Designing hybrid models between neural networks and multiple linear regression to predict the impact of energy carrier prices on the food costs of living for Syrian families

Abstract

The research aims to build hybrid models by combining the multiple regression model and the neural network model to predict the impact of energy carrier prices on the food costs of living for the Syrian family. The neural network model is one of the methods capable of learning and adapting to any model, and does not require assumptions about the nature of the time series.

The researcher used data on the monthly prices of diesel, gas, and gasoline and the monthly food cost of living calculated for the Syrian family for the period extending from (1/2020 - 5/2023), and by processing the data in the statistical program SPSS, the researcher arrived at building four hybrid models between the two methods, and through comparison between Hybrid models Using the evaluation criteria MSE, RMSE, MAE, MAPE, the researcher concluded that the hybrid model C was superior using previous observations and residuals and the estimated values ​​of the multiple linear regression model compared to the hybrid models.

Based on the hybrid model C, the monthly food costs of living for the Syrian family were predicted for the next 12 months, from June 2023 to May 2024, which showed an increase in the monthly food costs of living for the Syrian family in the coming periods