A Novel Data Mining Algorithm based on BP Neural Network and Its Applications on Stock Price Prediction

Abstract

As an indispensable part in financial industry, stock investment has gradually become more and more important in people's life. The major fluctuation in stock market directly affects the stability of financial markets and the healthy development of the economy. The successful stock price and trend prediction will help investors to yield profit and provide timely and reasonable guidance to market regulators. Therefore, how to forecast stock market is a crucial and worthy topic in financial area. Neural network simulated how brain functions, can automatically extract features from the historical data and related information to predicate future trend, with the primary characteristics of self-organization and adaption. Therefore, neural network is suitable for processing prediction in the field of nonlinear time series. This paper presents a novel data mining algorithm based on the BP neural network and its applications on stock price prediction. Taking improving the efficiency of the traditional back propagation network, we revise the perspective error, and modify incentive function and the adaptive learning rate in order to avoid the local minimum and decrease training time. Moreover, to eliminate the influences caused by noisy data, we combine the fuzzy rough set theory to pre-process the original data. The experimental result verifies feasibility of our model, without the local minimum and slow convergence speed, compared to the other state-of-the-art approaches. Therefore, our algorithm on the predications of the exact stock price value and future trend are applicable and satisfactory.