Lévy Flight Chaotic Runge Kutta Optimizer for Stock Price Forecasting

Maximizing investment returns is a key focus for investors and stakeholders, the challenge pertaining to stock market forecasting has assumed significant importance within this discipline, attributable to the substantial market volatility arising from a plethora of interconnected factors prevalent in the stock market. In this paper we introduced a hybrid machine learning model based on long short-term memory (LSTM) networks and lévy flight chaotic Runge Kutta optimizer (LCRUN) to predict three stock indices daily price. The LCRUN incorporates an initialization technique utilizing chaotic mapping to enhance population diversity. A method namely lévy flight are used in the LCRUN for enhancing ability to search globally optimal solutions and avoid falling into local optima. Our proposed model evaluates the performance with Runge Kutta-based LSTM (RUN-LSTM), sine cosine-based LSTM (SCA-LSTM), differential evolutionary-based LSTM (DE-LSTM) and particle swarm-based LSTM (PSO-LSTM) on these stock indices includes S&P500, NASDAQ100, and SPY. The experimental results show that the proposed LCRUN-LSTM has significant performance and predicts stock prices more accurately than the other four models.