Integration of MCDM and Machine Learning in B2B Customers Segmentation and Management

Zhu Yufei

Abstract

The recognition and retention of valuable customers are of great importance in a highly competitive business environment. In order to draw an accurate picture of their customers and meet their business requirements accordingly, a large proportion of Business-to-Business (B2B) marketers start dividing their target markets into discrete groups depending on their demand for certain products and services. Thus, methods and algorithms for Customer Relationship Management (CRM) and especially for accurate customer segmentation are developing rapidly these days, including big data analytics and machine learning techniques. In this paper, segmentation models were firstly built up in order to segment customers based their past purchasing records. Advanced machine learning algorithms including K-means and DBSCAN were used in both customer and supplier segmentations, combined with one of the multiple-criteria decision analysis (MCDM) method - analytic hierarchy process (AHP). A deep leaning module of Long short-term memory (LSTM) architecture was implemented to forecast future commercial behaviors of both customers and suppliers of the company. With insights to be gained from these proposed models, executives of the company will have better understanding regarding business potentials of current customers and allocate resources correspondingly, which will increase the satisfactory rate of customers and reduce product or resource waste and thus boost profits of the company to a considerable extent.