House Prices Advanced Regression

Hi! I'm Ali (PmD) PourMohammad! This Project is just a warm-up and mostly has an educational and training purpose.

Maybe it will be useful for those who are new to this field!

files:

• data_description.txt has information about data
• train.csv and House.csv are for train and evaluate a model
• test.csv doesn't have outcomes and it's for evaluation on the kaggle site
• and housePricePrediction.ipynp has codes that I've written to solve this problem

In this notebook I applied some preprocesses like:

• Drop columns with high outliers
• Drop some data with outlier values
• Drop ID col
• Convert all 'NaN' Values in 14 categorical Columns to 'NOT' (For example : "No Basement" stored as "NaN")
• Drop columns which have most null values
• Convert all categorical columns to numeric:

  In this data, we have three type of categorical columns: 1)Binary 2)Nominal and 3)Ordinal

  I encode ordinal with OrdinalEncoder or .map() func and encode Binary and Nominal with Get_Dummy

  *** Also, ‘MSSubClass’ Column has nominal values but encoded with numbers that may harm model, so I’m going to encode that with Get_dummy too.

• Find correlated columns and drop one of each
• get_dummy all remain categorical columns

And finally I've trained a GradientBoostingRegressor model and get 0.945 score.

Thanks for reading ;)