An Optimized Extreme Learning Machine for Predicting Loan Default in Peer-to-peer Lending Based on an Enhanced Honey Badger Algorithm

Abstract

Loan status prediction aims to reveal credit risks that are difficult to quantify and analyze in P2P lending. Since the complex data types used by web platforms involve unstructured information such as text, this task can be viewed as a new challenge in P2P. This paper proposes a novel hybrid intelligence model for loan default prediction in P2P lending based on Extreme Learning Machine (ELM) and an enhanced Honey Badger Algorithm (EHBA). The improved nature-inspired meta-heuristic algorithm is used to tune the parameters of ELM to improve predictive performance. An empirical analysis using real-world data from the Lending Club in the USA, which contains 3,000 records and 24 different variables, shows that the proposed model can improve loan default prediction performance compared with KNN, ANN, RF, SVM, KSVM, ELM, GA-ELM, PSO-ELM, GWO-ELM, AOS-ELM, MPA-ELM, and HBA-ELM.