Week 10: Polynomial Regression
Ishan B -
In my last post, I discussed the different metrics I used to analyze the performance of the linear regression (RMSE and score). I also discussed my belief for why the linear regression model performed poorly.
This week, I trained the polynomial regression model. The metrics gave numbers that corresponded to my beliefs, but were still much lower than expected. The score of the linear regression model was 0.002013770371433821 and the score of the polynomial regression model with 5 degrees (5 weights) was 0.022551080384189976. The RMSE of the linear regression model was 33303.971964744225 and the RMSE of the polynomial regression was 33206.212970112065. The polynomial regression model can be set to have different degrees, allowing for minor adjustments to its performance. From my testing, 5 degrees gave the best score. As the degrees got further away from 5, the score would continue to decrease. As I mentioned before, for these metrics, you want a score close to 1 and a low RMSE value. While the polynomial regression did indeed have a higher score and a lower RMSE, the polynomial regression did not perform well. It did perform much better than the linear regression as it had a score almost 10 times higher, but with a score of above 0.7 being the general consensus for good, the model was still quite far away from that. Regression models can be hyper-tuned to provide better results, but hyper-tuning is generally used once the model is already performing well, and used to optimize those models.
Comments:
All viewpoints are welcome but profane, threatening, disrespectful, or harassing comments will not be tolerated and are subject to moderation up to, and including, full deletion.