A Novel Artificial Intelligence-Based Model for Stock Price Prediction

Predicting stock price movements is a crucial, intriguing, and highly challenging task for researchers, traders, and market analysts. Daily stock price prediction is particularly difficult due to the nonlinear and chaotic nature of stock price fluctuations. Artificial intelligence (AI) techniques have been widely applied to predict data with nonlinear and chaotic structures. Most previous studies have utilized single artificial neural networks (ANNs) along with a limited number of technical indicators for stock price and index prediction. However, relying on a single ANN with fixed input variables often leads to increased prediction errors. Therefore, a more robust model is required to achieve higher prediction accuracy. In this study, a two-stage hybrid framework is proposed. In the first stage (feature selection), the genetic algorithm (GA) and multilayer perceptron (MLP) neural network are employed to identify the most suitable technical indicators. In the second stage (ensemble learning), three predictors namely, the multilayer perceptron (MLP), adaptive neuro-fuzzy inference system (ANFIS), and radial basis function (RBF) neural network are combined to forecast stock prices. The final output is obtained by averaging the results of these predictors. Experimental results demonstrate that the proposed model significantly reduces prediction errors compared to individual methods.

کلیدواژه ها:

Stock Price Prediction ، Artificial Neural Networks ، Metaheuristic Algorithms ، Feature selection ، Ensemble Learning