Quarterly Forecasting of Moroccan Gross Domestic Product (GDP) Using LASSO and Hybrid RNN-LSTM Models

Abstract

In this work, a new hybrid model based on deep learning techniques is proposed to forecast Morocco's quarterly GDP using several macroeconomic indicators. The proposed model integrates recurrent neural networks (RNN) and long-term memory (LSTM), which has the advantage of sequential learning and has been very successful in the past for predicting time series. The Pearson correlation coefficient and Lasso regression are used to select the important indicators linked to GDP. The introduced model is tested on quarterly indicators and GDP time series for Morocco from the Haut-Commissariat au Plan, covering a 43-year period from 1980 to 2023. Experimental results show that our hybrid RNN-LSTM model performs significantly better than individual RNN and LSTM models. It demonstrated better forecasting performance, achieving a minimum MSE of 0.001439, a MAE of 0.03068 and a MAPE of 0.04911. It also achieved a high coefficient of determination (R2) of 0.9983. These results highlight the model's precision and establish it as a dependable tool for economic decision-making and planning.