A Comparison of Feature Selection Methods for Clustering Algorithms on Financial Transactions

In today's competitive business environment, it is of great importance to take action by anticipating the needs and potentials of customers. Customer segmentation is used to take fast and correct action specific to the customer's needs and potential. This problem, which is valid in the financial sector, is solved by using segmentation models obtained by transaction information, an indicator of investors' behavioral data. The suitability of the machine learning model, dataset, and features in segmentation models is essential. In this article, feature extraction, feature selection, and several clustering machine learning algorithms have been applied and compared using order data generated by a brokerage firm. This comparison serves as an example for future studies by both supervised and unsupervised feature selection methods applied to unsupervised clustering algorithms to determine the best combination of procedures for investor segmentation. The acknowledge of this study is to compare unsupervised clustering model results for both unsupervised and supervised feature selection methods.