Analysis of sales influencing factors and prediction of sales in supermarket based on machine learning technique

Abstract

At this stage, the development of Internet technology and computer hardware has generated a huge amount of data, and this huge amount of data has become the sustenance for the development of machine learning, which is often used in business and industry to predict customer behavior, product production cycles, and the expected amount of product sales.

There is a relatively well-established process for using machine learning to predict influencing factors and sales, but for small and medium-sized enterprises, they often face problems such as low data volume and unrepresentative data types, and the large data requirements become the threshold for using machine learning methods to help business activities.

The original data for this study was sourced from publicly available data from Alibaba's Tianchi, containing sales data from a small shop in three different branches. This thesis examines the influencing factors in terms of the relevance of the data, using a random forest regression approach to ranking the importance of features.

In order to predict sales, this thesis applies the pre-trained model in the field of neural network fine-tuning (migration learning) to the analysis of small data. First, using publicly available big data to pre-trained models in related fields, and then using the influencing factors obtained from the previous study to fit the pre-trained models. The model training results show a high degree of interpretability, which indicates that this thesis is very accurate in selecting the features of the influencing factors. The thesis also uses different machine learning regression models (including bagging, boosting and lasso) to make predictions and to verify the validity of the pre-trained machine learning model for small data