Back to work

PropSight

Clearer HDB valuation for homeowners.

Valuation, confidence, SHAP explanations, maps, forecasts, comparison and assistant support in one dashboard.

Role
Frontend, UX
Users
HDB homeowners
Focus
Model clarity
PropSight valuation dashboard showing property prediction, confidence, and market signals
30-second takeaway Not another listing site.

Designed for value, risk and market clarity.

Model clarity Visible model reasoning.

Confidence, SHAP and comparables make trust visible.

Build Dashboard depth.

ML outputs translated into scannable homeowner UI.

Problem

Price estimates feel opaque.

Homeowners need the factors, confidence, comparables and future signals behind the number.

Solution

A homeowner-first analytics dashboard.

Prediction, explanation, comparables, heatmaps, forecasts, MRT simulation and assistant guidance.

PropSight explainable valuation screen with feature contribution breakdown
Explainable valuation
PropSight spatial analytics map showing property value patterns
Spatial analytics map
PropSight forecast dashboard showing future property value trends
Deeper Analytics - Forecast

Assistant Layer

The assistant explains without deciding for the user.

The assistant turns dashboard context into plain-language property insight.

  • Plain-language feature contributions.
  • Grounded in structured prediction data.
  • Supportive, not prescriptive.
PropSight assistant explaining property analytics context
Explanation panel

Key Features

Explainable Valuation

Shows why a home receives its estimate.

Five-Year Forecast

Shows future value direction beyond today's price.

Spatial Heatmap

Reveals location patterns faster than tables.

MRT What-if Simulation

Explores transport impact on future value signals.

Contextual Assistant

Turns analytics into clear explanations.

Admin and User Tiers

Shows product architecture beyond the dashboard.

Design Decisions

Trust before automation.

Prediction first, explanation second, deeper analytics after context is clear.

  • Visible uncertainty and comparables.
  • Model logic built into the UI.
  • Charts and maps over dense tables.
  • Assistant framed as context, not advice.

Tech

Frontend for model output, maps, and analytics workflows.

  • Responsive dashboards
  • Prediction API integration
  • CatBoost model output
  • SHAP explanations
  • Gemini assistant
  • Spatial heatmaps
  • Forecast simulation
  • Role-based user tiers

Reflection

Explainability is product design.

Trust comes from showing reasoning, uncertainty, and context.