Genetic algorithms approach to feature discretization in artificial neural networks for the prediction of stock price index
Introduction
For a long time, there has been much research interest in predicting the stock price index. Among them, there are many studies using data mining techniques including artificial neural networks (ANNs). However, most studies showed that ANN had some limitations in learning the patterns because stock market data has tremendous noise and complex dimensionality. ANN has preeminent learning ability while it is often confronted with inconsistent and unpredictable performance for noisy data. In addition, sometimes the amount of data is so large that the learning of patterns may not work well. In particular, the existence of continuous data and large amount of data may pose a challenging task to explicit concepts extraction from the raw data due to the huge amount of data space determined by continuous features (Liu & Setiono, 1996). Many researchers in the society of data mining are interested in the reduction of dimensionality. The reduction and transformation of the irrelevant or redundant features may shorten the running time and yield more generalized results (Dash & Liu, 1997).
This paper proposes a new hybrid model of ANN and genetic algorithms (GAs) for feature discretization to mitigate the above limitations. Feature discretization is to transform continuous values into discrete ones in accordance with certain thresholds. Feature discretization is closely related to the dimensionality reduction (Liu & Motoda, 1998a). Properly discretized data can simplify the process of learning and may improve the generalizability of the learned results. This study uses GA to search the optimal or near-optimal thresholds for feature discretization. In addition, this study simultaneously searches the connection weights between layers in ANN. The genetically evolved connection weights mitigate the well-known limitations of the gradient descent algorithm.
The rest of the paper is organized as follows. Section 2 reviews prior research. Section 3 proposes feature discretization using GA and describes the benefits of the proposed approach. Section 4 describes the research design and experiments. In Section 5, the empirical results are summarized and discussed. In Section 6, conclusions and the limitations of this study are presented.
Section snippets
Prior research on stock market prediction using ANN
Many studies on stock market prediction using artificial intelligence (AI) techniques were performed during the past decade. These studies used various types of ANN to predict accurately the stock index and the direction of its change.
One of the earliest studies, Kimoto, Asakawa, Yoda and Takeoka (1990) used several learning algorithms and prediction methods for developing the Tokyo stock exchange prices index (TOPIX) prediction system. They used the modular neural network to learn the
GA approach to feature discretization for ANN
Many fund managers and investors in the stock market generally accept and use certain criteria for technical indicators as the signal of future market trends. Even if a feature represents a continuous measure, the experts usually interpret the values in qualitative terms such as low, medium, and high (Slowinski & Zopounidis, 1995). For ‘Stochastic %K’, one of the most popular technical indicators, the value of 75 is basically accepted by stock market analysts as a strong signal if the value
Research data and experiments
The research data used in this study is technical indicators and the direction of change in the daily Korea stock price index (KOSPI). The total number of samples is 2928 trading days, from January 1989 to December 1998. Table 2 gives selected features and their formulas (Achelis, 1995, Chang et al., 1996, Choi, 1995, Edwards and Magee, 1997, Gifford, 1995).
The direction of daily change in the stock price index are categorized as “0” or “1”. “0” means that the next day's index is lower than the
Experimental results
Three models are compared according to the methods of determining the connection weights and the methods of feature transformation. Table 5 describes the average prediction accuracy of each model.
In Table 5, GAFD has higher prediction accuracy than BPLT and GALT by 10∼11% for the holdout data. It is worth giving attention to the fact that there is a shade of difference of prediction accuracy between the training data and the holdout data for GAFD. There is, however, a wide difference between
Concluding remarks
As mentioned earlier, previous studies tried to optimize the controlling parameters of ANN using global search algorithms. Some of them only focused on the optimization of the connection weights of ANN. Others had an interest in the optimization of the learning algorithms itself, but most studies had little interest in the dimensionality reduction and the elimination of irrelevant patterns. This paper has proposed a new hybrid GA and ANN to mitigate the above limitations. In this paper, GA not
Acknowledgements
The authors would like to thank Korea Science and Engineering Foundation for supporting this work under Grant No. 98-0102-08-01-3.
References (46)
- et al.
Unsupervised feature selection using a neuro-fuzzy approach
Pattern Recognition Letters
(1998) - et al.
Feature selection methods for classifications
Intelligent Data Analysis—An International Journal
(1997) - et al.
Comparing backpropagation with a genetic algorithm for neural network training
Omega
(1999) - et al.
Simultaneous design and training of ontogenic neural network classifiers
Computers and Operations Research
(1996) - et al.
Dimensionality reduction via discretization
Knowledge-Based Systems
(1996) - et al.
An initial comparison of a fuzzy neural classifier and a decision tree based classifier
Expert Systems with Applications
(1998) - et al.
Global optimization for artificial neural networks: a tabu search application
European Journal of Operational Research
(1998) - et al.
Toward global optimization of neural networks: a comparison of the genetic algorithm and backpropagation
Decision Support Systems
(1998) - et al.
Optimization of neural networks: a comparative analysis of the genetic algorithm and simulated annealing
European Journal of Operational Research
(1999) Analyzing discretizations of continuous attributes given a monotonic discrimination function
Intelligent Data Analysis—An International Journal
(1997)
Forecasting S&P 500 stock index futures with a hybrid AI system
Decision Support Systems
Technical analysis from A to Z
Machine learning: neural networks, genetic algorithms, and fuzzy systems
Genetic algorithms and investment strategies
Data mining with neural networks
Extreme events from the return-volume process: a discretization approach for complexity reduction
Applied Financial Economics
Technical indicators and analysis methods
Technical indicators
Business research methods
Genetic algorithms and financial applications
The use of parsimonious neural networks for forecasting financial time series
Journal of Computational Intelligence in Finance
Cited by (569)
A study of concave shoulder angle on the mechanical properties and fractography of friction stir welded AA6061-T651 joints
2024, Journal of Materials Research and TechnologyPractical machine learning: Forecasting daily financial markets directions
2023, Expert Systems with ApplicationsStock price forecasting using PSO hypertuned neural nets and ensembling[Formula presented]
2023, Applied Soft ComputingNews-based intelligent prediction of financial markets using text mining and machine learning: A systematic literature review
2023, Expert Systems with ApplicationsScientometric review and analysis of recent approaches to stock market forecasting: Two decades survey
2023, Expert Systems with Applications
- 1
Tel.: +82-2-958-3613; fax: +82-2-958-3604.