An Optimal Classifier discovery for diagnosing the account health in financial firms and a study of classifier performance on imbalanced data release_c2bnbvp76vfklhn7npoqdgeoau

Abstract

<jats:title>Abstract</jats:title> Customers are the backbone for any financial companies. The behaviour of customer changes over time and they disconnect when the services do not meet their expectations. Earning loyalty of the customer by providing remarkable services and adopting retention strategies are mandatory to run any user centric businesses. In view of the growth perception there is need for company to identify the churn and avoid them in time. The mechanism for churn prediction requires to explore the insight of data. Machine learning algorithms are capable of mining the patterns present in the data and able to discriminate between classes with statistical learning. A Standard bank dataset has been considered for the study and exploratory data analysis performed to understand the nature of the data. Suitable data pre-processing is done and training data split from dataset has been used to build classifier models. The dataset was found to be imbalanced and by adopting appropriate sampling the dataset was balanced. Linear, Non-linear and boosting classifiers were built and their performances on test data are summarized. A comparative study on the classifier performance for both imbalanced as well as balanced dataset was observed and an optimal classifier for diagnosing customer account health has been suggested.
In application/xml+jats format

Archived Files and Locations