Predicting Ethiopian gross domestic product using machine learning model

Abstract

The Gross Domestic Product (GDP) is an extensive indicator that reflects all of a country's economic activity over a certain time period. It calculates the total monetary value of all commodities and services produced within the country's borders. We employed a variety of algorithms and models to forecast Ethiopia's GDP using machine learning, including linear regression, Lasso regression, ridge regression, decision tree regression, random forest regression, gradient boosting regression, support vector machine regression, and neural network regression. Three phases comprise our investigation. First, we collect a dataset consisting of several economic statistics from the National Bank of Ethiopia. The gathered dataset is then preprocessed to ensure machine learning models can use it. Ultimately, we partition the dataset, designating 80% of it for model training and the remaining 20% for performance assessment. We employ a 5-fold cross-validation approach and consider evaluation metrics, including R-squared, mean absolute error, root mean square error, and mean squared error, to assess the efficacy of the model. Among all the models, Ridge Regression performs the best, achieving the lowest root mean squared error of 27,231,241,464.13, the highest R-squared value of 0.9950, a mean squared error of 1.06e+20, and a mean absolute error of 21,552,080,423.90. These results indicate that the model captures 99.5% of the variability in the data. Consequently, using the test dataset, the Ridge Regression model accurately forecasts Ethiopia's GDP.