README.md

House-Price-Prediction

Overview

This is an ML-based application that predicts the house prices based upon the features provided (such as area, num of bedrooms, etc.).

For testing using sample inputs, click here

Model Used	Accuracy
Linear Regressor	81.705317 (Cost)
Ordinary Least Squares	92.9 (R-squared value)

File Description:

• Delhi_load.ipynb : Application interface for users.
• CSV File/ Delhi.csv : dataset
• Codes/ Delhi.ipynb : Main file with all processing stuff
• Codes/ ols_results_delhi.pickle : Pickled OLS model after training on dataset

Problem Statement

Given a set of features, predict the price of any given house in the Delhi region.

Dataset and libraries used

Dataset Summary

1. Reference link for dataset : click here
2. df.info()

For full image Click here

Libraries Used

1. Pandas
2. Statsmodel
3. Sklearn
4. Numpy
5. Seaborn
6. Matplotlib
7. Google.colab

Feature analysis

Outliers

1. Price

Raw Data Rectified Data

2. Area

Raw Data Rectified Data

3. Price per sq. foot

Raw Data Rectified Data

New Features

All the features provided can be reframed to the format : {Area, AttributeScore, Resale, LogPremium, Bedrooms}

1. Area = Floor area of the property
2. AttributeScore = An integer based on features like num of bedrooms, gym facility, etc
3. Resale = A binary value denoting if the propery is first hand usage (0) or a resale (1)
4. LogPremium = An integer value depending on the Price per sq. foot value
5. Bedrooms = Number of bedrooms in the property

Final Correlation Matrix :

Model selection

Two models were cosidered as most optimum ones, whose predictions are depicted below:

Ordinary Least Squares	Linear Regressor
92.9 (R-squared value)	81.705317 (Score)

Results

The trained OLS model gives an accuracy of 92.9 (R-squared value). The model summary for the same is provided below