Academic journal article Academy of Accounting and Financial Studies Journal

Study of the Forecasting Performance of China Stocks' Prices Using Business Intelligence (BI): Comparison between Normalized and Denormalized Data

Academic journal article Academy of Accounting and Financial Studies Journal

Study of the Forecasting Performance of China Stocks' Prices Using Business Intelligence (BI): Comparison between Normalized and Denormalized Data

Article excerpt

INTRODUCTION

Business Intelligence (BI) represents a powerful technique for extracting meaningful information from vast amounts of raw data, interpreting inherited relationships among the data, and eventually facilitating the decision making process.

Stock markets, located throughout the world, generate numerous data covering various aspects on a daily basis. Investors who expect to make profits by forecasting stock market prices are confused by easily accessed but overwhelming information. A wide array of factors, such as the fluctuations in major global stock indices, competitors, investors' sentiments, and political elections make the prediction of stock prices even more complicated.

The stock market in China, one of the largest markets in the world, has its own unique characteristics. It has been more than 30 years since economic reform was launched in China but it is still considered young and state-controlled with a high degree of corruption. A study of the distinct differences between the stock markets in China and the markets in developed countries would yield compelling results.

The authors have been conducting research in predicting stock prices utilizing Business Intelligence (BI) for several years. Kwon, Tjung, and Tseng (2012) indicated that the Neural Network (NN) model, using a financial data mining technique, performs better in forecasting stock prices than the Ordinary Least Squares (OLS) model does in US stock markets. The forecasting accuracy is measured in terms of the significant percentage forecasting error of the mean and standard error. For the NN model, means vary from 2.1% to 12.31% and standard deviations change from 2.11% to 14.92%. The OLS model has means from 1.93% to 24.8% and standard deviations from 2.15% to 12.3%. In addition, the NN model provides more insights in identifying critical predictors which would

increase forecasting accuracy in stock market analysis. The study considers eight indicators which are macroeconomic indicators, microeconomic indicators, market indicators, market sentiments, institutional investors, political indicators, business cycles, and calendar anomalies. Furthermore, as a linear model, OLS has limited capacity and inconsistent performance in stock market prediction.

In this study, the NN model is applied to predict the price fluctuations in China's stock markets. The sample period ranges from 2002 to 2012 covering the dot-com bubble in late 2002, 2003's outbreak of SARS, party leadership transitions in 2002 and 2012, the global financial crisis and economic recession since 2007, and lastly the Wenchuan earthquake and the Beijing Olympic Games of 2008. In addition, 25 indicators, including macroeconomic indicators, market sentiment indicators, institutional investors, and microeconomic indicators, were added as independent variables to better predict the stock price changes in China. National holidays were also taken into consideration in the forecasting model. In our previous study, we received criticism that the results would lose physical meaning as stock prices after multiple steps of data manipulation. Thus, in this study, we de-normalized the predicted results and compared them with the original stock prices. The new generated dataset will be more compelling in explaining the forecasting accuracy of the BI approach.

LITERATURE REVIEW

Business Intelligence in Stock Price Forecasting

Business intelligence has been widely researched in stock market prediction in recent years. The NN model will be the primary form of business intelligence techniques in this study. In recent years, several studies have been conducted to compare the different versions of NN models regarding the productivity of China's stock markets. Dai, Wu & Lu (2012) proved that the combination of nonlinear independent component analysis (NLICA) and neural network has higher accuracy in forecasting Shanghai B shares than other NN models, including LICA-BPN, PCA BPN, and single BPN. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.