Application of Deep Learning Algorithms and Genetic Programming for Forecasting Stock Prices in Nepal: A Comparative Study

Abstract

As developers and software engineers enter the stock market with new strategies, such as machine learning models, the potential for stock price prediction has increased. However, other potential alternatives, such as GP, have been explored less in this area. This paper aims to validate the accuracy of existing Deep learning algorithms and compare their effectiveness with Gene Programming in forecasting stock prices in Nepal. Here are the stock data from 2014 to 2024, along with the DL and GP, to predict the stock prices of 29 stocks from 15 sectors within NEPSE. The performance metrics are evaluated in terms of accuracy and resistance to volatility. The results show that GP consistently outperforms the DL algorithms across all performance metrics, further validated by the Mann-Whitney U test. The findings also suggest the potential for integrating advanced forecasting methods, such as GP and GRU (or LSTM), into financial decision-making to enhance investment strategies.