Loading [MathJax]/jax/output/SVG/jax.js
Research article Special Issues

Approach to COVID-19 time series data using deep learning and spectral analysis methods

  • This article focuses on the application of deep learning and spectral analysis to epidemiology time series data, which has recently piqued the interest of some researchers. The COVID-19 virus is still mutating, particularly the delta and omicron variants, which are known for their high level of contagiousness, but policymakers and governments are resolute in combating the pandemic's spread through a recent massive vaccination campaign of their population. We used extreme machine learning (ELM), multilayer perceptron (MLP), long short-term neural network (LSTM), gated recurrent unit (GRU), convolution neural network (CNN) and deep neural network (DNN) methods on time series data from the start of the pandemic in France, Russia, Turkey, India, United states of America (USA), Brazil and United Kingdom (UK) until September 3, 2021 to predict the daily new cases and daily deaths at different waves of the pandemic in countries considered while using root mean square error (RMSE) and relative root mean square error (rRMSE) to measure the performance of these methods. We used the spectral analysis method to convert time (days) to frequency in order to analyze the peaks of frequency and periodicity of the time series data. We also forecasted the future pandemic evolution by using ELM, MLP, and spectral analysis. Moreover, MLP achieved best performance for both daily new cases and deaths based on the evaluation metrics used. Furthermore, we discovered that errors for daily deaths are much lower than those for daily new cases. While the performance of models varies, prediction and forecasting during the period of vaccination and recent cases confirm the pandemic's prevalence level in the countries under consideration. Finally, some of the peaks observed in the time series data correspond with the proven pattern of weekly peaks that is unique to the COVID-19 time series data.

    Citation: Kayode Oshinubi, Augustina Amakor, Olumuyiwa James Peter, Mustapha Rachdi, Jacques Demongeot. Approach to COVID-19 time series data using deep learning and spectral analysis methods[J]. AIMS Bioengineering, 2022, 9(1): 1-21. doi: 10.3934/bioeng.2022001

    Related Papers:

    [1] Ashebir Dingeto Hailu . Ethiopia hydropower development and Nile basin hydro politics. AIMS Energy, 2022, 10(1): 87-101. doi: 10.3934/energy.2022006
    [2] Oying Doso, Sarsing Gao . An overview of small hydro power development in India. AIMS Energy, 2020, 8(5): 896-917. doi: 10.3934/energy.2020.5.896
    [3] Gustavo Henrique Romeu da Silva, Andreas Nascimento, Christoph Daniel Baum, Nazem Nascimento, Mauro Hugo Mathias, Mohd Amro . Renewable energy perspectives: Brazilian case study on green hydrogen production. AIMS Energy, 2025, 13(2): 449-470. doi: 10.3934/energy.2025017
    [4] Jerome Mungwe, Stefano Mandelli, Emanuela Colombo . Community pico and micro hydropower for rural electrification: experiences from the mountain regions of Cameroon. AIMS Energy, 2016, 4(1): 190-205. doi: 10.3934/energy.2016.1.190
    [5] Gul Ahmad Ludin, Mohammad Amin Amin, Assadullah Aminzay, Tomonobu Senjyu . Theoretical potential and utilization of renewable energy in Afghanistan. AIMS Energy, 2017, 5(1): 1-19. doi: 10.3934/energy.2017.1.1
    [6] HVV Priyadarshana, MA Kalhan Sandaru, KTMU Hemapala, WDAS Wijayapala . A review on Multi-Agent system based energy management systems for micro grids. AIMS Energy, 2019, 7(6): 924-943. doi: 10.3934/energy.2019.6.924
    [7] Botelho Anabela, Lourenço-Gomes Lina, M.C. Pinto Lígia, Sousa Patrícia, Sousa Sara, Valente Marieta . Using Choice Experiments to Assess Environmental Impacts of Dams in Portugal. AIMS Energy, 2015, 2(3): 316-325. doi: 10.3934/energy.2015.3.316
    [8] Tilahun Nigussie, Wondwossen Bogale, Feyisa Bekele, Edessa Dribssa . Feasibility study for power generation using off- grid energy system from micro hydro-PV-diesel generator-battery for rural area of Ethiopia: The case of Melkey Hera village, Western Ethiopia. AIMS Energy, 2017, 5(4): 667-690. doi: 10.3934/energy.2017.4.667
    [9] Géremi Gilson Dranka, Jorge Cunha, José Donizetti de Lima, Paula Ferreira . Economic evaluation methodologies for renewable energy projects. AIMS Energy, 2020, 8(2): 339-364. doi: 10.3934/energy.2020.2.339
    [10] Ashebir Dingeto Hailu, Desta Kalbessa Kumsa . Ethiopia renewable energy potentials and current state. AIMS Energy, 2021, 9(1): 1-14. doi: 10.3934/energy.2021001
  • This article focuses on the application of deep learning and spectral analysis to epidemiology time series data, which has recently piqued the interest of some researchers. The COVID-19 virus is still mutating, particularly the delta and omicron variants, which are known for their high level of contagiousness, but policymakers and governments are resolute in combating the pandemic's spread through a recent massive vaccination campaign of their population. We used extreme machine learning (ELM), multilayer perceptron (MLP), long short-term neural network (LSTM), gated recurrent unit (GRU), convolution neural network (CNN) and deep neural network (DNN) methods on time series data from the start of the pandemic in France, Russia, Turkey, India, United states of America (USA), Brazil and United Kingdom (UK) until September 3, 2021 to predict the daily new cases and daily deaths at different waves of the pandemic in countries considered while using root mean square error (RMSE) and relative root mean square error (rRMSE) to measure the performance of these methods. We used the spectral analysis method to convert time (days) to frequency in order to analyze the peaks of frequency and periodicity of the time series data. We also forecasted the future pandemic evolution by using ELM, MLP, and spectral analysis. Moreover, MLP achieved best performance for both daily new cases and deaths based on the evaluation metrics used. Furthermore, we discovered that errors for daily deaths are much lower than those for daily new cases. While the performance of models varies, prediction and forecasting during the period of vaccination and recent cases confirm the pandemic's prevalence level in the countries under consideration. Finally, some of the peaks observed in the time series data correspond with the proven pattern of weekly peaks that is unique to the COVID-19 time series data.



    Spinors are structures used in many fields of applied sciences. In physics, the theory of spinors is especially used in applications electron spin and theory of relativity in quantum mechanics. The wave function of a particle whose spin is ½ is called spinor. In addition to that, the application of spinors in electromagnetic theory is very important. The construction to obtain a spin structure on a 4-dimensional space-time are extended to obtain spinors in physics, such as the Dirac spinor. The Dirac representation of the column vectors in Heisenberg representation corresponds to Ket|ψ=(ψ1ψ2). Here, ψ1 and ψ2 are complex numbers [7,16,24]. The Pauli spin matrices mentioned here are with dimension 2x2, and the two-component complex column matrices acting on these matrices are spinors. Spinors can be evaluated as concrete objects when choosing Cartesian coordinates. For example, in three dimensional Euclidean space, spinors can be generated by choosing Pauli spin matrices corresponding to the three coordinate axes.

    In quantum theory, spin is represented by Pauli matrices. Also, Pauli matrices are a representation of Clifford spin algebra and therefore all properties of Pauli matrices derive from basic algebra. One of the most important studies giving the relationship between Clifford algebra and spinors is [23]. Clifford algebras constitute a highly intuitive formalism, having an intimate relationship to quantum field theory. Among the existing approaches to Clifford algebras and spinors this book is unique in that it provides a didactical presentation of the topic. Also, the first person to study spinors in geometric terms is Cartan, one of the founders of Clifford algebra. Cartan studied spinors in a geometric sense [4]. Cartan's study gives the spinor representation of the basic geometric definitions. So, it is a very impressive reference in terms of the geometry of the spinors. In Cartan's study, it was emphasized that the set of isotropic vectors in the vector space C3 forms a two-dimensional surface in the space C2. It is seen that each isotropic vector in the complex vector space C3 corresponds to two vectors in the space C2. Conversely; these vectors in space C2 correspond to the same isotropic vector. Cartan expressed that these complex vectors with two-dimensional [4].

    In 1927, when Pauli introduced his spin matrices, he first applied the spinors to mathematical physics [20]. After that, the relationship between the spinors and the Lorentz group was demonstrated by Dirac, and the fully relativistic theory of electron spin was discovered [8]. In 1930, Juvet and Sauter represented the spinor space as the left ideal of a matrix algebra [13,18,21]. In further detail, instead of taking the spinors as complex valued two-dimensional column vectors, as Pauli did, if the spinors are taken as complex-valued 2x2-dimensional matrices where the left column of the elements are not zero, then the spinor space becomes a minimal left ideal in Mat (2, C) [17]. Later, in differential geometry the spinor representation of curve theory was expressed by Torres del Castillo and Barrales [6]. That study is one of the basic references of our study. Then, the spinor formulation of the Darboux frame on a directed surface and Bishop frame of curves in E3 were given in [15,22]. After that, the hyperbolic spinor corresponding to the Frenet frame of curves in Minkowski space E31 was expressed by Ketenci et al. [14]. Moreover, Erişir et. al obtained the spinor formulation of the alternative frame of a curve in Minkowski space, and the spinor formulation of the relationship between Frenet frame and Bishop frame [10]. Thus, Balcı et al. gave the hyperbolic spinor formulation of the Darboux frame in E31 [1]. Finally, Erişir and Kardağ have shown a new representation of the Involute Evolute curves in E3 with the help of spinors [11].

    In this study, answer to the following question has been sought: "What is the geometric construction of spinors for Bertrand curves?". For this, firstly, the spinor characterization of unit speed and not unit speed curve is calculated. Then, Bertrand curves have been considered and these curves have been represented by spinors in the space C2. Later, considering the relations between Bertrand curves, the spinor relations corresponding to these curves have been obtained. After that, the answer of question "If the angle between the tangent vectors of the Bertrand curves is θ, what is the angle between the spinors corresponding to these curves?" has been found. In this way, a different geometric construction of the spinors has been shown.

    A curve is considered as differentiable at the each point of an open interval. Thus, one can construct a set of mutually orthogonal unit vectors on this curve. These vectors are called tangent, normal and binormal unit vectors or the Serret-Frenet frame, collectively. So, let us consider that the regular curve (α) which is differentiable function so that α:IE3, (IR) has the arbitrary parameter t. Moreover, for tI the Frenet vectors on the point α(t) of the curve (α) are given by {T(t),N(t),B(t)}. So, these Frenet vectors are obtained by the equations T(t)=1α(t)α(t), N(t)=B(t)×T(t) and B(t)=1α(t)×α(t)(α(t)×α(t)) where "" is the derivative with respect to arbitrary parameter t, κ and τ are the curvature and torsion of this curve (α) [12]. Moreover, the Frenet derivative formulas of this curve (α) are given by T=ακN, N=α(κT+τB), B=ατN [12]. Consider that the parameter of the curve (α) is the arc-length parameter s. So, if there is the equation ˙α(s)=1 on the point sI of the curve α:IE3, (IR), the curve (α) is called a curve parameterized by the arc-length or a unit speed curve. Thus, the Frenet vectors of the curve (α) parameterized by arc-length can be obtained by T(s)=˙α(s), N(s)=1¨α(s)¨α(s) and B(s)=T(s)×N(s) where "." is the derivative with respect to arc-length parameter s. Moreover, the Frenet formulas of this curve are as ˙T=κN,˙N=κT+τB,˙B=τN [12].

    French mathematician Saint-Venant raised the question in 1845 on whether the principal normal vectors of a curve α on a ruled surface produced by the principal normal vectors of a curve α in three dimensional Euclidean space can be the same as the principal normal vectors of a second curve. This question was answered by an article published by Bertrand in 1850. For this reason, the curves with this feature started to be called Bertrand curves after that study. Bertrand expressed that to have a second curve with this property, there is a linear relationship between the curvature and torsion of the original curve. These curve pairs, which are a remarkable subject for scientists, have been treated in many studies [2,3,5,9,12,19]. So, for the Bertrand curves, the following definition and theorems can be given.

    Definition 2.1. Let α,β:IE3 be two curves and the Frenet vector fields of these curves be {T,N,B} and {T,N,B}, respectively. If the vector pair {N,N} is linearly dependent, this pair (α,β) is called Bertrand curves [12].

    Theorem 2.2. Let the curves α,β:IE3 be Bertrand curves. Then, the distance between mutual points of these curves is constant [12].

    Theorem 2.3. If the curves α,β:IE3 are considered Bertrand curves then the angle between the tangent vectors at mutual points of these curves is constant [12].

    Theorem 2.4. Let the Frenet frames of the Bertrand curves (α,β) be {T,N,B} and {T,N,B}, respectively. So, especially if N=N is chosen, the relationship between these Frenet frames is

    [TNB]=[cosθ0sinθ010sinθ0cosθ][TNB] (2.1)

    where θ is the angle between the tangent vectors of Bertrand curves (α,β) [12].

    Spinors form a vector space equipped with a linear group representation of the spin group. Here this vector space is usually over the complex numbers. Cartan [4] introduced the spinors geometrically over the complex numbers as follows. Let the vector x=(x1,x2,x3)C3 be an isotropic vector (x12+x22+x32=0) and the space C3 be the 3-dimensional complex vector space. A two-dimensional surface in the space C2 is formed by the set of isotropic vectors in the vector space C3. If this two-dimensional surface is parameterized by coordinates γ1 and γ2, then x1=γ12γ22, x2=i(γ12+γ22), x3=2γ1γ2 is obtained. So, the equations γ1=±x1ix22, γ2=±x1ix22 are obtained. Each isotropic vector in the complex vector space C3 corresponds to two vectors, (γ1,γ2) and (γ1,γ2) in the space C2. Conversely; both vectors so given in space C2 correspond to the same isotropic vector x. Moreover, Cartan expressed that the two-dimensional complex vectors γ=(γ1,γ2) described in this way are called spinors

    γ=(γ1γ2).

    Using Cartan's study [4], Torres del Castillo and Barrales [6] matched the isotropic vector a+ibC3 with spinor γ=(γ1,γ2) where a,bR3. So, considering the Pauli matrices they created the matrices σ and gave

    a+ib=γtσγ,c=ˆγtσγ (2.2)

    where a+ib is the isotropic vector in the space C3, cR3 and the mate ˆγ of the spinor γ is

    ˆγ=(0110)¯γ=(0110)(¯γ1¯γ2)=(¯γ2¯γ1).

    When all necessary operations are considered, it is seen that the vectors a, b and c have the same length a=b=c=¯γtγ and these vectors are mutually orthogonal. The correspondence between spinors and orthogonal bases given by equation (2.2) is two to one; the spinors γ and γ correspond to the same ordered orthonormal basis {a,b,c}, with a×b,c>0. It is important to notice that the ordered triads {a,b,c}, {b,c,a} and {c,a,b} correspond to different spinors [6]. Moreover, the following proposition can be given.

    Proposition 2.5. Let two arbitrary spinors be γ and ψ. Then, the following statements hold;

    i)¯ψtσγ=ˆψtσˆγii)^λψ+μγ=¯λˆψ+¯μˆγiii)ˆˆγ=γiv)ψtσγ=γtσψ

    where λ,μC [6].

    Now, let a curve parameterized by arc-length be α:IE3, (IR). So, α(s)=1 where s is the arc-length parameter of the curve (α). Moreover, consider the Frenet vectors of this curve as {N,B,T} and the spinor ζ represents the Frenet vector triad {N,B,T}. Thus, from Eq (2.2) the following equations can be obtained

    N+iB=ζtσζ=(ζ12ζ22,i(ζ12+ζ22),2ζ1ζ2),T=ˆζtσζ=(ζ1¯ζ2+¯ζ1ζ2,i(ζ1¯ζ2¯ζ1ζ2),|ζ1|2|ζ2|2) (2.3)

    with ¯ζtζ=1 [6]. Moreover, the following theorem can be given.

    Theorem 2.6. If the spinor ζ with two complex components represents the triad {N,B,T} of a curve (α) parameterized by its arc-length s, the Frenet equations are equivalent to the single spinor equation

    dζds=12(iτζ+κˆζ)

    where κ and τ denote the curvature and torsion of the curve (α), respectively [6].

    Before I get to the main section, I have to emphasize that: While Torres del Castillo and Barrales matched a spinor with Frenet frame {N,B,T} of an unit speed curve α:IE3, they separated Frenet vectors as N+iB and T where N+iB is an isotropic vector. Unlike Torres del Castillo and Barrales, in this paper, the Frenet frame {B,T,N} corresponds to a different spinor and this Frenet frame is separated as B+iT and N where B+iT is an isotropic vector.

    Now let the Frenet vectors on the point α(s) of the regular and unit speed curve α:IRE3 be {B,T,N}. Moreover, the Frenet vectors {B,T,N} correspond to a spinor ξ. So, the spinor equations of this Frenet frame can be written by

    B+iT=ξtσξ (3.1)

    and

    N=ˆξtσξ (3.2)

    where ˉξtξ=1. So, the following theorems can be given.

    Theorem 3.1. Let ξ be the spinor corresponding to the Frenet vectors {B,T,N} of the unit speed and regular curve α:IE3. So, the Frenet equations are equivalent to the single spinor equation

    dξds=(τiκ)2ˆξ (3.3)

    where τ and κ denote the torsion and curvature of the curve (α) and ˆξ is the mate of spinor ξ, respectively.

    Proof. If the derivative of the Eq (3.1) with respect to arc-length parameter s of the curve (α) is considered, one can write

    d(B+iT)ds=dξdstσξ+ξtσdξds.

    So, since {ξ,ˆξ} is a basis for spinors with two complex components in the space C2, the spinor dξds can be written dξds=fξ+gˆξ where f and g are arbitrary complex valued functions and ˆξ is the mate of spinor ξ. If Proposition 2.5 is considered, one can obtain f=0,g=τiκ2. Thus, dξds=(τiκ)2ˆξ.

    Now, in addition to reference [6], I give the components of the vector B+iT, separately, with the following theorem.

    Theorem 3.2. Let a spinor ξ represent the Frenet frame {B,T,N} of the regular and unit speed curve (α). So, the spinor equations of Frenet vectors T and B from the complex vector B+iT can be written as

    T=i2(ξtσξ+ˆξtσˆξ),B=12(ξtσξˆξtσˆξ).

    Proof. Let a spinor ξ represent the Frenet vectors {B,T,N} of the curve (α). If Eq (3.1) is considered, B=Re(ξtσξ) and T=Im(ξtσξ) hold. Moreover, from Proposition 2.5 one can write T=i2(ξtσξ+ˆξtσˆξ), B=12(ξtσξˆξtσˆξ).

    Corollary 3.3. If the spinor ξ represents the Frenet frame {B,T,N} of the curve (α), then the spinor components of Frenet vectors can be written as

    T=i2(ξ12ξ22+¯ξ22¯ξ12,i(ξ12+ξ22+¯ξ12+¯ξ22),2(¯ξ1¯ξ2ξ1ξ2)),N=(ξ1¯ξ2+¯ξ1ξ2,i(ξ1¯ξ2¯ξ1ξ2),|ξ1|2|ξ2|2),B=12(ξ12ξ22+¯ξ12¯ξ22,i(ξ12+ξ22¯ξ12¯ξ22),2(ξ1ξ2+¯ξ1¯ξ2)).

    On the other hand, let any curve (β) which is not parameterized by arc length be considered and the Frenet vectors of this curve be {B,T,N}. Moreover, let the different spinor ϕ represent the Frenet vectors of the curve (β). So, similar to the Eqs (3.1) and (3.2), one can write

    B+iT=ϕtσϕ,N=ˆϕtσϕ.

    Thus, the spinor equations of the Frenet vectors of the curve (β) similar to Corollary 3.3 can be written by components as

    T=i2(ϕ12ϕ22+¯ϕ22¯ϕ12,i(ϕ12+ϕ22+¯ϕ12+¯ϕ22),2(¯ϕ1¯ϕ2ϕ1ϕ2)),N=(ϕ1¯ϕ2+¯ϕ1ϕ2,i(ϕ1¯ϕ2¯ϕ1ϕ2),|ϕ1|2|ϕ2|2),B=12(ϕ12ϕ22+¯ϕ12¯ϕ22,i(ϕ12+ϕ22¯ϕ12¯ϕ22),2(ϕ1ϕ2+¯ϕ1¯ϕ2)).

    So, one can give the following theorem without proof. The proof of this theorem can be easily done similar to Theorem 3.1.

    Theorem 3.4. Let the Frenet vectors of the curve (β) which is not parameterized by arc length be {B,T,N} and the spinor corresponding to this curve be ϕ. So, the Frenet equation of this curve in terms of a single spinor equation is written as

    dϕdt=dβdt(τiκ)2ˆϕ

    where t is any arbitrary parameter of the curve (β).

    Here, if s is considered arc-length parameter of the curve (β), dβdt=dsdt holds. So, one can write

    dϕds=(τiκ)2ˆϕ.

    Now, the curves α and β mentioned above are considered Bertrand curves. Let the spinors ξ and ϕ represent the Bertrand curves α and β, respectively. Moreover, the Frenet vectors of the curve α are {B,T,N} and the Frenet vectors of the curve β are {B,T,N}. So, the geometric interpretation of angles between spinors can be done with the following theorems.

    Theorem 3.5. Let the curves α,β:IE3 be Bertrand curves and the curves α and β correspond to the spinors ξ and ϕ, respectively. Moreover, the Frenet vectors of the curves α and β shall be {B,T,N} and {B,T,N}, respectively. So, the relationship between spinors of corresponding Bertrand curves is

    ϕ=±eiθ2ξ. (3.4)

    Proof. Let the spinors ξ and ϕ represent the Frenet vectors {B,T,N} and {B,T,N} of Bertrand curves α,β:IE3, respectively. So, if the relationship between Bertrand curves in the Eq (2.1) is used, one can obtain

    B+iT=sinθT+cosθB+icosθTisinθB=(cosθisinθ)(B+iT)=eiθ(B+iT)

    where θ is the angle between the tangent vectors of the Bertrand curves (α,β). From here it appears that the angle between the B+iT and B+iT vectors is also θ. So, one can write

    ϕtσϕ=eiθ(ξtσξ)

    and finally

    ϕ1=±eiθ2ξ1,ϕ2=±eiθ2ξ2.

    It is known from Section 2.2 that the spinors γ and γ correspond to the same ordered orthonormal basis. Namely, the spinors ϕ and ϕ correspond to the vector B+iT, while the spinors ξ and ξ correspond to the vector B+iT. So, one can write

    ϕ=±eiθ2ξ.

    Theorem 3.6. (Main Theorem): Let the spinors ξ and ϕ represent the Bertrand curves α,β:IE3, respectively. So, if the angle between the tangent vectors of these curves is θ, then the angle between the spinors ξ and ϕ is θ2.

    The following corollaries of the main theorem can be given.

    Corollary 3.7. Let the spinors ξ and ϕ represent the Bertrand curves α,β:IE3, respectively. Then, the spinor ξ returns to the spinor ϕ, while the vector ˆξ makes a reverse rotation to the vector ˆϕ. Moreover, the rotation angle is also θ2,

    ˆϕ=±eiθ2ˆξ.

    Corollary 3.8. Let the curves α,β:IE3 be Bertrand curves and the curves α and β correspond to the spinors ξ and ϕ, respectively. Thus, the angle between the derivatives of spinors ξ and ϕ is also θ2.

    Corollary 3.9. Let the curves (α,β) be Bertrand curves and the spinors ξ and ϕ represent these curves, respectively. Thus, the derivative of spinor ϕ which represent the curve (β) with respect to the curvatures of curve (α) can be rewritten as

    dϕdt=eiθ(τiκ2)ˆϕ.

    Spinors can be considered as concrete objects using a set of Cartesian coordinates. For example, spinors can be formed from Pauli spin matrices corresponding to coordinate axes in three-dimensional Euclidean space. By using matrices with these 2x2 dimensional complex elements, spinors, which are column matrices, are created. In this case, the spin group is isomorphic to the group of 2x2 unitary matrices with determinant one, which naturally sits inside the matrix algebra. This group acts by conjugation on the real vector space spanned by the Pauli matrices themselves, realizing it as a group of rotations among them, but it also acts on the column vectors (that is, the spinors). So, in this study, a different geometric interpretation of spinors has been created by using Pauli matrices and spinors. A different (spinor) representation of Bertrand curves, one of the most important curve pairs in differential geometry, has been obtained. Thus, this study in geometry brings a new perspective as an interesting geometric interpretation of spinors used in many fields of science.

    The author declares no conflict of interest.


    Acknowledgments



    The authors would like to thank the Petroleum Technology Development Fund (PTDF) Nigeria doctoral fellowship in collaboration with the Campus France Africa Unit. The authors appreciate the helpful comments and suggestions of the five anonymous reviewers, which have contributed to the overall improvement of the article.

    Conflict of interest



    The authors state that they have no conflicts of interest.

    [1] Seligmann H, Iggui S, Rachdi M, et al. (2020) Inverted covariate effects for mutated 2nd vs 1st wave Covid-19: high temperature spread biased for young. Medrxiv ppmedrxiv-20151878.
    [2] Demongeot J, Seligmann H (2020) SARS-CoV-2 and miRNA-like inhibition power. Med Hypotheses 144: 110245https://doi.org/10.1016/j.mehy.2020.110245.
    [3] Demongeot J, Griette Q, Magal P (2020) SI epidemic model applied to COVID-19 data in mainland China. Roy Soc Open Sci 7: 201878https://doi.org/10.1098/rsos.201878.
    [4] Soubeyrand S, Demongeot J, Roques L (2020) Towards unified and real-time analyses of outbreaks at country-level during pandemics. One Health 11: 100187https://doi.org/10.1016/j.onehlt.2020.100187.
    [5] Gaudart J, Landier J, Huiart L, et al. (2021) Factors associated with spatial heterogeneity of Covid-19 in France: a nationwide ecological study. The Lancet Public Health 6: e222-e231. https://doi.org/10.1016/S2468-2667(21)00006-2.
    [6] Oshinubi K, Al-Awadhi F, Rachdi M, et al. (2021) Data analysis and forecasting of COVID-19 pandemic in Kuwait based on daily observation and basic reproduction number dynamics. Kuwait J Sci Special Issue: 1-30. https://doi.org/10.48129/kjs.splcov.14501.
    [7] Oshinubi K, Ibrahim F, Rachdi M, et al. Functional data analysis: Transition from daily observation of COVID-19 prevalence in France to functional curves (2021) .https://doi.org/10.1101/2021.09.25.21264106.
    [8] Demongeot J, Oshinubi K, Rachdi M, et al. (2022) The application of ARIMA model to analyse incidence pattern in several countries. J Math Comput Sci 12: 10https://doi.org/10.28919/jmcs/6541.
    [9] Demongeot J, Oshinubi K, Seligmann H, et al. Estimation of daily reproduction rates in COVID-19 outbreak (2021) .https://doi.org/10.1101/2020.12.30.20249010.
    [10] Griette Q, Demongeot J, Magal P (2021) A robust phenomenological approach to investigate COVID-19 data for France. Math Appl Sci Eng 3: 149-218. https://doi.org/10.5206/mase/14031.
    [11] Worldometers (2021) .Available from: https://www.worldometers.info/coronavirus/.
    [12] Ahmed HM, Elbarkouky RA, Omar OAM, et al. (2021) Models for COVID-19 Daily confirmed cases in different countries. Mathematics 9: 659https://doi.org/10.3390/math9060659.
    [13] Tojanovic J, Boucher VG, Boyle J, et al. COVID-19 is not the flu: Four graphs from four countries (2021) .https://doi.org/10.3389/fpubh.2021.628479.
    [14] Bakhta A, Boiveau T, Maday Y, et al. (2021) Epidemiological forecasting with model reduction of compartmental models: application to the COVID-19 pandemic. Biology 10: 22https://doi.org/10.3390/biology10010022.
    [15] Abioye AI, Umoh MD, Peter OJ, et al. (2021) Forecasting of COVID-19 pandemic in Nigeria using real statistical data. Commun Math Biol Neurosci 2021: 2https://doi.org/10.28919/cmbn/5144.
    [16] Oshinubi K, Rachdi M, Demongeot J (2021) Analysis of daily reproduction rates of COVID-19 using current health expenditure as gross domestic product percentage (CHE/GDP) across countries. Healthcare 9: 1247https://doi.org/10.3390/healthcare9101247.
    [17] Deb S, Majumdar M (2020) A time series method to analyze incidence pattern and estimate reproduction number of COVID-19. ArXiv 2003.10655.
    [18] Hochreiter S, Schmidhuber J (1997) Long short-term memory. Neural Comput 9: 1735-1780. https://doi.org/10.1162/neco.1997.9.8.1735.
    [19] Chung J, Gulcehre C, Cho K, et al. (2014) Empirical evaluation of gated recurrent neural networks on sequence modeling. ArXiv 1412.3555.
    [20] Chimmula VKR, Zhang L (2020) Time series forecasting of COVID-19 transmission in Canada using LSTM networks. Chaos Soliton Fract 135: 109864https://doi.org/10.1016/j.chaos.2020.109864.
    [21] Yahia NB, Kandara MD, Saoud NBB Deep ensemble learning method to forecast COVID-19 outbreak (2020) .https://doi.org/10.21203/rs.3.rs-27216/v1.
    [22] Yang Z, Zeng Z, Wang K, et al. (2020) Modified SEIR and AI prediction of the epidemic trend of COVID-19 in China under public health interventions. J Thorac Dis 12: 165-174. http://dx.doi.org/10.21037/jtd.2020.02.64.
    [23] Cho K, Van Merriënboer B, Gulcehre C, et al. (2014) Learning phrase representations using RNN encoder-decoder for statistical machine translation. ArXiv 1406.1078.
    [24] Hochreiter S, Schmidhuber J (1996) LSTM can solve hard long time lag problems. Adv Neural Inf Proces Syst 9: 473-479.
    [25] Ren Y, Chen H, Han Y, et al. (2020) A hybrid integrated deep learning model for the prediction of citywide spatio-temporal flow volumes. Int J Geogr Inf Sci 34: 802-823. https://doi.org/10.1080/13658816.2019.1652303.
    [26] Zhang Y, Cheng T, Ren Y, et al. (2020) A novel residual graph convolution deep learning model for short-term network-based traffic forecasting. Int J Geogr Inf Sci 34: 969-995. https://doi.org/10.1080/13658816.2019.1697879.
    [27] Chollet F, Allaire JJ (2018)  Deep Learning with R New York: Manning Publications.
    [28] Liu YH, Maldonado P (2018)  R Deep Learning Projects: Master the Techniques to Design and Develop Neural Network Models in R UK: Packt Publishing.
    [29] Ma X, Dai Z, He Z, et al. (2017) Learning traffic as images: A deep convolutional neural network for large-scale transportation network speed prediction. Sensors 17: 818https://doi.org/10.3390/s17040818.
    [30] Jeong MH, Lee TY, Jeon S-B, et al. (2021) Highway speed prediction using gated recurrent unit neural networks. Appl Sci 11: 3059https://doi.org/10.3390/app11073059.
    [31] Omran NF, Abd-el Ghany SF, Saleh H, et al. (2021) Applying deep learning methods on time-series data for forecasting COVID-19 in Egypt, Kuwait and Saudi Arabia. Complexity 2021: 6686745https://doi.org/10.1155/2021/6686745.
    [32] Chae S, Kwon S, Lee D (2018) Predicting infectious disease using deep learning and big data. Int J Environ Res Public Health 15: 1596https://doi.org/10.3390/ijerph15081596.
    [33] Frank RJ, Davey N, Hunt SP (2001) Time series prediction and neural networks. J Intell Robot Syst 31: 91-103. https://doi.org/10.1023/A:1012074215150.
    [34] Gu J, Wang J, Kuen J, et al. (2017) Recent advances in convolutional neural networks. ArXiv 1512.07108.
    [35] Huang CJ, Chen Y-H, Ma Y, et al. Multiple-Input deep convolutional neural network model for COVID-19 forecasting in China (2020) .https://doi.org/10.1101/2020.03.23.20041608.
    [36] Miotto R, Wang R, Wang S, et al. (2018) Deep learning for healthcare: review, opportunities and challenges. Briengs Bioinformatics 19: 1236-1246. https://doi.org/10.1093/bib/bbx044.
    [37] Pascanu R, Gulcehre C, Cho K, et al. (2014) How to construct deep recurrent neural networks. ArXiv 1312.6026.
    [38] Ravi D, Wong D, Deligianni F, et al. (2017) Deep learning for health informatics. IEEE J Biomed Health 21: 4-21. https://doi.org/10.1109/JBHI.2016.2636665.
    [39] Priestley MB (1981)  Spectral Analysis and Time Series, volume 1 of Probability and mathematical statistics: A series of monographs New York: Academic Press.
    [40] Priestley MB (1981)  Spectral Analysis and Time Series, volume 2 of Probability and mathematical statistics: A series of monographs New York: Academic Press.
    [41] Parker RL, O'Brien MS (1997) Spectral analysis of vector magnetic field profiles. J Geophys Res 102: 24815-24824. https://doi.org/10.1029/97JB02130.
    [42] Percival D, Walden A (1993)  Spectral Analysis for Physical Applications Cambridge: Cambridge University Press.
    [43] Prieto GA, Parker RL, Thomson DJ, et al. (2007) Reducing the bias of multitaper spectrum estimates. Geophys J Int 171: 1269-1281. https://doi.org/10.1111/j.1365-246X.2007.03592.x.
    [44] Thomson DJ (1982) Spectrum estimation and harmonic analysis. Proc IEEE 70: 1055-1096. https://doi.org/10.1109/PROC.1982.12433.
    [45] Rahim KJ, Burr WS, Thomson DJ (2014)  Applications of Multitaper Spectral Analysis to Nonstationary Data [PhD thesis] Canada: Queen's University.
    [46] Ord K, Fildes R, Kourentzes N (2017)  Principles of business forecasting New York: Wessex Press Publishing.
    [47] Kourentzes N, Barrow BK, Crone SF (2014) Neural network ensemble operators for time series forecasting. Expert Syst Appl 41: 4235-4244. https://doi.org/10.1016/j.eswa.2013.12.011.
    [48] Crone SF, Kourentzes N (2010) Feature selection for time series prediction – A combined filter and wrapper approach for neural networks. Neurocomputing 73: 1923-1936. https://doi.org/10.1016/j.neucom.2010.01.017.
    [49] Barrow D, Kourentzes N (2018) The impact of special days in call arrivals forecasting: A neural network approach to modelling special days. Eur J Oper Res 264: 967-977. https://doi.org/10.1016/j.ejor.2016.07.015.
    [50] Golyandina N, Korobeynikov A (2014) Basic singular spectrum analysis and forecasting with R. Comput Stat Data Anal 71: 934-954. https://doi.org/10.1016/j.csda.2013.04.009.
    [51] Zhang Z, Moore J (2011) New significance test methods for Fourier analysis of geophysical time series. Nonlin Processes Geophys 18: 643-652. https://doi.org/10.5194/npg-18-643-2011.
    [52] Shorten C, Khoshgoftaar TM, Furht B (2021) Deep Learning applications for COVID-19. J Big Data 8: 18https://doi.org/10.1186/s40537-020-00392-9.
    [53] Bergman A, Sella Y, Agre P, et al. (2020) Oscillations in U.S. COVID-19 incidence and mortality data reflect diagnostic and reporting factors. mSystems 5: e00544-20https://doi.org/10.1128/mSystems.00544-20.
    [54] Frescura FAM, Engelbrecht CA, Frank BS (2007) Significance tests for periodogram peaks. ArXiv 0706.2225.
    [55] Grzesica D, Wiecek P (2016) Advanced forecasting methods based on spectral analysis. Procedia Engineering 161: 253-258. https://doi.org/10.1016/j.proeng.2016.08.546.
    [56] Kalantari M (2021) Forecasting COVID-19 pandemic using optimal singular spectrum analysis. Chaos Soliton Fract 142: 110547https://doi.org/10.1016/j.chaos.2020.110547.
    [57] Castillo O, Melin PA (2021) Novel method for a COVID-19 classification of countries based on an intelligent fuzzy fractal approach. Healthcare 9: 196https://doi.org/10.3390/healthcare9020196.
    [58] Castillo O, Melin PA (2021) A new fuzzy fractal control approach of non-linear dynamic systems: The case of controlling the COVID-19 pandemics. Chaos Soliton Fract 151: 111250https://doi.org/10.1016/j.chaos.2021.111250.
    [59] Sun TZ, Wang Y (2020) Modeling COVID-19 epidemic in Heilongjiang province, China. Chaos Soliton Fract 138: 109949https://doi.org/10.1016/j.chaos.2020.109949.
    [60] Boccaletti S, Ditto W, Mindlin G, et al. (2020) Modeling and forecasting of epidemic spreading: the case of Covid-19 and beyond. Chaos Soliton Fract 135: 109794https://doi.org/10.1016/j.chaos.2020.109794.
    [61] Castillo O, Melin P (2020) Forecasting of COVID-19 time series for countries in the world based on a hybrid approach combining the fractal dimension and fuzzy logic. Chaos Soliton Fract 140: 110242https://doi.org/10.1016/j.chaos.2020.110242.
    [62] Mansour RF, Escorcia-Gutierrez J, Gamarra M, et al. (2021) Unsupervised deep learning based variation antoencoder model for COVID-19 diagnosis and classification. Pattern Recogn lett 151: 267-274. https://doi.org/10.1016/j.patrec.2021.08.018.
    [63] Jaiswal AK, Tiwari P, Kumar S, et al. (2019) Identifying pneumonia in chest X-rays: a deep learning approach. Measurement 145: 511-518. https://doi.org/10.1016/j.measurement.2019.05.076.
    [64] Wikipedia (2021) .Available from: https://www.google.com/search?q=neural+network+picture.
    [65] Ahmadian S, Jalali SMJ, Islam SMS, et al. (2021) A novel deep neuroevolution-based image classification method to diagnose coronavirus disease (COVID-19). Comput Biol Med 139: 104994https://doi.org/10.1016/j.compbiomed.2021.104994.
    [66] Jalali SMJ, Ahmadian M, Ahmadian S, et al. (2021) An oppositional-Cauchy based GSK evolutionary algorithm with a novel deep ensemble reinforcement learning strategy for COVID-19 diagnosis. Appl Soft Comput 111: 107675https://doi.org/10.1016/j.asoc.2021.107675.
    [67] Melin P, Monica JC, Sanchez D, et al. (2020) Multiple ensemble neural network models with fuzzy response aggregation for predicting COVID-19 time series: the case of Mexico. Healthcare 8: 181https://doi.org/10.3390/healthcare8020181.
    [68] Magal P, Seydi O, Webb G, et al. (2021) A model of vaccination for dengue in the Philippines 2016–2018. Front Appl Math Stat 7: 760259https://doi.org/10.3389/fams.2021.760259.
    [69] Oshinubi K, Rachdi M, Demongeot J Modelling of COVID-19 pandemic vis-à-vis some socio-economic factors (2021) .https://doi.org/10.1101/2021.09.30.21264356.
  • This article has been cited by:

    1. Willis Awandu, Robin Ruff, Jens-Uwe Wiesemann, Boris Lehmann, Status of Micro-Hydrokinetic River Technology Turbines Application for Rural Electrification in Africa, 2022, 15, 1996-1073, 9004, 10.3390/en15239004
    2. Richard Cartland, Al-Mas Sendegeya, Jean de Dieu Khan Hakizimana, Performance Analysis of a Hybrid of Solar Photovoltaic, Genset, and Hydro of a Rural-Based Power Mini-Grid: Case Study of Kisiizi Hydro Power Mini-Grid, Uganda, 2023, 11, 2227-9717, 175, 10.3390/pr11010175
    3. A.O. Onokwai, I.P. Okokpujie, M.O. Ibiwoye, H.I. Owamah, G.C. Ayuba, J.O. Dirisu, Effect of thermal and flow properties on the performance of Jebba Hydro-Power Plant, Jebba, Nigeria, 2022, 65, 22147853, 2245, 10.1016/j.matpr.2022.06.342
    4. Kevin Lo, Benard Kibalya, Electric cooperatives and the political economy of rural electrification in Africa: Insights from Uganda, 2023, 36, 10406190, 107238, 10.1016/j.tej.2023.107238
    5. Mamdouh El Haj Assad, Siamak Hoseinzadeh, Editorial to the 'Special Issue-Analyzing energy storage systems for the applications of renewable energy sources' of AIMS Energy, 2022, 10, 2333-8334, 1074, 10.3934/energy.2022050
    6. Brendah Akankunda, Muyiwa S. Adaramola, Stephen K. Nkundabanyanga, Twaha Kaawaase Kigongo, Nicholas Mukisa, Techno-economic performance assessment of selected power plants in Uganda, 2024, 43, 1478-6451, 10.1080/14786451.2024.2350980
    7. Petras Punys, Gitana Vyčienė, Linas Jurevičius, Algis Kvaraciejus, Small Hydropower Assessment of Uganda Based on Multisource Geospatial Data, 2023, 15, 2073-4441, 2051, 10.3390/w15112051
    8. Tarek Ibrahim, Farouk Hachem, Mohamad Ramadan, Jalal Faraj, Georges El Achkar, Mahmoud Khaled, Cooling PV panels by free and forced convections: Experiments and comparative study, 2023, 11, 2333-8334, 774, 10.3934/energy.2023038
    9. Petras Punys, Linas Jurevičius, Andrius Balčiūnas, HYPOSO Map Viewer: A Web-Based Atlas of Small-Scale Hydropower for Selected African and Latin American Countries, 2024, 16, 2073-4441, 1276, 10.3390/w16091276
    10. Sherif Okewale Okesiji, Adegboyega Matthew Olaniyi, Victor Ogbonnaya Okorie, Innovations in Hydroelectric Power for Sustainable Development in Africa, 2025, 2692-2924, 10.1089/scc.2025.0003
    11. Anthony Basooma, Astrid Schmidt-Kloiber, Rose Basooma, Herbert Nakiyende, Johannes Kowal, Andrea Funk, Thomas Hein, Florian Borgwardt, Halil Ibrahimi,, Spatiotemporal changes in river network connectivity in the Nile River Basin due to hydropower dams, 2025, 20, 1932-6203, e0322338, 10.1371/journal.pone.0322338
  • Reader Comments
  • © 2022 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

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(3945) PDF downloads(240) Cited by(20)

Figures and Tables

Figures(11)  /  Tables(2)

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog