How does SHAP model interpretation works?

[@StrikeasVR]

In series 1, does that mean one observation of the data? Or subset of data? Or as a whole? Thanks for this video

[@vivek4march]

The baseline value explained for series 1 talks about the shapley values which is the average of all predictions(whole data). In the diagram, each shapley value is an arrow that pushes to increase (positive value) or decrease (negative value) the prediction. These forces balance each other out at the actual prediction of the particular single data instance. I hope this will clarify your question. If you find my video useful then please subscribe to my youtube channel.

[@StrikeasVR]

@@vivek4march Subscribed 👍 and one more question arises in my mind is that if we know that feature is decreasing the probability of sucess, then why are we not removing that feature from our dataset?

[@vivek4march]

@@StrikeasVR For most of the real-world data, All features interact with each other in a certain way in a non-linear fashion and it may happen that a given feature will increase the prediction score for some of the data instances but for the other data instances it will decrease the data instance. SHAP will help you explain your model output so that you could make a decision or understand why my model has provided a given score or prediction. I hope this explanation will help you.

[@StrikeasVR]

@@vivek4march Yes, thanks a lot :) I think your videos are quite informative 😀 but these are not coming as top searches in youtube, maybe due to less keywords 😅