Bipolar Fuzzy Hypersoft Set with Heuristic Search Based Customer Retention Prediction Model in Financial Sectors

From a philosophical viewpoint, the theory of neutrosophic set (NS) is a simplification of the concept of Fuzzy Set (FS) and intuitionistic FS (IFS). An NS is illustrated by a truth, an indeterminacy, and a falsity membership functions and every membership degree is an actual standard or a non-standard sub-set of the non-standard unit range of] −0, 1+ [. Customer churn is when clients stop utilizing a company's service or product. Moreover, it is also named customer retention, which is vastly significant metric as it is much less costly to keep the existing customers than to obtain novel customers. The prediction of churn plays an essential part in customer retention because it forecasts clients who are in danger of leaving the organization. In the banking sector, the customer attrition arises when clients quit utilizing the services and goods provided by the bank for some time. So, customer churn is vital in today's economic banking industry. This study proposes a Leveraging Bipolar Fuzzy Hypersoft Set with Heuristic Optimization Algorithms-based Customer Retention Prediction (BFHSS-HOACRP) technique in financial sectors. The BFHSS-HOACRP technique applies optimized techniques to predict the customer retention behavior in the industry of bank. Initially, the mean normalization technique is utilized in the data pre-processing stage to prepare raw data into a suitable format for analysis and modeling. For the selection of feature process, the grasshopper optimization algorithm (GOA) method is employed to identify and select the most relevant features from an input data. In addition, the proposed BFHSS-HOACRP technique implements bipolar fuzzy hypersoft set (BFHSS) method for the classification process. Additionally, the spider monkey optimization (SMO)-based hyperparameter selection process is performed to optimize the classification results of BFHSS model. The efficacy of the BFHSS-HOACRP approach is examined under the bank customer churn prediction dataset. The comparison analysis of the BFHSS-HOACRP approach portrayed a superior accuracy value of 95.41% over existing techniques.