Application of a Novel Deep Fuzzy Dual Support Vector Regression Machine in Stock Price Prediction

Abstract

The desire of any investor is to accurately predict market behavior in order to maximize profits. This is a daunting task because market behavior is random, volatile, and influenced by many factors. Deep learning has excellent feature learning ability, and support vector machine has excellent reasoning ability. In recent years, the deep support vector machine network that perfectly combines the advantages of the two has attracted the attention of many scholars. Compared with the traditional deep neural network, the deep support vector machine network has the following advantages: (1) It has higher reasoning ability; (2) It is more suitable for tasks with insufficient training samples. This paper proposes a new deep fuzzy dual support vector regression network to predict stock price through the numerical data of stock prices. The method proposed in this study is a hybrid model that combines the advantages of: (a) evolutionary computation, (b) ensemble learning, (c) deep learning and (d) multi-kernel function learning. In addition to providing the most probable prediction results, the deep fuzzy dual support vector regression machine proposed in this study can also provide the inner and outer boundaries of the fuzzy range of the prediction results, as well as the confidence level of the prediction results.