Comparison of Machine Learning Algorithms in Predicting the Age Distribution Parameters of H&M Product Customers

Abstract: Over the past decade, the fashion industry has shifted towards dynamic product assortments with shorter life cycles. As a result, the role of the analysis of product sales has increased and become crucial for fashion retailers. Using H&M Group’s dataset on their product sales, I analyze and compare the performance of two machine learning algorithms in predicting the standard deviation and average age of product customers. These algorithms are Random Forest and Artificial Neural Networks. Since both dependent variables are estimated with noise, I fitted the models to the dataset with products having only a high number of observations. The paper describes in detail the performance of each of the algorithms and compares the accuracy. Random Forest works better in predicting the standard deviation of the age of customers per product, while ANN shows slightly better performance in predicting the average age of product customers. Both models perform better on a restricted sample and the performance of the models increases significantly while predicting the standard deviation of age.