A Spatiotemporal Deep Learning-Based Multisource Data Analytics Framework for Basketball Game

Data analytics has been an important business demand for basketball game. Conventionally, it was implemented with use of statistical approaches, yet neglecting the perception of spatiotemporal characteristics data. To deal with the problem, we introduce deep learning to reveal potential spatiotemporal features from multisource data. Therefore, a spatiotemporal deep learning-based multisource data analytics framework for basketball game, is proposed in this paper. Firstly, advanced deep learning models under spatiotemporal characteristics are elaborated, and a back propagation-based convolution neural network structure is utilized to extract the spatiotemporal features. Then, a systematic data collection and preprocessing method under spatiotemporal characteristics is defined, and thus formulating the spatiotemporal deep learning-based data analytics framework. Furthermore, we conduct a case study to make empirical analysis for the proposed technical framework. The results show that the proposal can well make prediction for some key factors of basketball game, and that it can provide some predictive information for the game preparation from the technical dimension.