Man’s common goal from time immemorial is to make his life easier. The prevailing notion in most societies is that wealth affords one a certain level of comfort and luxury, so it is not surprising that so much effort has gone into ways of predicting the financial markets, a key source of wealth creation and preservation. Forecasting stock price or financial markets has been one of the biggest challenges to the Artificial Intelligence (AI) community. Various technical, fundamental, and statistical indicators have been proposed and used with varying degrees of success. However, none of these techniques or combination of techniques has been deemed successful enough, hence, the continued search for that ‘holy grail’.
The objective of forecasting research has been largely beyond the capability of traditional AI research which has mainly focused on developing intelligent systems that are supposed to emulate human intelligence. By its nature, the stock market is mostly complex (non-linear) and volatile. With the development of neural networks, researchers and investors are hoping that the market mysteries can be unravelled. Artificial Neural Networks inspired by human brain cells’ activity can learn the data patterns and generalize their knowledge to recognize the future new patterns.
Researches on neural networks show that neural networks have great capability in pattern recognition and machine learning problems such as classification and regression.
These days neural networks are considered as a common data mining method in different fields like finance, business, industry, and science. The application of neural networks to prediction problems is very promising due to some of their special characteristics. First, traditional methods such as linear regression and logistic regression are model-based while neural networks are self-adjusting methods based on training data, so they have the ability to solve the problem with a little knowledge about its model and without constraining the prediction model by adding any extra assumptions. Besides, neural networks can find the relationship between the input and output of the system even if this relationship might be very complicated because they are general function approximators.
Consequently, neural networks are well applied to the problems in which extracting the relationships among data is really difficult but, on the other hand, there exists a large enough training data sets. It should be mentioned that, although sometimes the rules or patterns that we are looking for might not be easily found or the data could be corrupted due to the process or measurement noise of the system, it is still believed that the inductive learning or data-driven methods are the best ways of dealing with real world prediction problems.
Furthermore, neural networks have generalization ability, meaning that after training they can recognize the new patterns even if they haven’t been in training set. Since in most of the pattern recognition problems predicting future events (unseen data) is based on previous data (training set), the application of neural networks would be very beneficial. Neural networks have been proved to be general function approximators, particularly the Multi-layer Perceptron (MLP) neural network which has been used to approximate many complex continuous functions that enable us to learn any complicated relationship between the input and the output of the system.
In the recent decades, many researches have been done on the application of neural networks to predict the stock market changes. We attempt to replicate the works of Tseng, Kwon and Tjung (2012) in which they combined neural networks and genetic algorithms for the forecast of the direction of Singapore stock market, achieving 81 percent precision. In an attempt to predict the daily value of the NSE-ASI, we will be applying the Tseng et al. methodology to the Nigerian Stock Exchange, using the daily series of the All-Share Index from January 2000 to December 2014. We employ the one-day lags of the NSE-ASI, the NGN/USD, and 3-months treasury bills rate, along with five dummy variables to capture each trading day of the week.
The obtained results from our simulation show that for predicting the direction of next day’s changes in the value of the NSE-ASI, our model achieves a precision of 83 percent. This outperforms the simple linear regression model and MLP neural network which achieved 46 percent and 68 percent, respectively. This is contrary to the Efficient Market Hypothesis postulation, and suggests additional premium can be gained by applying this model on the Nigerian Stock Exchange.
Olugbenga A. Olufeagba
