Analyzing and Forecasting the Purchases Made on Black Friday

Abstract

Abstract— Things are sold at a substantial discount on the eve of Black Friday, resulting in sales that are 30 times larger than on normal flash sale days. Customers' data from purchases made on this day canbe examined, resulting in a quick declaration of their preferences for specific products. We looked at data that contained packets of clients, and also the factors that influenced their purchases and the amounts they spent. This data is analysed and forecasted purely for the purpose of providing clients with customized discounts on goods depending on individualpreferences and purchase budget. Four models were employed to forecast significant variations in trainingand test data (50:50, 70:30, 30:70), as well as a distinct sample training and testing dataset with two additional examples of prediction: xgboost, tfidftransform, both combination, and extra trees regressor. The two scenarios involve forecasting and analysing another dataset, as well as projected on the train data and testing data on a different testing data set. The dataset would be analyzed to learn about consumer behavior and trends of product the sale's popularity. For each of the five scenarios, the feature significance and benefit importance are displayed.All of the models' accuracy in various settings has been given in the manner of accuracy graphs and the accuracy findings have been displayed in the form of an RMSE score. Keywords— Black Friday, XGBoost, Accuracy.