Statistic-Based CRM Approach via Time Series Segmenting RFM on Large Scale Data

CRM (Customer relationship management) plays a crucial role in detecting and gathering valuable users from externality and internality, where externality refers to customer relationship and internality is regarded as customer characteristics. However, conventional approaches based on RFM (recency, frequency and monetary) model have encountered with three challenges. First, traditional approaches that derive from survey data rather than objective large-scale data fail to apply the method in general scenario, Second, since there is several trial to experiment on RFM model changing over time, different segmentation of time leads to different results, Last, analysis of multiple characteristics either on externality or internality is sparse and separate, which betray the exploration purpose for CRM and make results unconvincing. To overcome the three limitations, a multiple statistic-based approach to value users via time series segmenting time interval of RFM on large scale data is proposed in the paper. In the aspect of telecom service data, we experiment on segmenting time interval methodologically for RFM model on data set more than millions of users. Besides, the most significant part there is formal mechanism to apply MCA (multiple corresponding analysis) on multiple characteristics for internality correspondingly with RFM for externality, leading to the deep relationships of users and their characteristics. Subsequently, we improve the traditional RFM model overtime from the different clustering steps on large-scale data.