ChatHDB - Property Valuations
A web application providing accurate HDB property valuations by integrating fine-tuned XGBoost models with real-time economic data (GDP, inflation) and sentiment analysis from news, forums, and Google Trends.
The Problem
Property valuation in Singapore's HDB market is complex, influenced by numerous factors including location, flat type, remaining lease, and broader economic conditions. Traditional valuation methods often fail to capture real-time market sentiment and economic shifts, leading to inaccurate predictions.
The Solution
ChatHDB combines fine-tuned XGBoost machine learning models with real-time economic indicators and sentiment analysis to provide accurate property valuations. The system ingests data from multiple sources including government APIs for GDP and inflation data, news articles, property forums, and Google Trends to understand market sentiment.
Users can interact with the system through a conversational interface, asking questions about property values, market trends, and receiving personalized recommendations based on their requirements.
Technical Architecture
The machine learning pipeline includes feature engineering from historical transaction data, economic indicators integration, and NLP-based sentiment scoring from multiple text sources. The XGBoost model was fine-tuned with extensive hyperparameter optimization to achieve high accuracy on HDB resale price predictions.
The web application is built with modern technologies, featuring a clean interface for users to input property details and receive instant valuations along with confidence intervals and market insights.
Key Features
The platform offers instant property valuations, historical price trend analysis, sentiment-adjusted predictions, and comparative market analysis. Users can explore how different factors affect property values and make informed decisions based on comprehensive data-driven insights.