For any queries, please raise an issue.
This project is a price prediction system for resendential property prices in Delhi NCR. Locations covered are Delhi, Ghaziabad, Noida, Greater Noida, Faridabad, Gurugram(Gurgaon).
Data is collected from the websites selling property in India, specially in Delhi. Python and Ruby scripts are use to collect data. Price is predicted using features like BHK, Latitude, Longitude, Area, Years Old, Bathrooms, Floor, Distance from Nearest Metro Station and IGI Airport.
- Scripting folder contains two scripts(one in Python, one in Ruby) that are used for website scraping.
- Airport.py: script for calculating distance of property from Indira Gandhi International Airport.
- metro.py: script for calculating distance of property from nearest Metro distance.
- lat_long.py: script for converting Address of property to latitude and longitude.
- FINALregressor.py: script used for predicting prices. It contains regression models - Ridge Regression, Bayesian Ridge Regression, kNN Regression, Decision Tree Regression, Multi Layer Perceptron Regression.
- FINALsvr.py: Support Vector Regressor for price prediction.
- FINALrandom.py: Random Forest Regressor for price prediction.
- xgboost_gridsearch.py: GridSearchCV to tune parameters for XGBoost Regression model.
- FINALxgboost.py: XGBoost Regressor for price prediction.
- numpy
- pandas
- geopy
- sklearn
- xgboost
| Model | Mean | Standard Deviation |
|---|---|---|
| Ridge | 0.5663 | 0.0246 |
| Bayesian Ridge | 0.5734 | 0.0213 |
| kNN | 0.7077 | 0.027 |
| Decision Tree | 0.63 | 0.0199 |
| Neural Net (MLP) | 0.6556 | 0.0321 |
| Random Forest | 0.7726 | 0.0246 |
| XGBoost | 0.7386 | 0.05217 |
| Maximum Depth | Min Child Weight | Mean | Standard Deviation |
|---|---|---|---|
| 3 | 1 | 0.7376 | 0.05975 |
| 3 | 3 | 0.73856 | 0.05217 |
| 3 | 5 | 0.73351 | 0.05993 |
| 5 | 1 | 0.73014 | 0.04219 |
| 5 | 3 | 0.73267 | 0.04214 |
| 5 | 5 | 0.73632 | 0.04516 |
| 7 | 1 | 0.72185 | 0.04562 |
| 7 | 3 | 0.71809 | 0.04393 |
| 7 | 5 | 0.71722 | 0.04400 |