Sales Forecasting Based on CatBoost

Abstract

Sales forecasting is a vital technology nowadays in the retail industry. With the help of advanced machine learning and deep learning algorithms, business owners can accurately predict the sales of thousands of products and make optimum decisions based on them. In this paper, we proposed a sales forecasting system based on CatBoosting. The algorithm is trained on the Walmart sales dataset, by far the largest dataset in this field. We performed effective feature engineering to boost prediction accuracy and speed. In the experiments, our model outperforms traditional machine learning methods like Linear Regression and SVM, reaching an RMSE of 0. 605. Our method doesn't need as much finetuning as other methods thus improving its generalization ability on other custom datasets, expanding its potential use.