An Efficient Multi-layer Ensemble Framework with BPSOGSA-Based Feature Selection for Credit Scoring Data Analysis

摘要: Credit scoring is extensively used by credit industries and financial institutions for decision-making. It a way to assess the risk associated with an applicant based on historical data. However, data may have large number of redundant noisy features which could affect performance models. Main focus this paper develop hybrid model combining feature selection multi-layer ensemble classifier framework improve prediction model. The proposed uses binary particle swarm optimization gravitational search algorithm (BPSOGSA) five heterogeneous classifiers. A novel V-shaped transfer function BPSOGSA also designed effective selection, transform continuous space space. Also, fitness calculate value each agent. Further, along aggregation generalized convex function. validated using Australian, German-categorical, German-numerical Japanese datasets. experimental results all datasets demonstrate that outperforms other methods such as random forest frameworks, namely majority voting, layered weighted voting in terms accuracy, sensitivity, G-measure ROC characteristics.