ANN Model to Predict Stock Prices at Stock Exchange Markets (1502.06434v1)

Abstract: Stock exchanges are considered major players in financial sectors of many countries. Most Stockbrokers, who execute stock trade, use technical, fundamental or time series analysis in trying to predict stock prices, so as to advise clients. However, these strategies do not usually guarantee good returns because they guide on trends and not the most likely price. It is therefore necessary to explore improved methods of prediction. The research proposes the use of Artificial Neural Network that is feedforward multi-layer perceptron with error backpropagation and develops a model of configuration 5:21:21:1 with 80% training data in 130,000 cycles. The research develops a prototype and tests it on 2008-2012 data from stock markets e.g. Nairobi Securities Exchange and New York Stock Exchange, where prediction results show MAPE of between 0.71% and 2.77%. Validation done with Encog and Neuroph realized comparable results. The model is thus capable of prediction on typical stock markets.

• The paper demonstrates that a refined ANN model with a 5:21:21:1 configuration reduces prediction error to a MAPE between 0.71% and 2.77%.
• The study employs a feedforward multi-layer perceptron with backpropagation to overcome limitations of traditional technical, fundamental, and time series methods.
• The model generalizes across varying market conditions, offering stockbrokers a robust tool for more precise investment decision-making.

Artificial Neural Network Model for Stock Market Prediction

The paper entitled "ANN Model to Predict Stock Prices at Stock Exchange Markets" explores the application of artificial neural networks (ANNs) for forecasting stock prices, as an alternative to traditional predictive methods like technical, fundamental, and time series analysis. The authors propose a model utilizing a feedforward multi-layer perceptron (MLP) with error backpropagation, and they present experimental results that demonstrate the viability of this approach.

In the domain of stock market prediction, the paper targets addressing the limitations of existing methodologies that often fail to reliably forecast exact stock price values. The authors argue that while stockbrokers commonly employ these methods to advise clients, they primarily indicate trends rather than precise figures, making them insufficient for optimal decision-making. Hence, the research aims to create an ANN-powered predictive tool to enhance the accuracy of stock price predictions.

The paper employs a meticulous tuning process to refine the architecture of the ANN model. The model is initialized with a baseline configuration of 5:11:11:1, indicating network layers and proportional training versus testing data sets. It is then experimented upon and refined to a configuration of 5:21:21:1, using 80% of historical data over 130,000 training cycles. This approach signifies the model's robustness by achieving a Mean Absolute Percentage Error (MAPE) between 0.71% and 2.77% when tested across various stock data from both the Nairobi Securities Exchange and New York Stock Exchange.

The methodology of tuning the ANN involves incremental adjustments of neurons in hidden layers and varying data volumes, among other parameters. Through these experiments, the authors ascertain the optimum configuration conducive to minimizing prediction error, illustrating a strategic enhancement over baseline model performance. Notably, the final ANN model demonstrates the capability to generalize stock price movements across diverse market conditions, thereby showcasing adaptability.

Practical and theoretical implications of this research are significant. Practically, the developed ANN model presents a high-precision tool capable of assisting stockbrokers in making more informed investment recommendations. Theoretical contributions include demonstrating the efficacy of ANN in addressing non-linear prediction challenges inherent in financial data analysis, underscoring the potential for future developments in AI-driven finance tools.

For future research directions, the authors suggest extending the model to accommodate a broader range of stocks and refining neural network configurations beyond the studied parameters. Additionally, exploring the longevity of a trained ANN's efficacy in prediction before necessitating retraining could offer insights into maintaining model relevance in rapidly changing economic climates.

Overall, this paper provides a comprehensive analysis of employing artificial neural networks for stock price prediction and highlights the potential for ANN-based models to surpass traditional forecasting methods in terms of precision and reliability. The insights garnered from the research suggest promising advancement potential for AI applications in financial markets.