Analysis Of The Promotion Strategy For Pet Company

As one of the heated research propositions in academic research and business application, data mining and big data processing involves statistics, computational science and machine learning. This paper applied some basic method to accomplish tasks of predict specific properties and describe the potential relationships in data structure.Our project based on client data from a listed pet company. In this paper we discussed data cleaning and variable selection method for look-alike model in logistic regression analysis. Then we focused on explaining how to evaluate the accuracy and stability of models in real business issue. Based on the look-alike model we have built, we used decision tree and random forest to apply optimization and validation.For big data processing, we used SQL server to implement relational database and prepare the modeling dataset. We used SAS enterprise for data pre-processing and variable exploring. We cleaned data and created final variable list, iterated and tested model, evaluated the rationality and improved model performance based on data demand and business demand. We scored all the customer and divided them into different deciles, client will send promotion mails based on our predict value. Then we used Rstudio to implement decision tree, customer segmentation can describe the different dimensions of unknown customers, such as demographic, behavior, purchase intention and so on. So we can identify high value customer and customize precision marketing. In the end, we compared the accuracy of response model and random forest.