Skip to content

Application of learnings in the Machine Learning course , this project mainly gives first hand idea of elaborative exploratory data analysis performed on data sets and various advanced regressions models are used for predicting House Prices.

License

Notifications You must be signed in to change notification settings

naureen20/Predicting-Housing-prices-using-Advanced-Regression-Techniques.

Repository files navigation

Predicting Housing prices using Advanced Regression Technique

Objective is to perform a detailed EDA and estimate House prices using Ensemble technique.

Applying the learnings in the Machine Learning course, Detailed exploratory data analysis and various Advanced regression models XG-Boost, Ridge, Lasso, Elastic-net, Gradient Boost, SVR are employed.

STEPS 2 to 5 are for exploratory data analysis.

Predicting Sales Price of a given house in real estate market using various statistical analysis tools. Obtained the closest price that a client might sell their house utilizing machine learning. Presented in the form of iPython Notebooks.

Data is extracted from Kaggle Running code file can be found Here

Contents

  1. Loading and Exploring Data

Data is read from files and SalesPrice in training dara, ID for houses in test data are located

Exploratory Data Analysis

  1. Cleanup of data and Handling missing values

Many categorical features hold ratings which can be converted into numbers and missing values can be taken as 0.

  1. Analyzing Features having Object data type

Handling missing values manually and eliminating features with majority missing values

  1. Analyzing non-Categorical Features

Handling missing values manually by looking at features having close correlation with them and looking for possible collinearity and making new features from available features that have better correlation with Target prediction.

  1. Feature Engineering and EDA

Imputing if there are any remaining missing values, looking for outliers, skewness, observing feature importances and analysing correlation between each category of features together

Advanced Regression Models

  1. Creating Models with Grid Search

Ridge regression, Lasso regression, ElasticNet, XGBoost,SVR, Gradient Boost Regression models have been formed with Grid search for Best Hyper Parameters. Stacking all models and using XGBoost for 2nd level training improved accuracy of predictions.

Ensemble

  1. Ensembling and Obtaining result

For improving accuracy, all trained model predictions are Weight Averaged with most weightage for Stacking to obtain a Ensembled prediction of Sales Price

  1. References

About

Application of learnings in the Machine Learning course , this project mainly gives first hand idea of elaborative exploratory data analysis performed on data sets and various advanced regressions models are used for predicting House Prices.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published