Customer Segmentation Using Machine Learning

Abstract

In the face of huge competition among business organizations and with the modern economy, organizations have a huge number of customers hence, it is required to mine customer resources in order to achieve targeted measures for different types of customers and provide them with the services they want. Since it is a complex task to treat each and every customer separately, this can be achieved through customer segmentation by grouping the customer base into refined customer groups based on their similar needs and behaviors. This is very crucial to the development of the organization as it enables to improve their customer satisfaction further the implementation of customer segmentation leads to gaining new customers and this will be beneficial to extract a higher value from the existing customers through maintaining a better customer relationship. When the segmentation system is efficiently designed, customers of one segment have similar interests and behaviors, and they will most probably respond similarly to the situations where the elements of the marketing mix for example pricing, promotions, and for sales channels. This will be very significant for financial organizations to improve profit-driving opportunities targeting each unique customer group. This research project focuses on a banking sector dataset and this study explores multiple machine learning models for segmenting customers and for identifying the most valuable customer group according to the customer payment behaviors. I have used a hybrid approach utilizing both supervised learning model and unsupervised machine learning model in this study. The banking dataset was analyzed and processed in order to train the machine learning models. The customer base was segmented into four customer segments and each customer group was analyzed to recognize the most valuable customer group. And the output of the trained clustering model was used to develop the customer segmentation prediction system using supervised machine learning models to predict the customer group of the user input customer. The customer dataset was trained using six different unsupervised machine learning algorithms and the obtained customer segments from the best-performance machine learning model were used for training the prediction model supervised machine learning algorithms were trained on the clustered dataset and the best-performing model was selected to build the prediction model. The prediction model guarantees an accuracy of 0.97 along with the other performance metrices.