Heterogeneous local model networks for time series prediction

Oh, Sang-Keon; Kim, Min-Soeng; Eom, Tae-Dok; Lee, Ju-Jang

doi:10.1016/j.amc.2004.08.018

Cited by 13 publications

(2 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We select representative threelevel countries, including fragile countries (South Sudan and Somalia), vulnerable countries (Indonesia and China), and stable countries (the UK and USA). In the absence of extreme weather events, natural disasters, and economic and political crises, intervention in the climate environment will gradually improve the postintervention climate conditions [19].…”

Section: Empirical Analysis Of Mitigating Climatementioning

confidence: 99%

Analysis of the Impact of Climate Change on National Vulnerability Based on Fuzzy Comprehensive Evaluation

Zhu

Chen

Zhang

2020

Discrete Dynamics in Nature and Society

View full text Add to dashboard Cite

Climate change has become one of the major threats to global security. On the impact of climate change on national vulnerability, the researchers firstly select relevant factors and data and analyze the principal components to determine the weight of climate change on vulnerability so as to analyze the impact of climate change on national vulnerability. Secondly, based on the influencing factors, we obtain the national vulnerability score through the fuzzy comprehensive evaluation method, delineate the impact scope according to the score, and predict climate change by means of regression analysis and time series analysis. Finally, according to the characteristics of cities and continents, the model is revised, a PSR model and a cluster analysis model are established, and the prediction accuracy is improved.

show abstract

Section: Empirical Analysis Of Mitigating Climatementioning

confidence: 99%

Analysis of the Impact of Climate Change on National Vulnerability Based on Fuzzy Comprehensive Evaluation

Zhu

Chen

Zhang

2020

Discrete Dynamics in Nature and Society

View full text Add to dashboard Cite

show abstract

“…Consequently, it is difficult for a single predictor to capture such a switching input-output relationship. Inspired by the so-called "divide-and-conquer" principle that is often used to attack complex problems, the approaches of local modeling have emerged as one of the promising methods of time series prediction (Oh [28]). This study employed a sparse multi-manifold clustering (SMMC) algorithm (Elhamifar and Vidal [14]) for partitioning the feature space into several disjointed regions for different time series dynamics.…”

Section: Introductionmentioning

confidence: 99%

Financial Data Analysis by Manifold Clustering and Kernel Machines

2016

International Conference on Engineering Technologies and Big Data Analytics (ETBDA’2016) Jan. 21-22, 2016 Bangkok (Thailand)

View full text Add to dashboard Cite

Abstract-The challenge of modern financial forecasting comes from the high-dimensionality, nonlinearity, and non-stationarity of financial data. To address the problem, this study employs wavelet analysis to map time domain inputs to time-frequency (or wavelet) domain, and a sparse multi-manifold clustering (SMMC) to partition the high-dimensional feature space into several disjointed regions according to their dynamics. In the final stage, hierarchical multiple kernel machines (HMKM) are employed to perform the high-dimensional forecasting and trading. In our system, SMMC can effectively cluster data on multiple manifolds that are very close to each other, manifolds with non-uniform sampling and holes. HMKM embeds basis kernels in a directed acyclic graph, and optimized them by a graph-adapted sparsity-inducing norm, which performs the feature selection in polynomial time in the number of selected kernels. The empirical results demonstrate that the proposed model outperforms traditional neural networks, support vector machines, statistical models, and significantly reduces the forecasting errors. However, the tight correlations among financial markets provide investors with valuable information to make accurate forecasts regarding the co-movements of stock indices. International investors are a diverse group, operating on very different time scales. As a result, the correlation pattern between international market indices are not fixed between each time scale. Consequently, this study aims to address the problem by a new model in wavelet domain that fully exploits time-frequency features from high-dimensional financial time series. Namely, this study will implement a new forecasting strategy which Another problem in forecasting is that financial time series are usually non-stationary; namely, time series switch their dynamics between different regions. This leads to changes in the dependency structure between input and output variables. Consequently, it is difficult for a single predictor to capture such a switching input-output relationship. Inspired by the so-called "divide-and-conquer" principle that is often used to attack complex problems, the approaches of local modeling have emerged as one of the promising methods of time series prediction (Oh [28]). This study employed a sparse multi-manifold clustering (SMMC) algorithm (Elhamifar and Vidal [14]) for partitioning the feature space into several disjointed regions for different time series dynamics. We then employed an architecture involving multiple experts to overcome the problem, namely, using different experts for different feature regions. Keywords-ManifoldSparse manifold clustering is outstanding at partitioning data points non-linearly distributed on multiple manifolds. In contrast to traditional nearest neighbors-based methods for manifold modeling, which fix the number of neighbors or the neighborhood radius and then compute the weights between points in each neighborhood, SMMC finds both the neighbors and the weights automatically. SMMC automat...

show abstract

Integrating spectral clustering with wavelet based kernel partial least square regressions for financial modeling and forecasting

Huang

2011

Applied Mathematics and Computation

View full text Add to dashboard Cite

Heterogeneous local model networks for time series prediction

Cited by 13 publications

References 16 publications

Analysis of the Impact of Climate Change on National Vulnerability Based on Fuzzy Comprehensive Evaluation

Analysis of the Impact of Climate Change on National Vulnerability Based on Fuzzy Comprehensive Evaluation

Financial Data Analysis by Manifold Clustering and Kernel Machines

Integrating spectral clustering with wavelet based kernel partial least square regressions for financial modeling and forecasting

Contact Info

Product

Resources

About