Predicting customer behavior at an online retail store

Abstract

Nowadays, the predicting customer behavior is important to increase the selling product at an online retail store. Because the transaction of each customer looks like sequence, one of the approaches to data mining which is sequence mining is suitable for this case. In particular, the prediction model called Compact Prediction Tree + (CPT+) have suggested and applied to useful information discovery for customer behavior prediction in an online retail store. In the online retail store, there is customer services, especially the conversation tools between the store staff and customers. But it will limit the time of buying products. To keep services running 24/7 and combine with the recommend system, there use a Chatbot. In this research, a chatbot is made, that can learn customer behavior and then recommend suitable products. So, there are two things which are the bot and the API prediction. In the bot, it will handle the natural language processing (NLP). In the prediction API, we use the longest common subsequence (LCS) for finding the common transactions. Then, the prediction model using CPT+ integrated with the classification for speeding up and increasing the accuracy prediction will predict three next patterns given this transaction. Keyword: Sequence mining, Compact Prediction Tree+ (CPT+), Natural Language Processing (NLP), Longest Common Subsequence (LCS), Chatbot.