FinXCredit: An AI-Powered Alternative Credit Risk Assessment System for India's Gig Economy Workers

FinXCredit demonstrates that AI-powered alternative credit scoring is both technically feasible and practically valuable for India's 400 million gig and informal economy workers. By training an XGBoost classifier on a novel 100,000+ record pan-India dataset, implementing a 5-factor gig-worker-sensitive heuristic scoring engine with empirically validated differentiated weights, and deploying SHAP-powered explainability compliant with RBI Digital Lending Guidelines, FinXCredit bridges the gap between advanced fintech capabilities and academic reproducibility. The system achieves 78.6% accuracy and 0.85 ROC-AUC, outperforming logistic regression by 11 percentage points and the bureau score proxy by 8.5 points. The KS statistic of 0.55 exceeds the 0.40 industry threshold for production credit models. All 16 security test scenarios and 6 cross-browser environments pass. The critical finding — that transaction activity is 2.3× more predictive for gig workers than salaried workers, validated by a 29.7-point default rate differential — constitutes a novel contribution to alternative credit scoring methodology that is directly applicable to the broader financial inclusion agenda in developing economies. Future work includes: integration with live RBI Account Aggregator APIs (Sahamati/Setu); real UIDAI Aadhaar OTP and NSDL PAN verification; LSTM/Transformer-based temporal income modeling; algorithmic fairness auditing using demographic parity and equalised odds metrics; and MLOps pipeline deployment for continuous model retraining as real loan outcome data becomes available.

[1]  NITI Aayog, "India's Booming Gig and Platform Economy: Perspectives and Recommendations on Future of Work," Government of India, New Delhi, 2022. [2]  W. Fair and E. Isaac, "A method of estimating the probability of credit granting," U.S. Patent 2 885 672, May 12, 1959. [3]  K. Brevoort, P. Grimm, and M. Kambara, "Data Point: Credit Invisibles," Consumer Financial Protection Bureau, Tech. Rep., 2015. [4]  D. Bjorkegren and D. Grissen, "Behavior Revealed in Mobile Phone Usage Predicts Credit Repayment," The World Bank Economic Review, vol. 34, no. 3, pp. 618–634, 2020. [5]  World Bank, "Identifying the Unbanked: Alternative Data and Machine Learning for Financial Inclusion," Financial Sector Advisory Center, 2017. [6]  R. Hao, Z. Wang, and Y. Chen, "Credit scoring with social network data," Marketing Science, vol. 35, no. 2, pp. 234–258, 2016. [7]  S. Lessmann, B. Baesens, H. Seow, and L. Thomas, "Benchmarking State-of-the-Art Classification Algorithms for Credit Scoring," European Journal of Operational Research, vol. 247, no. 1, pp. 124–136, 2015. [8]  T. Chen and C. Guestrin, "XGBoost: A Scalable Tree Boosting System," in Proc. 22nd ACM SIGKDD Int. Conf. Knowledge Discovery Data Mining, 2016, pp. 785–794. [9]  I. Brown and C. Mues, "An Experimental Comparison of Classification Algorithms for Imbalanced Credit Scoring Data Sets," Expert Systems with Applications, vol. 39, no. 3, pp. 3446–3453, 2012. [10]  Y. Xia et al., "A Boosted Decision Tree Approach Using Bayesian Hyper-Parameter Optimization for Credit Scoring," Expert Systems with Applications, vol. 78, pp. 225–241, 2017.