Using AI models in stocks prices prediction: Case of the Moroccan stock market

Abstract

The use of machine learning and deep learning to anticipate and capture stock market trends has captured the attention of the financial sector, offering the prospect of more accurate and efficient prediction of market dynamics. Faced with the growing complexity and highly volatile nature of financial markets, the adoption of advanced predictive models has become a necessity. This paper investigates the performance of four machine learning models for predicting the Moroccan stock market, namely Multilayer Perceptron (MLP), Recurrent Neural Networks (RNN), Convolutional Neural Networks (CNN) and the LSTM model, using historical daily stock price data. The results obtained, supported by the mean absolute percentage error (MAPE) values retained, show an alternation between MLP and RNN for the prediction and recognition of patterns in the time series studied. However, RNN stands out as a model offering, in the majority of cases, the best accuracy for anticipating stock price variations, underlining its effectiveness in this context.

Keywords: Prediction, Artificial Neural Network, CNN, RNN, MLP.

JEL Classification: G12

Paper type: Empirical research