Identifying Interesting Customers through Web Log Classification

Retention recommendation has been an important topic in e-commerce. Subjective classification is a potentially useful approach for both better understanding customer Web logs and identifying information actionable to customer retention. Subjective classification seems attractive because obtaining a large set of objective data, with labeling for training and testing, is often difficult. In particular, building a classifier when a training data set is small and possibly inaccurate is important. Thatʹs because decision makers find that identifying user purchase patterns from a Web log is difficult-thereʹs no direct relationship between Web log data and purchase patterns. Itʹs also difficult because the information in the small training data set is insufficient. A proposed method to build a classifier further selects a small subset of the training data set to build a classifier that possibly leads to high accuracy. This approach can help identify whether customers have purchase interest. The result of such classification provides actionable patterns and helps companies gain high customer retention.