A travel customer segmentation method based on improved RFM and k-means++

Customer segmentation is an important approach for customer relationship management, in which many methods are achieved by the Recency, Frequency and Monetary model(RFM) and clustering techniques. However, most methods based on the Recency, Frequency and Monetary model do not consider customer loyalty. In addition, these methods need to use all the historical data when updating the clustering, which has high data storage requirements. In this paper, a clustering method with a time window is proposed to solve these problems. The proposed method is divided into a feature selection stage and a clustering stage. In the feature selection stage, an important factor is considered in an improved Recency, Frequency and Monetary model, called the Length, Recency, Frequency and Monetary model(LRFM). In the clustering stage, a sliding time window is added to intercept the most recent data before the clustering. The proposed method differs from many other methods in that the model takes into consideration a new feature Length to identify customers more accurately, and uses the sliding time window to reduce data storage requirements. Based on the proposed method, the travel customer value analysis is explored on real customer anonymous transaction data. The experimental results show that the proposed method can classify travel customers into different groups effectively. The proposed method has a better clustering performance compared to other baseline algorithms.