Tesla-arch/payroll-anomaly-detection

School management system with Ghana payroll and explainable anomaly detection (Laravel + React) # AI-Based Payroll Anomaly Detection System School management system for Ghanaian basic schools with automated payroll (SSNIT, loans and absence) and an **explainable, rule-based anomaly engine**. Built as a final-year project: Laravel REST API + React SPA. ## What it does - Staff, students, parent emails, attendance, class-teacher assignment (Headteacher / HR), and two-step leave approval (HR → Headteacher) - Payroll runs using 5.5% employee SSNIT, staff loans and absence — no PAYE deduction - Automatic scan after every payroll run - Anomaly dashboard with evidence, risk score, narrative, and recommended action - Payroll cannot be approved or marked paid while **critical** anomalies are still open ## Detection rules | Code | What it catches | | Ghost_No_User_Account | Salary without a linked user account | | Ghost_Inactive_Staff | Inactive employee on a payroll run | | Payment_Without_Attendance | Pay with no attendance (approved leave suppresses this) | | Duplicate_Period_Payment | Two payments for the same period | | Unusually_High_Salary | Gross pay above 1.5× grade/basic | | Ssnit_Mismatch | Employee SSNIT not 5.5% of capped basic | | Bank_Details_Missing | Missing bank name or account | | Loan_Over_Deduction | Deduction above outstanding loan | | Unauthorized_Allowance | Unauthorized or over-cap allowances | ## Requirements - PHP 8.3+ (developed on PHP 8.5) and Composer - Node.js 20+ - SQLite (default) or MySQL ## Open in your browser (GitHub Codespaces) GitHub hosts the source at Tesla-arch/payroll-anomaly-detection. The live app needs PHP and Node, so GitHub Pages cannot run it. Use **Codespaces** instead: 1. Open that badge (or on GitHub: **Code → Codespaces → Create codespace on main**) 2. Wait until the codespace finishes **setup** (first start can take several minutes; a 502 means the app is not running yet) 3. Open forwarded port **5173** — that is the school app 4. Sign in with a demo account below (password: `password`) If port 5173 shows …