InvestSmart: Personalized Commercial Real Estate Deal Discovery Engine

App Demo can be found here: https://drive.google.com/file/d/19g9VlpRqpXvio-09uGBdSsLu6ZiC61J0/view?usp=sharing

InvestSmart is an AI-driven tool designed to simplify commercial real estate investments by delivering personalized property recommendations with a focus on high return on investment (ROI). By integrating natural language processing, geospatial analysis, and machine learning, InvestSmart helps investors quickly discover well-priced commercial properties aligned with their preferences.

Key Features

• Personalized Recommendations:
  Uses a hybrid recommendation model combining Sentence-BERT embeddings and geo-spatial nearest neighbor searches to match user preferences such as location, budget, property size, and transit proximity.

• Deal Scoring:
  Employs Random Forest models to predict property price percentiles and calculates an unsupervised deal score to identify fairly priced, high-value opportunities.

• Large-Scale Dataset:
  Built on a dataset of over 1 million U.S. commercial properties from diverse public and private sources.

• Interactive Dashboard:
  Provides real-time visualizations of market trends, ROI estimates, pricing ranges, and key attributes to support informed decision-making.

How It Works

1. Users input preferences (e.g., location, budget, size, transit access).
2. The hybrid model finds top 5 listings using semantic similarity and spatial filtering.
3. Deal scoring highlights properties with the highest estimated ROI.
4. Results and trends are visualized through an interactive dashboard.

Performance Metrics

• Diversity score between 0.73 and 0.85, balancing relevance and variety.
• Fast execution with recommendation times ranging from 0.02 to 0.06 seconds even on 1 million+ listings.

Data Sources

• Commercial Property Listings: Dewey Data
• Transit Location Data: U.S. Department of Transportation Bureau of Transportation Statistics

Technology Stack

• Languages/Tools: Python, Pandas, Scikit-learn, GeoPandas, Sentence-BERT
• Models: Hybrid recommender (BERT + geospatial), Random Forest percentile prediction
• Spatial Search: BallTree for efficient transit proximity filtering

Future Enhancements

• Incorporate collaborative filtering to leverage user behavior.
• Integrate user feedback for continuous recommendation improvements.
• Add wishlist functionality for saved properties.
• Enhance price and ROI prediction models.
• Expand geospatial analysis for neighborhood trend insights.

Built for smarter investing, InvestSmart empowers commercial property investors with intelligent, data-driven insights.