[FINCN-252] Machine Learning Scorecard for Credit Risk Assessment Phase 5

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: fineract-cn-ml
• Labels: gsoc2021 mentor
• Assignee: Unassigned
• Reporter: Ed Cable
• Created: 11/Mar/21 01:15
• Updated: 18/Jun/22 00:15

Description

Mentors

• Lalit Mohan S
• Fintecheando

Overview & Objectives

Financial Organizations using Mifos/Fineract are depending on external agencies or their past experiences for evaluating credit scoring and identification of potential NPAs. Though information from external agencies is required, financial organizations can have an internal scorecard for evaluating loans so that preventive/proactive actions can be done along with external agencies reports. In industry, organizations are using rule based, Statistical and Machine learning methods for credit scoring, predicting potential NPAs, fraud detection and other activities. This project aims to implement a scorecard based on statistical and ML methods for credit scoring and identification of potential NPAs.

Description

The approach should factor and improve last year's GSOC work (https://gist.github.com/SupreethSudhakaranMenon/a20251271adb341f949dbfeb035191f7) on Features/Characteristics, Criteria and evaluation (link). The design and implementation of the screens should follow Mifos Application standards. Should implement statistical and ML methods with explainability on decision making. Should also be extensible for adding other functionalities such as fraud detection, cross-sell and up-sell, etc.

Helpful Skills

JAVA, Integrating Backend Service, MIFOS X, Apache Fineract, AngularJS, ORM, ML, Statistical Methods, Django

Impact

Streamlined Operations, Better RISK Management, Automated Response Mechanism

Other Resources

2019 Progress: https://gist.github.com/SupreethSudhakaranMenon/a20251271adb341f949dbfeb035191f7

https://gist.github.com/lalitsanagavarapu
 

Issue Links

• links to: GitHub Pull Request #1811
• links to: Scorecard REST API module
• links to: Web-app Integration

Sub-Tasks

• Integrate scorecard configurations: Resolved
• Import rule based in the API.: Resolved
• Import statistical scoring in the API: Resolved
• Import Machine Learing scoring: Resolved
• Create ML pipeline to auto-update ML scoring: Resolved
• Implement and integrate scoring with H2o.ai: Open
• Research and implement a federated learning model: Open
• Integrate and Deploy Federated model: Open