Overview

This pull request delivers 5 end-to-end full-stack advanced financial features for FinWise Mini App, introducing algorithmic financial forecasting, discrete-event simulation, statistical anomaly detection, automated subscription tracking, and a deterministic natural language query engine.

All features are fully integrated across the stack with Type-Safe Backend Services, Mathematical Engines, Automated Tests, Claymorphism UI Components, TanStack Query Hooks, and Bilingual Localization (VI/EN).

Upgrades Breakdown

1. Cash Flow Runway & Budget Forecaster (/forecast)

• Backend:
  • Exponential decay weighting for daily burn rate ($\lambda = 0.04$).
  • Mathematical runway projection: $\text{Runway} = \frac{\text{Current Balance}}{\text{Weighted Burn Rate}}$.
  • Data sufficiency validation based on sample variance & count.
  • Active budget exhaustion date prediction and daily spending cap recommendations.
• Frontend:
  • Interactive forecast horizon switcher (14, 30, 60, 90 days).
  • Runway gauge card with finite countdown or positive cash flow indicator.
  • Risk badges (LOW, MEDIUM, HIGH, CRITICAL) with localized date formatting.
  • Claymorphism loading skeleton & error retry handling.

2. What-if Financial Simulation Sandbox (/simulations)

• Backend:
  • Discrete-event simulation over a 1–24 month horizon.
  • Support for custom recurring & one-time perturbations (income & expense shifts).
  • First-month deficit detection & saving goal collision evaluation.
  • Built-in scenario presets (New Job, Major Purchase, Inflation Surge, Aggressive Saving).
• Frontend:
  • 1-click preset loader and interactive month duration slider.
  • Custom cash flow adjustment builder with real-time currency formatting.
  • Comparative trajectory breakdown (Baseline vs. Simulated) and goal delay alerts.

3. Multi-dimensional Anomaly Detection (/anomalies)

• Backend:
  • Modified Z-Score using Median Absolute Deviation (MAD): $$M_i = \frac{0.6745 \cdot |x_i - \tilde{x}|}{\text{MAD}}$$
  • Multi-factor evaluation: Velocity bursts ($\ge 3$ txns / 20 min), off-peak transactions (02:00–05:00 UTC+7), and wallet balance ratio triggers.
  • Integration with NotificationService for real-time alerts.
• Frontend:
  • Pre-purchase interactive transaction evaluator.
  • Flagged history cards with severity badges (CRITICAL, HIGH, ELEVATED) and tagged reason codes.

4. Auto Subscription Discovery (/subscriptions)

• Backend:
  • Merchant token normalization and string cleaning.
  • Periodic regularity autocorrelation scoring ($R \ge 0.75$).
  • Price drift detection ($>3%$ increase between billing cycles).
  • 1-click conversion endpoint creating active tracked Reminder entries.
• Frontend:
  • Discovered subscription cards with frequency badges (Monthly, Weekly, Yearly).
  • Amber warning banners for detected subscription price hikes.
  • 1-click "Add to Reminders" conversion action flow.

5. Hybrid NL to Deterministic DSL Query Engine (/query)

• Backend:
  • Vietnamese natural language tokenizer and slot extractor.
  • Compiler generating type-safe Abstract Syntax Trees (AST).
  • Parameterized database aggregation executor (preventing raw SQL injection risks).
• Frontend:
  • Natural language input bar with quick prompt chips.
  • AST filter inspector tags.
  • Verifiable summary card, category breakdown groups, and detailed matching transaction list.