Housing Price Analysis Using Regularized Linear Regression

We are required to build a regression model using regularisation in order to predict the actual value of the prospective properties and decide whether to invest in them or not.

The company wants to know:

1. Which variables are significant in predicting the price of a house, and
2. How well those variables describe the price of a house.

Table of Contents

• General Info
• Technologies Used
• Conclusions
• Running the code

General Information

In this assignment, we will

1. Use a hybrid combination of RFE and manual menthods for feature selection.
2. build a linear regression model with ridge and lasso regularization for predicting 'SalePrice', which is the final selling price of a property.
3. Find optimal regularization parameters for each of the methods using Grid search with K-Fold cross validation.
4. Use R-squared score on the test set to evaluate our model
5. Decide which model to go with.

Note that our main criterion of selecting a model would be $R^2$ scores, especially on testing data. Further consideration will be given to Lasso if it successfully selects fewer variables in the model.

Conclusions

The most important features to determine the price of a property are:

1. The overall material and finish of the house
2. First Floor Area
3. Second Floor Area
4. Basement Area

Technologies Used

• numpy - 1.23.1
• pandas - 1.4.3
• scikit-learn - 1.1.1

Running the code

• The data is present in train.csv.
• To understand the data, please read data_description.txt
• Download the repository, making sure that train.csv and Housing Price Analysis.ipynb are in the same folder.
• Now you can run the whole Notebook (BHousing Price Analysis.ipynb) from top to bottom.

Contact

Created by [@showman-sharma] - feel free to contact me!