The challenge was a Kaggle competition.

My kernel walks through some of the basic data preprocessing and preparation steps -

• Encoding categorical values
• Handling missing values
• Normalizing data
• Feature engineering by creating interactions
• Dimensionality reduction using PCA

Outlier detection is not part of the kernel. I checked for outliers but removing/imputing outlier values deterioriated performance on the test dataset.

For the modeling step, I used only linear regression. Kaggle used root mean squared logarithmic error to measure model performance. My model had a RMSLE of 0.01254 on the validation set and a RMSLE of 0.14918 on the test set.

Here is a link to my notebook on Kaggle. to my notebook on Kaggle.