{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + {{\bar z}_{ts(2)}}}}} \right)$	$t_{}^{8 - erp} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar x}_{ts(1)}} + {{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar z}_{ts(2)}} + {{\bar Z}_{ts}}}}} \right)$	2	2
$t_{}^{9 - epr} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(1)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}}}}} \right)$	$t_{}^{10 - erp} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}}}}} \right)$	1	1
$t_{}^{11 - epr} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + \left({b - 1} \right){{\bar z}_{ts(2)}}}}} \right)$	$t_{}^{12 - erp} = {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + \left({b - 1} \right){{\bar z}_{ts(2)}}}}} \right)$	$a$	$b$

| Show Table

DownLoad: CSV

Table B2. Some members of the generalized shrinkage estimator (

$t_s^G$ ).

Ratio-cum-product estimator $\alpha = 1, \beta = 1$	Product -cum-product estimator $\alpha = - 1, \beta = - 1$	$a$	$\lambda$	$b$
$t_s^{1 - err} = \lambda {\bar y_{ts}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}} + {{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar Z}_{ts}} - {{\bar z}_{ts(2)}}}}{{{{\bar Z}_{ts}} + {{\bar z}_{ts(2)}}}}} \right)$	$t_s^{2 - epp} = \lambda {\bar y_{ts}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}} + {{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + {{\bar z}_{ts(2)}}}}} \right)$	2	$\lambda$	2
$t_s^{3 - err} = \lambda {\bar y_{ts}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar Z}_{ts}} - {{\bar z}_{ts(2)}}}}{{{{\bar Z}_{ts}}}}} \right)$	$t_s^{4 - epp} = \lambda {\bar y_{ts}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}}}}} \right)$	1	$\lambda$	1
$t_s^{5 - erp} = \lambda {\bar y_{ts}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts}}}}} \right)\exp \left({\frac{{{{\bar Z}_{ts}} - {{\bar z}_{ts(2)}}}}{{{{\bar Z}_{ts}} + \left({b - 1} \right){{\bar z}_{ts(2)}}}}} \right)$	$t_s^{6 - epp} = \lambda {\bar y_{ts}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + \left({b - 1} \right){{\bar z}_{ts(2)}}}}} \right)$	$a$	$\lambda$	$b$
Product -cum-ratio estimator $\alpha = - 1, \beta = 1$	Ratio-cum-product estimator $\alpha = 1, \beta = - 1$	$a$	$\lambda$	$b$
$t_s^{7 - epr} = \lambda {\bar y_{ts}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}} + {{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + {{\bar z}_{ts(2)}}}}} \right)$	$t_s^{8 - erp} = \lambda {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar x}_{ts(1)}} + {{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar z}_{ts(2)}} + {{\bar Z}_{ts}}}}} \right)$	2	$\lambda$	2
$t_s^{9 - epr} = \lambda {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(1)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}}}}} \right)$	$t_s^{10 - erp} = \lambda {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}}}}} \right)$	1	$\lambda$	1
$t_s^{11 - epr} = \lambda {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar X}_{ts}} - {{\bar x}_{ts(1)}}}}{{{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + \left({b - 1} \right){{\bar z}_{ts(2)}}}}} \right)$	$t_s^{12 - erp} = \lambda {\bar y_{ts(2)}}\exp \left({\frac{{{{\bar x}_{ts(1)}} - {{\bar X}_{ts}}}}{{{{\bar X}_{ts}} + (a - 1){{\bar x}_{ts(1)}}}}} \right)\exp \left({\frac{{{{\bar z}_{ts(2)}} - {{\bar Z}_{ts}}}}{{{{\bar Z}_{ts}} + \left({b - 1} \right){{\bar z}_{ts(2)}}}}} \right)$	$a$	$\lambda$	$b$

| Show Table

DownLoad: CSV

Appendix C

Table C1. Data Statistics for Population-Ⅰ.

	Population-Ⅰ (unequal fsu’s)
Cluster	1	2	3	4
${M_i}$	18	14	12	20
${m_i}$	9	7	6	10
${\bar Y_i}$	25.77722	22.79286	28.43500	23.0905
${\bar X_i}$	51.06389	46.49700	67.00217	57.11855
${C_{{y_i}}}^2$	0.58025	0.39297	0.34783	0.31545
${C_{{x_i}}}^2$	0.43322	0.29984	0.41947	0.40689
${\rho _{i1}}$	0.88373	0.83895	0.82425	0.82113

| Show Table

DownLoad: CSV

Table C2. MSEs and PREs of the adapted and suggested class of estimators.

Estimators	MSE	PRE With new fraction
${\overline y _s}$	24.44649	100
$t_{}^{1 - err}$	11.66857	209.51
$t_{}^{3 - err}$	20.27969	120.5467
$t_{}^{5 - err}$	11.46787	213.1738
$t_s^{1 - err}$	11.58499	211.0229
$t_s^{3 - err}$	18.92219	129.1974
$t_s^{5 - err}$	11.25906	217.1317
$t_s^{G - opt}$	11.25906	217.1317

| Show Table

DownLoad: CSV

References

[1]	Utama JP, Syafrudin, Nugraha WD (2015) Penentuan Daya tampung beban pencemaran BOD dan Fecal Coliform Sungai Plumbon Kota Semarang Dengan software QUAL2E. J Teknik Lingkungan 4: 1–9.
[2]	Warsito B, Sumiyati S, Yasin H, et al. (2021) Evaluation of river water quality by using hierarchical clustering analysis, IOP Conference Series: Earth and Environmental Science. https://doi.org/10.1088/1755-1315/896/1/012072 doi: 10.1088/1755-1315/896/1/012072
[3]	Zubaidah T, Karnaningroem N, Slamet A (2018) K-means method for clustering water quality status on the rivers of Banjarmasin, Indonesia. ARPN J Eng Appl Sci 13. https://doi.org/10.31227/osf.io/s9n2u doi: 10.31227/osf.io/s9n2u
[4]	Novianta MA, Syafrudin S, Warsito B (2023) K-means clustering for Grouping Rivers in DIY based on water quality parameters. JUITA J Inform 11: 155–163. https://doi.org/10.30595/juita.v11i1.16986 doi: 10.30595/juita.v11i1.16986
[5]	Di Z, Chang M, Guo P (2019) Water quality evaluation of the Yangtze River in China using machine learning techniques and data monitoring on different time scales. Water 11: 339. https://doi.org/10.3390/w11020339 doi: 10.3390/w11020339
[6]	Mohammadrezapour O, Kisi O, Pourahmad F (2020) Fuzzy c-means and K-means clustering with genetic algorithm for identification of homogeneous regions of groundwater quality. Neural Comput Appl 32: 3763–3775. https://doi.org/10.1007/s00521-018-3768-7 doi: 10.1007/s00521-018-3768-7
[7]	Widyastuti M, Marfai MA (2004) Kajian daya tampung sungai gajahwong thdp beban pencemaran. Majalah Geografi Indonesia. 8: 81–97.
[8]	Riza H, Santoso EW, Tejakusuma IG, et al. (2020) Utilization of artificial intelligence to improve flood disaster mitigation. Jurnal Sains dan Teknologi Mitigasi Bencana 30: 1–11. https://doi.org/10.29122/jstmb.v15i1.4145 doi: 10.29122/jstmb.v15i1.4145
[9]	Makubura R, Meddage DPP, Azamathulla HM, et al. (2022) A simplified mathematical formulation for water quality index (WQI): A case study in the Kelani River Basin, Sri Lanka. Fluids 7: 147. https://doi.org/10.3390/fluids7050147 doi: 10.3390/fluids7050147
[10]	Madhushani C, Dananjaya K, Ekanayake IU, et al. (2024) Modeling streamflow in non-gauged watersheds with sparse data considering physiographic, dynamic climate, and anthropogenic factors using explainable soft computing techniques. J Hydrol 631: 130846. https://doi.org/10.1016/j.jhydrol.2024.130846 doi: 10.1016/j.jhydrol.2024.130846
[11]	Perera UAKK, Coralage DTS, Ekanayake IU, et al. (2024) A new frontier in streamflow modeling in ungauged basins with sparse data: A modified generative adversarial network with explainable AI. Results Eng 21: 101920. https://doi.org/10.1016/j.rineng.2024.101920 doi: 10.1016/j.rineng.2024.101920
[12]	Pohan S, Warsito B, Suryono S (2020) Backpropagation artificial neural network for prediction plant seedling growth. J Phys 1524: 012147. https://doi.org/10.1088/1742-6596/1524/1/012147 doi: 10.1088/1742-6596/1524/1/012147
[13]	Purba RA, Samsir S, Siddik M, et al. (2020), The optimization of backpropagation neural networks to simplify decision making, IOP Conference Series: Materials Science and Engineering, 830: 022091. https://doi.org/10.1088/1757-899X/830/2/022091 doi: 10.1088/1757-899X/830/2/022091
[14]	Nurjaya IK, Estananto E, Murti A (2022) Pemodelan sistem kendali suhu otomatis pada Smart poultry farm menggunakan metode jaringan saraf tiruan. eProc Eng 9.
[15]	Haekal M, Wibowo WC (2023) Prediksi kualitas air sungai menggunakan metode pembelajaran mesin: Studi kasus Sungai Ciliwung: Prediction of river water quality using machine learning methods: Ciliwung River case study. Jurnal Teknologi Lingkungan 24: 273–282. https://doi.org/10.55981/jtl.2023.795 doi: 10.55981/jtl.2023.795
[16]	Mustafa HM, Mustapha A, Hayder G, et al. (2021) Applications of iot and artificial intelligence in water quality monitoring and prediction: A review, In 2021 6th International Conference on inventive computation technologies (ICICT), 968–975. https://doi.org/10.1109/ICICT50816.2021.9358675
[17]	DLHK DIY (2021) Dokumen Informasi Kinerja Pengelolaan Lingkungan Hidup Daerah (DIKPLHD) Provinsi DIY Tahun 2021, Yogyakarta.
[18]	Masruroh M (2020) Perbandingan metode regresi linear dan neural network backpropagation dalam prediksi nilai Ujian nasional siswa smp menggunakan software R. Joutica 5: 331–336. https://doi.org/10.30736/jti.v5i1.347 doi: 10.30736/jti.v5i1.347
[19]	Wanto A (2019) Prediksi produktivitas jagung di Indonesia sebagai upaya antisipasi impor menggunakan jaringan saraf tiruan backpropagation, SINTECH 2: 53–62. https://doi.org/10.31598/sintechjournal.v2i1.355 doi: 10.31598/sintechjournal.v2i1.355
[20]	Sudarsono A (2016) Jaringan syaraf tiruan untuk memprediksi laju pertambahan penduduk menggunakan metode backpropagation (studi Kasus di Kota Bengkulu). Jurnal Media Infotama 12. https://doi.org/10.37676/jmi.v12i1.273 doi: 10.37676/jmi.v12i1.273
[21]	Cynthia EP, Ismanto E (2017) Jaringan syaraf tiruan algoritma backpropagation dalam memprediksi ketersediaan komoditi pangan provinsi riau. RABIT Jurnal Teknologi Dan Sistem Informasi Univrab 2. https://doi.org/10.36341/rabit.v2i2.152 doi: 10.36341/rabit.v2i2.152
[22]	Aggarwal L, Sahoo BM (2018) Back propagation algorithm for computerized paper evaluation using neural network. IJSRST 4.
[23]	Bande S, Shete VV (2017) Smart flood disaster prediction system using IoT & neural networks, 2017 International Conference on Smart Technologies For Smart Nation (SmartTechCon), 189–194. https://doi.org/10.1109/SmartTechCon.2017.8358367 doi: 10.1109/SmartTechCon.2017.8358367
[24]	Mitra P, Ray R, Chatterjee R, et al. (2016) Flood forecasting using Internet of things and artificial neural networks, 2016 IEEE 7th Annual Information Technology, Electronics and Mobile Communication Conference (IEMCON), 1–5. https://doi.org/10.1109/IEMCON.2016.7746363 doi: 10.1109/IEMCON.2016.7746363
[25]	Alquisola GLV, Coronel DJA, Reolope BMF, et al. (2018) Prediction and visualization of the disaster risks in the Philippines using discrete wavelet transform (DWT), autoregressive integrated moving average (ARIMA), and artificial neural network (ANN), 2018 3rd International Conference on Computer and Communication Systems (ICCCS), 146–149. https://doi.org/10.1109/CCOMS.2018.8463238
[26]	Wu J, Ming H, Xu J, (2021) Research on intelligent disaster prevention and mitigation method in high flood risk area of river basin based on artificial neural network, 2021 7th International Conference on Hydraulic and Civil Engineering & Smart Water Conservancy and Intelligent Disaster Reduction Forum (ICHCE & SWIDR), 943–952. https://doi.org/10.1109/ICHCESWIDR54323.2021.9656352 doi: 10.1109/ICHCESWIDR54323.2021.9656352

Reader Comments

Your name:*

Email:*
© 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)

通讯作者: 陈斌, bchen63@163.com

1.
沈阳化工大学材料科学与工程学院沈阳 110142

AIMS Environmental Science

1.6 2.9

Metrics

Article views(1411) PDF downloads(81) Cited by(2)

Preview PDF

Download XML

Export Citation

Article outline

Show full outline

Figures and Tables

Figures(5) / Tables(3)

AIMS Environmental Science

Monitoring river water quality through predictive modeling using artificial neural networks backpropagation

Related Papers:

Abstract

1. Introduction

2. Materials and method

2.1. Proposed generalized estimator

2.2. Bias and mean square error of generalized estimator

3. Proposed two-exponential general Shrinkage estimator

3.1. Theorem for estimating any population characteristic

4. Efficiency comparison

4.1. When is known and is unknown

4.2. If a is unknown λ, b is known

5. Numerical study

6. Conclusion

Acknowledgments

Conflict of interest

Appendix A

Appendix B

Appendix C

References

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Figures and Tables

Other Articles By Authors

Catalog

AIMS Environmental Science

Monitoring river water quality through predictive modeling using artificial neural networks backpropagation

Related Papers:

Abstract

1. Introduction

2. Materials and method

2.1. Proposed generalized estimator

2.2. Bias and mean square error of generalized estimator

3. Proposed two-exponential general Shrinkage estimator

3.1. Theorem for estimating any population characteristic

4. Efficiency comparison

4.1. When is known and is unknown

4.2. If a is unknown λ, b is known

5. Numerical study

6. Conclusion

Acknowledgments

Conflict of interest

Appendix A

Appendix B

Appendix C

References

Reader Comments

通讯作者: 陈斌, bchen63@163.com

Metrics

Figures and Tables

Other Articles By Authors

Related pages

Tools

Export File

Citation

Format

Content

Catalog