# An Explainable Machine-Learning Pipeline for Financial Risk, Fraud and Compliance Screening

Project: FinGuard AI (MF-2026-07)  
Status: title and abstract working draft; not submitted or accepted  
Authorship and venue: unassigned pending contribution and editorial review

## Abstract

FinGuard AI investigates aI-driven financial risk detection, fraud prevention and intelligent compliance system. The proposed study frames the system around traceable data preparation, model or rule selection, measurable evaluation and responsible human review. The final manuscript will report only results reproduced from the approved repository and project evidence package.

## Keywords

FinTech; Risk; Applied AI; explainability; reproducibility; responsible AI

## 1. Introduction

Frame the domain problem, affected users, research gap and contribution. Do not describe a feature as novel until related-work review supports the statement.

## 2. Related work

Placeholder for peer-reviewed and authoritative sources. Each source must be recorded with DOI or stable URL, relevance note and the exact claim it supports.

## 3. Methodology

Document data acquisition, preprocessing, feature design, algorithm or system selection, baselines, train/validation/test strategy, cross-validation where appropriate, hyperparameters, stopping criteria and reproducibility controls.

## 4. Experiments

Pending verified experiment protocol. Pre-register research questions, comparison baselines, metrics, hardware/software environment and ablation plan before reporting results.

## 5. Results and discussion

Pending reproducible evidence. Negative and inconclusive results must be preserved alongside successful runs.

## 6. Limitations and responsible use

Current engineering limitation: Backend, Flutter, model and graph components remain unverified.

Document failure modes, bias, privacy, security, human oversight and conditions in which the system must not be used.

## 7. Conclusion and future work

Pending evidence review. Capstone development may extend the system after the current technical gate is cleared: Establish the canonical source tree and reproduce backend, mobile and model workflows.

## References

Placeholder — no citation will be invented.

## Reproducibility statement

Current verified outcome: React lint and production build passed; the frontend is connected for local review at port 5175.