Modeling Funding for Industrial Projects Using Machine Learning: Evidence from Morocco

Abstract

Moroccan manufacturing companies investing in the metallurgical, mechanical, and electromechanical industries sector are among the contributors to the growth of the national economy. The projects they are awarded do not have the same specific features as those of operating activities within other companies. They share several common features, making them particularly complex to fund. In such circumstances, supervised machine learning seems to be a suitable instrument to help such enterprises in their funding decisions, especially given that linear regression methods are inadequate for predicting human decision making as human thinking is a complicated system and not linear. Based on 5198 industrial projects of 53 firms operating in the said sector, four machine learning models are used to predict the funding method for some industrial projects, including are decision tree, random forest, gradient boosting, and K-nearest neighbors (KNN). Among the four machine learning methods, the gradient boosting method appears to be most effective overall, with an accuracy of 99%.