Abstract: This paper presents a comprehensive study and analysis on customer segmentation using banking data, aiming to enhance the accuracy and effectiveness of segmentation techniques. The rationale for conducting this research lies in the growing need for personalized services in the banking sector, where understanding customer behavior and preferences is crucial for strategic decision-making. The problem addressed in this study is the challenge of accurate and meaningful customer segmentation, considering the intricate patterns and complexities inherent in banking data. Conventional segmentation methods like k-mean, improve k-mean, and fuzzy c have been widely applied; however, their limitations in handling non-linear and complex data structures necessitate the exploration of more advanced techniques. The methodology employed involves a multi-faceted approach to address the segmentation challenge. Initially, conventional methods such as k-means, improved k-means, and fuzzy c-means are applied to the banking data to establish a benchmark for comparison. These methods are effective for relatively simple data distributions but may fall short in capturing intricate patterns. To address this, a novel approach utilizing spectral clustering is proposed. The proposed method, spectral clustering, leverages the spectral properties of the data to capture underlying structures and relationships. Unlike traditional methods, spectral clustering can effectively identify non-linear and complex patterns in the data, making it suitable for the nuances of banking customer behavior. Through experimentation and analysis, the proposed method's performance is evaluated against the established benchmarks, showcasing its potential to yield more accurate and meaningful customer segments. This research contributes to the field of customer segmentation in the banking sector by highlighting the limitations of traditional methods and introducing a novel spectral clustering approach. The customer segmentation using Neural Network and Spectral Clustering performs well compared to the previous research our proposed system gives an accuracy of 99.54 and also gives the best Gini obtained.
Keywords: Customer Segment, K–Means, Machine Learning, Banking Profiling, Spectral Clustering.
|
DOI:
10.17148/IJARCCE.2023.12805
