Improving Sales Performance using Machine Learning Prediction Model

Sales prediction in Nigeria in previous years has been limited to traditional forecasting and expert judgment which are not always effective especially for large datasets. The aim of this study is to utilize machine learning algorithms to improve sales prediction using an historical dataset. The study investigated the application of machine learning models for predicting sales of medical supplies at Ekoistic Diagnostics, a medical facility in Nigeria using Regression analysis. We utilized a live dataset containing historical sales data and explored four machine learning algorithms: Support Vector Regression (SVR), Random Forest, and XGBoost, Multilayer Perceptron. These models were combined in a stacking ensemble. Predictions were generated from the models and hyperparameter optimization techniques were employed to enhance the performance of the models including the ensemble model. The evaluation metrics focused on Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) to assess the accuracy of sales predictions. Furthermore, all models exhibited significant improvement in prediction accuracy after undergoing hyperparameter optimization. The results revealed that the ensemble model emerged as the most effective model, achieving the lowest RMSE and MAE values of 1.38 and 0.97 respectively. Prior literature emphasized the high accuracy of these algorithms. Multilayer Perceptron, a deep learning model was used to improve the prediction of the ensemble model. From these findings, it was discovered that sales prediction using historical data is more of a time series problem.