Hi Trevor, thank you for the wonderful tutorial. One thing that wasn't clear when reading this part
> Points off the Line: The two points that do not lie on the red diagonal line represent instances where the model's predictions do not match the actual outcomes.
Sorry for the confusion. Let me try to explain this better.
The points on the line at (False, False) and (True, True) show that the model correctly predicted instances of both non-purchases and purchases. These points are more important to me because they confirm the model’s ability to identify both positive and negative outcomes accurately. The confusion matrix from Step 12 also showed the model's overall accuracy was good, so I was looking at the Predicted vs. Actual Values Scatter Plot in the context of that confusion matrix.
So to summarize, I was looking to confirm that the model can reliably predict both True (purchases) and False (non-purchases) cases. The (True, False) and (False, True) points are less relevant in this context. (I should have explained this better in the article. My bad.)
If you see wrong predictions in your own data, you can follow these steps:
- Look Closer: Check the data points where the model was wrong. See if there are any patterns or reasons why those predictions failed. In this article, I noticed the model correctly predicted the things that mattered to me, so it was okay to move forward.
- Clean Your Data: Bad data can lead to wrong predictions. In the tutorial, we cleaned the data in Step 6, so I knew there were no missing values or data issues.
- Tweak the Model: Try adjusting the model settings, using different features, or even a different model type to improve accuracy. In the tutorial, we used the feature importance plot (Step 14) to understand which features were most influential and confirm our model was on the right track. If I were to see something fishy in Step 14, I would have gone back and tweaked more.
- Get More Data: More data can help the model learn better and make more accurate predictions. In the tutorial, we used a dataset that was big enough, but more data can improve model performance in real life. Do this step last though. The other 3 steps will usually give you the info you need before gathering more data.
Hope this helps! Let me know if you have any more questions.
Hi Trevor, thank you for the wonderful tutorial. One thing that wasn't clear when reading this part
> Points off the Line: The two points that do not lie on the red diagonal line represent instances where the model's predictions do not match the actual outcomes.
What do I do if that's the case?
Hi Linh,
Sorry for the confusion. Let me try to explain this better.
The points on the line at (False, False) and (True, True) show that the model correctly predicted instances of both non-purchases and purchases. These points are more important to me because they confirm the model’s ability to identify both positive and negative outcomes accurately. The confusion matrix from Step 12 also showed the model's overall accuracy was good, so I was looking at the Predicted vs. Actual Values Scatter Plot in the context of that confusion matrix.
So to summarize, I was looking to confirm that the model can reliably predict both True (purchases) and False (non-purchases) cases. The (True, False) and (False, True) points are less relevant in this context. (I should have explained this better in the article. My bad.)
If you see wrong predictions in your own data, you can follow these steps:
- Look Closer: Check the data points where the model was wrong. See if there are any patterns or reasons why those predictions failed. In this article, I noticed the model correctly predicted the things that mattered to me, so it was okay to move forward.
- Clean Your Data: Bad data can lead to wrong predictions. In the tutorial, we cleaned the data in Step 6, so I knew there were no missing values or data issues.
- Tweak the Model: Try adjusting the model settings, using different features, or even a different model type to improve accuracy. In the tutorial, we used the feature importance plot (Step 14) to understand which features were most influential and confirm our model was on the right track. If I were to see something fishy in Step 14, I would have gone back and tweaked more.
- Get More Data: More data can help the model learn better and make more accurate predictions. In the tutorial, we used a dataset that was big enough, but more data can improve model performance in real life. Do this step last though. The other 3 steps will usually give you the info you need before gathering more data.
Hope this helps! Let me know if you have any more questions.
Thanks again!
Trevor
That makes sense. Thank you for the detailed response!