House_Price_Iowa_Project

Repo for the Regression House Price competition project:

https://www.kaggle.com/c/iowa-house-prices-regression-techniques/data

Project Overview

• Exploratory Data Analysis, outliers identification and data cleaning.

• Modelling using Random Forest, CatBoost amd XGBoost Regressors.

• Hyperparameters tuning usings RandomizedSearchCV amd GridSearchCV.

• Test set predictions.

Code and Resourses used

Python Version: 3.8.2

Packages: Pandas, Numpy, Matplotlib, Seaborn, SKlearn, XGBoost, CatBoost

EDA: Exploratory Data Analysis

The EDA made shows how data is distributed and relation between different features. Following few highlights from the graphs dispalyed:

Data Cleaning

First use the pandas api to clean the Train dataset (df1) as follow:

• Fill missing numerical values with feature median
• Convert Object data into numerical
• Create a binary column for missing data with Boolean values

Then functionize the whole process with a preprocess_data(df) function that performs same transformations.

Model Building

• Split Data into train and test data

• Create fit_and_score(model) function to instantiate and compare accuracy from different estimators simultaneously.

• 3 different estimators: Random Forest Classifier XGBoost Classifier CatBoost Classifier

• Hyperparameter tuning using RandomizedSearchCV and GridSearchCV for the two best performant classifiers.

Make predictions

Once evaluated the best model which gives the lowest RSLME, use it to make predictions on test data.