Prediction of Consumption Behavior Based on Data Mining

Data can record all kinds of information, including the behavior of a business. Therefore, mining valuable information from massive data and using big data algorithms to promote marketing have become the focus of enterprises. Rossman is the largest grocery store in Germany. In order to improve the management level, the purpose of this study is to predict the daily sales volume of this store, and to investigate the prediction effect of different algorithms. According to the provided data from Rossman many factors have to be taken into consideration, such as promotions, holiday and so on. The main method used is data analysis and modeling prediction through python. The data is from Rossmann store sales of Kaggle, and the data volume reaches more than 1 million. There are three tables: train.csv, store.csv, test.csv. Through the modeling and prediction of python, we find that there are differences in the prediction result and efficiency of different models, among which GBDT algorithm has the best prediction effect and is considered to be adopted. At the same time, among the many influencing factors, whether the store is open or not, promotional activities and the distance of competitors have a more important impact on the turnover.