House_prices-prediction

Problem Statement

Help Alzar, the record keeper for finding lost details of 3.5k houses with the help of Machine Learning.

Installation

• Clone repository and run dataset_creation.py and then train_test_files_creation.py to create dataset.
  • Dataset is given in form of text files so preprocessing is required to convert them into csv file
  • Firstly dataset_creation.py extracts data from text files and make finaldataset.csv.
  • Then train_test_files_creation.py uses pandas library to split finaldataset.csv into testing and training dataset on basis of house prices.

Requirements

• This problem statement uses xgboost Regressor so it must be installed through either of these ways.
  • Using pip- pip install xgboost
  • Using conda- conda install -c py-xgboost
• Python2.7 is preferred for this project.

Usage

• Run dataset_creation.py followed by train_test_files_creation.py to create and split dataset from raw text files to processed csv files.
• Run features_analysis.py to visualize train and test dataset using functions of pandas dataframe. It helps in visualizing relations between features and target value with the help of histogram, scatter plots and Heat Map.

• Run algo.py for trying new features and feature selection and filling NaN values through interpolation.
  • After this data is ready to fit for different models.
  • This gives detail time and r2_score analysis after tuning hyperparameters of different types of regressions.
  • This will run cross validation across the training set on LinearRegression, LassoRegression, Ridge Regression which prints r2_score.

Results

• Finally run final_solution_xgboost.py to get the final results
• With the help of xgboost regressor we are able to achieve r2_score of 0.99512.
• solution.csv is also given in repository to match results of test dataset.
• xgboost with tuned parameters gives final r2_score of 0.99553 on test dataset.