STOCK MARKET ANALYSIS AND PREDICTION
ABSTRACT
Our aim is to create software that analyzes previous stock data of certain companies, with help of certain parameters that affect stock value. We are going to implement these values in data mining algorithms. This will also help us to determine the values that particular stock will have in near future. We will determine the Month’s High and Low with help of data mining algorithms. A number of research efforts had been devoted to forecasting stock price based on technical indicators which rely purely on historical stock price data. However, the performances of such technical indicators have not always satisfactory. The fact is, there are other influential factors that can affect the direction of stock market which form the basis of market experts’ opinion such as interest rate, inflation rate, foreign exchange rate, business sector, management caliber, investors’ confidence, government policy and political effects, among others.
In this study, the effect of using hybrid market indicators such as technical and fundamental parameters as well as experts’ opinions for stock price prediction was examined. Values of variables representing these market hybrid indicators were fed into the artificial neural network (ANN) model for stock price prediction.
The empirical results obtained with published stock data show that the proposed model is effective in improving the accuracy of stock price prediction. Also, the performance of the neural network predictive model developed in this study was compared with the conventional Box-Jenkins autoregressive integrated moving average (ARIMA) model which has been widely used for time series forecasting. Our findings revealed that ARIMA models cannot be effectively engaged profitably for stock price prediction. It was also observed that the pattern of ARIMA forecasting models were not satisfactory. The developed stock price predictive model with the ANN-based soft computing approach demonstrated superior performance over the ARIMA models; indeed, the actual and predicted value of the developed stock price predictive model were quite close.
CHAPTER ONE
INTRODUCTION
The ability to accurately predict the future is crucial for decision processes in planning, organizing, scheduling, purchasing, strategy formulation, policy making and supply chains management and so on. Therefore, prediction/forecasting is an area where a lot of research efforts have been carried out in the past. This area is presently still an important and active field of human activity and will continue to be in the future (Zhang, 2004). Stock price prediction has always been a subject of interest for most investors and financial analysts, but clearly, finding the best time to buy or sell has remained a very difficult task for investors because there are other numerous factors that may influence stock prices (Weckman et al., 2008, Chang and Liu, 2008 and Adebiyi et al 2009). Therefore, stock market prediction has remained an important research topic in business. However, stock markets environments are very complicated, dynamic, stochastic and thus difficult to predict (Wei, 2005; Gerasimos et al., 2005; Yang and Wu, 2006; Tsanga et al., 2007; and Tae 2007).
Presently, financial forecasting is regarded as one of the most challenging applications of time series forecasting. Financial time series presents complex behaviour, resulting from a huge number of factors, which could be economic, political, or psychological. They are inherently noisy, non-stationary, and deterministically chaotic.
Data mining technology has found increasing acceptance in business areas that need to analyze large amounts of data in order to discover knowledge which could not be found using traditional methods. Time series data mining is identified as one of the 10 challenging problems in data mining research (Yang and Wu, 2006; Lay-Ki and Sang, 2007). Financial time series forecasting has been a subject of research since 1980s. The objective is to beat financial markets and win much profit. However, due to complexity of financial time series, there is some skepticism about predictability of financial time series. Due to dynamic nature and unpredictable environment of stock market domain, predicting the future has always been the desire of mankind because of their inability to deal with uncertain, fuzzy, or insufficient data which fluctuate rapidly in very short periods of time.
It is against this backdrop that this study intends to develop an improved model for Stock Market Analysis and Prediction.
1.2 STATEMENT OF THE PROBLEM
Generally in stock markets, investors are often faced with difficulties of inability to:
- determine and predict the stock market behaviour due to the dynamism and unpredictable environment of stock market domain.
- take decision on the appropriate stock to buy or sell for better profit due to unpredictable nature of stock markets.
- analyze and extract useful knowledge from a vast amount of information in order to make qualitative investment decision. .
- engage effectively technical trading strategies and buy-hold strategy.
Hence, in this study, an ANN-based predictive model to overcome the problems stated above, and profer solutions to the following research questions, is being proposed.
- Could our proposed stock price predictive model be effective in improving the accuracy of stock price prediction with the combination of the parameters of technical, fundamental analysis and experts’ opinion variables?
- Could our proposed stock price predictive model enhance investmentdecisions of investors?
1.3 OBJECTIVES OF THE RESEARCH
The aim of this research work is to develop an improved model for Stock Market Analysis and Prediction
Consequently, the objectives of the research are to:
- develop a predictive model for stock price prediction using hybridized market indicators for enhanced decision making.
- compare the predictive performance of the statistical technique of autoregressive integrated moving average (ARIMA) and the proposed ANNbased predictive model.
- evaluate the proposed stock price predictive model with performance measures.
1.4 METHODOLOGY
The research methodology used in this study is as follows:
- The forecastingtechniques engaged in this study were ARIMA model and soft computing technique of the artificial neural network with multilayer perceptron (MLP) model trained with backpropagation (BP) algorithm. The procedures involved in using ARIMA model for forecasting include: (1) data collection and examination; (2) testing for stationarity; (3) model identification and estimation;
(4) model diagnostics checking; and (5) forecasting and forecast evaluation. Similarly, the steps required for ANN model in developingfinancial predictive model include: (1) variable selection; (2) data collection; (3) data preprocessing; (4) training, testing, and validation sets; (5) neural network design (number of hidden layers, number of hidden neurons, number of output neurons, transfer functions); (6) evaluation criteria; (7) neural network training (number of training iterations, learning rate and momentum); and (8) implementation.
- The stock data used in this study were obtained from New York Stock Exchange (NYSE) and Nigerian Stock Exchange (NSE) respectively. The historical stock data of four different companies, two from each of the stock exchange mentioned were used. The two companies’ stock data from NYSE are Dell Inc. and Nokia
Inc. relatively from information technology industry sector. The companies from NSE are from banking industry sector which are UBA bank and Zenith bank. The technical data used in this study are raw daily opening price, highest price, lowest price, closing price and volume traded in each day. The fundamental data used consist of price per earning, return on asset and return equity and the expert opinions were obtained through interactions with financial experts and through administration of questionnaires.
- Some performance measures: root mean square error, mean square error, and confusion matrix; were used to evaluate the predictive models developed.
- The predictive models were implemented using Eviews software for ARIMA model and Matlab for artificial neural network model.
1.5 SIGNIFICANCE OF THE STUDY
The stock price predictive model developed in this study is of immense benefits to the stakeholders such as traders, investors and stock brokers in stock market domain. It can serve as a useful guide to individual investors in making investment decision on which stock to buy or sell. Moreover, it can enable individual investors to increase wealth through profit gained from usage of the predictive model. Furthermore, it will stimulate interest of individuals to invest in stock market indexes thereby making the sector vibrant, robust and healthy.
1.6 MOTIVATION FOR THE STUDY
The motivation for this study stems from the following reasons: firstly, investors in stock market are desirous to make profits from their investment; however, lack of adequate knowledge of the right stock to buy or sell at the right time poses a big challenge. Secondly, to further contradict the hypothesis formulated in stock market known as the Efficient Market Hypothesis, which says there is no way to make profit by predicting the stock market.
1.7 CONTRIBUTION TO KNOWLEDGE
The specific contributions of this research work pertain to stock forecasting modeling in stock market domain both at local and global levels. Firstly, the study provides an improved predictive model for stock price prediction using the soft computing approach with hybrid market indicators that combined the parameters of technical and fundamental analyses as well as experts’ opinion.
Secondly, this study was able to resolve and clarify contradictory findings reported in literature on the superiority of statistical techniques of ARIMA model over soft computing technique of ANN model in time series prediction and vice-versa. The findings in this study showed that ANN model outperformed the statistical forecasting techniques, and in particular, ARIMA, which is the most widely used statistical forecasting technique.
1.8 LIMITATION AND SCOPE OF THE STUDY
In the study only one soft computing technique, namely ANN, was used. Furthermore the composition of expert’s opinion was limited to five parameters. Also, Eviews and Matlab were used to simulate the proposed model
1.9 ORGANIZATION OF STUDY
The organization of the thesis is summarized as follows: Chapter one is the introduction which is composed of background information of the study, statement of the problem and research question, aim and objective of the study, research methodology, significance of the study, motivation for the study, limitation and scope of the study. Chapter two presents literature review – soft computing theories, related works with use of soft computing in stock prediction and gaps in literature. Chapter three consists of research methodology used in this study which includes sources of data, details of forecasting techniques employed and performance measures used to evaluate the predictive model. Chapter four presents detailed results of the study and finally, chapter five is composed of summary, conclusion and future research work.
Attached Files
STOCK MARKET ANALYSIS AND PREDICTION.docx
₦5,000.00
STOCK MARKET ANALYSIS AND PREDICTION
ABSTRACT
Our aim is to create software that analyzes previous stock data of certain companies, with help of certain parameters that affect stock value. We are going to implement these values in data mining algorithms. This will also help us to determine the values that particular stock will have in near future. We will determine the Month’s High and Low with help of data mining algorithms. A number of research efforts had been devoted to forecasting stock price based on technical indicators which rely purely on historical stock price data. However, the performances of such technical indicators have not always satisfactory. The fact is, there are other influential factors that can affect the direction of stock market which form the basis of market experts’ opinion such as interest rate, inflation rate, foreign exchange rate, business sector, management caliber, investors’ confidence, government policy and political effects, among others.
In this study, the effect of using hybrid market indicators such as technical and fundamental parameters as well as experts’ opinions for stock price prediction was examined. Values of variables representing these market hybrid indicators were fed into the artificial neural network (ANN) model for stock price prediction.
The empirical results obtained with published stock data show that the proposed model is effective in improving the accuracy of stock price prediction. Also, the performance of the neural network predictive model developed in this study was compared with the conventional Box-Jenkins autoregressive integrated moving average (ARIMA) model which has been widely used for time series forecasting. Our findings revealed that ARIMA models cannot be effectively engaged profitably for stock price prediction. It was also observed that the pattern of ARIMA forecasting models were not satisfactory. The developed stock price predictive model with the ANN-based soft computing approach demonstrated superior performance over the ARIMA models; indeed, the actual and predicted value of the developed stock price predictive model were quite close.
CHAPTER ONE
INTRODUCTION
The ability to accurately predict the future is crucial for decision processes in planning, organizing, scheduling, purchasing, strategy formulation, policy making and supply chains management and so on. Therefore, prediction/forecasting is an area where a lot of research efforts have been carried out in the past. This area is presently still an important and active field of human activity and will continue to be in the future (Zhang, 2004). Stock price prediction has always been a subject of interest for most investors and financial analysts, but clearly, finding the best time to buy or sell has remained a very difficult task for investors because there are other numerous factors that may influence stock prices (Weckman et al., 2008, Chang and Liu, 2008 and Adebiyi et al 2009). Therefore, stock market prediction has remained an important research topic in business. However, stock markets environments are very complicated, dynamic, stochastic and thus difficult to predict (Wei, 2005; Gerasimos et al., 2005; Yang and Wu, 2006; Tsanga et al., 2007; and Tae 2007).
Presently, financial forecasting is regarded as one of the most challenging applications of time series forecasting. Financial time series presents complex behaviour, resulting from a huge number of factors, which could be economic, political, or psychological. They are inherently noisy, non-stationary, and deterministically chaotic.
Data mining technology has found increasing acceptance in business areas that need to analyze large amounts of data in order to discover knowledge which could not be found using traditional methods. Time series data mining is identified as one of the 10 challenging problems in data mining research (Yang and Wu, 2006; Lay-Ki and Sang, 2007). Financial time series forecasting has been a subject of research since 1980s. The objective is to beat financial markets and win much profit. However, due to complexity of financial time series, there is some skepticism about predictability of financial time series. Due to dynamic nature and unpredictable environment of stock market domain, predicting the future has always been the desire of mankind because of their inability to deal with uncertain, fuzzy, or insufficient data which fluctuate rapidly in very short periods of time.
It is against this backdrop that this study intends to develop an improved model for Stock Market Analysis and Prediction.
1.2 STATEMENT OF THE PROBLEM
Generally in stock markets, investors are often faced with difficulties of inability to:
Hence, in this study, an ANN-based predictive model to overcome the problems stated above, and profer solutions to the following research questions, is being proposed.
1.3 OBJECTIVES OF THE RESEARCH
The aim of this research work is to develop an improved model for Stock Market Analysis and Prediction
Consequently, the objectives of the research are to:
1.4 METHODOLOGY
The research methodology used in this study is as follows:
(4) model diagnostics checking; and (5) forecasting and forecast evaluation. Similarly, the steps required for ANN model in developingfinancial predictive model include: (1) variable selection; (2) data collection; (3) data preprocessing; (4) training, testing, and validation sets; (5) neural network design (number of hidden layers, number of hidden neurons, number of output neurons, transfer functions); (6) evaluation criteria; (7) neural network training (number of training iterations, learning rate and momentum); and (8) implementation.
Inc. relatively from information technology industry sector. The companies from NSE are from banking industry sector which are UBA bank and Zenith bank. The technical data used in this study are raw daily opening price, highest price, lowest price, closing price and volume traded in each day. The fundamental data used consist of price per earning, return on asset and return equity and the expert opinions were obtained through interactions with financial experts and through administration of questionnaires.
1.5 SIGNIFICANCE OF THE STUDY
The stock price predictive model developed in this study is of immense benefits to the stakeholders such as traders, investors and stock brokers in stock market domain. It can serve as a useful guide to individual investors in making investment decision on which stock to buy or sell. Moreover, it can enable individual investors to increase wealth through profit gained from usage of the predictive model. Furthermore, it will stimulate interest of individuals to invest in stock market indexes thereby making the sector vibrant, robust and healthy.
1.6 MOTIVATION FOR THE STUDY
The motivation for this study stems from the following reasons: firstly, investors in stock market are desirous to make profits from their investment; however, lack of adequate knowledge of the right stock to buy or sell at the right time poses a big challenge. Secondly, to further contradict the hypothesis formulated in stock market known as the Efficient Market Hypothesis, which says there is no way to make profit by predicting the stock market.
1.7 CONTRIBUTION TO KNOWLEDGE
The specific contributions of this research work pertain to stock forecasting modeling in stock market domain both at local and global levels. Firstly, the study provides an improved predictive model for stock price prediction using the soft computing approach with hybrid market indicators that combined the parameters of technical and fundamental analyses as well as experts’ opinion.
Secondly, this study was able to resolve and clarify contradictory findings reported in literature on the superiority of statistical techniques of ARIMA model over soft computing technique of ANN model in time series prediction and vice-versa. The findings in this study showed that ANN model outperformed the statistical forecasting techniques, and in particular, ARIMA, which is the most widely used statistical forecasting technique.
1.8 LIMITATION AND SCOPE OF THE STUDY
In the study only one soft computing technique, namely ANN, was used. Furthermore the composition of expert’s opinion was limited to five parameters. Also, Eviews and Matlab were used to simulate the proposed model
1.9 ORGANIZATION OF STUDY
The organization of the thesis is summarized as follows: Chapter one is the introduction which is composed of background information of the study, statement of the problem and research question, aim and objective of the study, research methodology, significance of the study, motivation for the study, limitation and scope of the study. Chapter two presents literature review – soft computing theories, related works with use of soft computing in stock prediction and gaps in literature. Chapter three consists of research methodology used in this study which includes sources of data, details of forecasting techniques employed and performance measures used to evaluate the predictive model. Chapter four presents detailed results of the study and finally, chapter five is composed of summary, conclusion and future research work.
Attached Files