Stock Market Forecasting Using Long Short-Term Memory Recurrent Neural Networks: A Case Study of Moroccan Companies

This paper presents a study on forecasting daily stock exchange closing prices for four prominent Moroccan companies over four years, from January 2020 to January 2024. We meticulously prepared the data, addressing missing values and anomalies. Our approach is a recurrent neural network architecture based on LSTM layers. We tested the model and compared the forecasts to validate our strategy. For model construction, we employed specific architecture and parameters, optimized using the Adam optimizer, and evaluated its performance using mean squared error (MSE) and mean absolute error (MAE) metrics. The results demonstrated promising performance, with the model achieving an overall validation score between 0.0412 and 0.1230 (MSE) for the four companies. A visual comparison between forecasted and actual values confirmed the model's effectiveness.