An analytics system for time-series-cross-section data and its applications

Project Details


This project aims to propose a new type of analytics system for time-series-crosssection (TSCS) data. The first step is to establish models for TSCS data from different perspectives separately and establish the associated marginal networks by the extracted features. Next, we propose to develop a feature fusion method to combine all the components extracted from different perspectives into a single network. The final step is to establish a network time series model with the fusion network for prediction. This project plans to apply the proposed system to financial data. We will start from the perspectives of classical time series modeling, the economic theory of investment, and deep neural network models separately for collecting useful features of market trend predictions. We complete the proposed analytics system's development by performing feature fusion and network time series modeling with the collected features. In our empirical study, we plan to compare the in-sample and out-of-sample performances of the proposed approach with each marginal network. Besides, market participants can use the features extracted from each method to predict market trends separately. One can also create associated investment strategies according to these market trend predictions. A comparison study of the investment performances will investigate whether the proposed system would improve market trend predictions or classifications.
Effective start/end date1/08/2231/10/23


  • feature fusion
  • feature extraction
  • market portfolio
  • network time series


Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.