SIDAKU: AI-Powered Financial Anomaly Detection System for Community Funds
An intelligent, microservices-based platform designed to detect anomalous financial transactions in grassroots community organizations using the Isolation Forest algorithm and a Human-in-the-Loop validation process.
The Challenge & Approach
SIDAKU (Sistem Informasi Deteksi Anomali Keuangan) is a comprehensive financial management solution tailored for community organizations (RT/RW). Traditional manual record-keeping often leads to human error and potential financial mismanagement. SIDAKU addresses this by implementing an automated, AI-driven monitoring system that ensures transparency and security.
The platform utilizes a robust microservices architecture comprising three core components:
- Mobile Application: A user-friendly Android interface for residents to input daily transactions effortlessly.
- Machine Learning Engine: Powered by the unsupervised Isolation Forest algorithm, it analyzes transaction features (amount, time, payment method, cash flow type) in real-time. It calculates an anomaly risk score and automatically flags suspicious transactions, providing dynamic heuristic explanations for why a transaction was flagged.
- Web Administrator Dashboard: A centralized hub for community leaders (Head of RT/RW and Treasurers) to oversee finances. It features a Human-in-the-Loop mechanism where administrators receive instant Firebase Push Notifications to validate or reject anomalies. This collaborative validation ensures high accuracy and creates a continuous feedback loop for the system.
Technologies
Links