Another use example was applying the RFM (Recency, Frequency, Monetary) strategy to explore client segments, demonstrating how to interactively display and Customer Segmentation as well as employ Guided Insights to pinpoint high-value customer categories. We will choose a more sophisticated strategy in this use case that makes use of machine learning and the clustering method.
In order to examine and group a set of points in a way that keeps the gap between points in a cluster small (within the cluster distance) and the distance between points from other clusters big, we will use an unsupervised machine learning clustering model (inter-cluster distance). Unsupervised algorithms come in a variety of forms (e.g., hierarchical, probabilistic, overlapping), but K-Means clustering is the most often used method.
Predictive models can be trained, evaluated, and used in Tellius' robust machine learning layer, which is based on Apache Spark and the Spark ML open-source framework. A model can be trained using one of two methods provided by the platform. One is called AutoML, where the user chooses a target variable and depends on Tellius to choose the proper algorithm, carry out feature transformation, and fine-tune the parameters. Another method is known as Point-n-Click, which gives users more flexibility over model selection and hyper parameter tuning strategy. For the construction of our model, we'll use a point-and-click method.
In this post, we provide an introduction of two techniques to customer segmentation analysis that a marketing team may use using Tellius's Augmented Analytics platform: a classic RFM approach and a Bisecting K-Means clustering predictive model. Other eCommerce and retail applications can benefit from Tellius, or you can download our Guide to AI-Driven Analytics for eCommerce.
