Mysore Real-Estate Analysis and House Price Prediction

The aim was to analyse the real estate market and predict the house prices in Mysore, Karnataka. A streamlit web-app is built for better visualization and prediction.

Data Extraction

• The data is extracted from websites https://housing.com and https://99acres.com
• Only houses and apartment prices were considered leaving behind empty lands and commercial properties

Data Cleaning

• The raw data is further cleaned resulting in five columns in the file combined_cleaned.csv are Location, Beds, Price, Area, PricePerSqft

Model Building

• The data features were one-hot encoded for location and Number of beds and fed to the model.
• Linear Regression performed the worst as expected for a small dataset
• gridsearchcv was performed and was inferred that decision tree algorithm performed the best among linear, lasso, decisionTree algorithms. Later, randomforestRegressor was used to increase the accuracy.

Deployment

Streamlit offers a great platform to host webapps easily.

REFERENCES

• https://www.geeksforgeeks.org/random-forest-regression-in-python/
• https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.RandomForestRegressor.html
• https://medium.com/analytics-vidhya/predicting-house-prices-using-classical-machine-learning-and-deep-learning-techniques-ad4e55945e2d
• https://housing.com/
• https://99acres.com