This is a competition held by Kaggle from August 30, 2016 to March 2, 2017.
In this competition, each team needs to predict the final price of the house based on 79 provided features by using some advanced regression techniques such as Random Forest and Gradient Boosting.
My kernel is motivated by some top-voted kernels in this competition which give good approach to the model. All reference models are arranged in reference_model folder. Some parts of the reference were modified according to my intuitive thinking.
Final result:
- Best score: 0.11565 (rmse)
- Best ranking: 393/4,915
- Personal summary: Even though I tried to implement many different models to see the improvement, the score was not improved that much. The model seems to be overfitting all the time. I guess to improve significantly, the data exploration and processing (outliers removal, dimension reduction and feature creation) should be focused.
- Further step: Check the kernel of leaderboard when they release their methods.