THE STUDY OF THE PURCHASE PROPENSITY OF THE USER OF THE ONLINE STORE ON THE BASIS OF TECHNICAL DATA ON VISITS OF VISITORS TO THE ONLINE STORE

Abstract

The paper describes the development of an automatisation service for marketing campaigns on the basis of web usage data mining. The source web usage data were obtained from Yandex.Metrika service. On the basis of exploratory data analysis it was decided to divide the work into the following stages: analysis, building an assembly of models, assesing the efficiency of the models and selecting the best one, and publishing the web service with the purpose of its further usage in the Internet shop CMS within marketing campaigns. The data mining was conducted on the basis of Loginom Analytics Platform which allows to create analytic models as well as to present them out as web services. For the data mining 3 approaches - association rules, time series and scoring - were used. The scoring model demonstrated the best results, it showed excellent results on test data, with that the coefficient expert analysis confirms its correctness and applicability. In case of necessity the developed analytic model can be learned on web usage data of any Internet shop provided by Yandex.Metrica service.

Keywords

• IAD
• scoring
• association rules
• logistic regression
• time series