A Comparative Study on Stock Market Forecasting using Lstm, Gru and Rf

Abstract

Stock market prediction is a thriving topic across the globe, millions of new dematerialised accounts were opened in past few years. Stock market prediction has been a challenging problem for both economists and data scientists. For the sake of building effective prediction model, various researchers are coming with new techniques and approaches achieving good yet capricious results as market depends upon various stochastic factors.Our experiment focused on comparing three models: LSTM, GRU, and Random Forest (RF). We evaluated the performance of each model using R-squared (R�) and Root Mean Squared Error (RMSE) metrics. The results revealed that the GRU model delivered the highest prediction accuracy, successfully capturing complex price patterns in NEPSE stock data. The LSTM model performed well, coming in a close second. On the other hand, the Random Forest model showed signs of over-fitting, which led to less reliable predictions. The GRU model�s outstanding performance is largely due to its ability to effectively manage sequential data, making it highly suitable for stock price forecasting, with the LSTM model also demonstrating strong predictive capabilities.