RFM Analysis Applied to Customer Segmentation Using Machine Learning Models

Abstract

Prediction, classification, and anomaly detection are made easier by a wide range of supervised and
unsupervised methods that are included in machine learning (ML). Customer churn prediction is one of the most
well-known applications for these approaches. Data scientists use a range of demographic, social, transactional,
and behavioral information and traits to predict customer switching. Regretfully, a large number of UK
businesses still lack the thorough and flexible customer data needed to conduct precise assessments. Because of
this, they frequently rely significantly on data generated by enterprise resource planning systems, which are
essentially transactional. As a result, companies are frequently only able to model and forecast using
transactional data and are reluctant to make large investments in marketing research or other sources pertaining
to customers. Companies are frequently restricted to using transactional data for modeling and forecasting, which
are typically not based on sophisticated methodologies like ML and recency, frequency, and monetary (RFM).
Therefore, the primary goal of the current effort is to offer a combination of RFM and ML analysis methods for
churn prediction with a focus on transactional data. The dataset was extracted from a service that searches for
internet retail datasets. Based on the information at hand, each customer's RFM scores are calculated. a churn
metric that shows if a consumer has completed a transaction within a certain amount of time. This report presents
a comparison of various strategies. Density-Based Spatial Clustering of Applications with Noise clustering and
K-means were employed. By the end of this research, it can be concluded that the division of customers into six
different clusters is a more easy and practical method