This paper applies the social network theory to analyze the FOODMART sales dataset which is from a large supermarket company in the United States. We first measure the node degree distribution, the average path length and the clustering coefficient. The results show that the basket network accords with the characteristics of a small world network, but its topology is different from a number of actual large social networks. Its point degree distribution follows a Poisson distribution rather than a power-law distribution. We then try to find the cliques in the network and conclude that products which have same attributes connect more closely each other than the products which have different attributes. Furthermore, we also find that family members with similar age structure buy the similar products.

On the O2O platform, accurate forecasting of customer repeat purchase is critical to the management of these businesses. Based on the NBD model and SMC model, and taking nearly thousands of different categories of businesses on the O2O platform "koubei" as study object, we implement empirical research on customer repeat purchase forecast. The main contents include: (1) the introductions about NBD and SMC model; (2) forecasting the customer repeat purchase of different categories of businesses based on the models and analyzing the models fitting; (3) analyzing the models applicability based the results. The results shows that, in addition to the hot pot category, predicted results of the SMC model in each category of business is better than the NBD model, while the NBD model is more excellent in the forecast of the hot pot category.

No abstract