Dandelion Optimizer with LSTM Model (DO-LSTM) for Stock Input Feature Selection for Price Forecasting

Stock price prediction is a critical decision-making function in finance, which offers investors valuable information regarding future market trends. This paper presents a novel hybrid model that combines the Dandelion Optimizer (DO) and Long Short-Term Memory (LSTM) networks for forecasting the stock market close price. Dandelion Optimizer is employed as a feature selection method to identify the most significant technical indicators, e.g., Simple Moving Average (SMA), Exponential Moving Average (EMA), Momentum, and Volatility, which have a direct influence on the stock price movements. Two feature selection approaches are explored: DO with Linear Regression-based Feature Selection and DO with LSTM-Based Feature Selection. Experimental results indicate that the DO with Linear Regression Feature Selection + LSTM Prediction Model outperforms the LSTM-based feature selection model in terms of prediction accuracy and reliability with the lowest Mean Squared Error (MSE) of 14.4632 and the highest R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Score of 0.93. The findings validate the importance of technical indicators in enhancing model performance and the promise of hybrid optimization-based deep learning models for financial time series forecasting. By offering a reliable and computationally effective framework, this study positively impacts the advancement of intelligent forecasting of stocks systems.