Prediction of Stock Prices Using Unstructured and Semi-structured Qualitative Data – A Neural Network Approach

Abstract

As more interest by the people about stock markets is growing substantially bigger, the more their thinking is directed to a systematic method to predict stock prices that vary. There is an increasing appreciation that unstructured qualitative data contain awfully precious information, helpful for a variety of reason. It is this motivation that has fascinated awareness of researchers and specialized into the interesting and demanding area of stock market prediction using qualitative news and events. In this paper, a framework for stock price movement prediction using news, company newsletter and financial information was presented. Financial information and events were processed and multilayer feed forward network trained using back propagation was employed to empirically explore the intricate relationships between events and the stock market prices. Among the various industries we chose to focus on the Information Technology (IT) sector since it is most prevalent and applicable to the Indian markets. The stock prices and the relevant unstructured information for six IT companies were collected and filtered based on set of keywords. Our research has proven that the stock price can be accurately predicted from news articles by furnishing the empirical results that were obtained after training the multilayer neural networks.

Stock Price Forecasting Using Univariate Singular Spectral Analysis through Hadamard Transform

“…Kim and Han proposed a genetic algorithm (GAs) and artificial neural networks (ANN) to predict the stock price index [4]. The neural network approach is also used by Rajakumar et al [5]. Kim proposed a support vector machine (SVM) and comparing to the backpropagation neural network (BPNN) and casebased reasoning (CBR) to evaluate the stock price prediction [6].…”