Stochastic Forward-Looking Prediction for Entropy-Based Multimodal Fused Data

Abstract

Nowadays data available are vast and versatile in constant change along with time, showing a dynamic behavior. The fusion of multimodal data sources in a specific domain provides more accurate and reliable information than data sources being used separately. The dynamic behavior of a random variable in time series can be simulated by a general stochastic index -- fused entropy process for prediction. This paper proposes a stochastic forward-looking prediction for multi-modal fused data. Based on the general entropy of multimodal data to indicate the total chaotic index, a stochastic model based on the general entropy is proposed to simulate the dynamic behavior of the general entropy and make the forward-looking prediction for it by calculating an expected value of the predicted general index. The general index can accurately indicate the predicted uncertainty of multiple dynamic variables. An experiment is carried out for weather data fusion and the uncertainty of future weather status is simulated and then predicted. The experimental result is confirmative of our method.