Partial Dependence Plot (PDP) in Python 

That means a negative 'direction' impact on the target variable because target var is substracted by its mean to show 'direction'.

@@dylanlee777 so all y values is: mean-prediction? Cant understand) Could you say pls as an example, why here 2:55 we have negative values of PD on MedInc graf? I thought that all this points is mean prediction on certain iterate of target variable.

@@vladvlad6399 You are right in the sense that zero price means the predicted price is somewhere around the mean of the price. Negative value means it is less than the mean value. In other words, if you assume all observations' MedInc is 2 and use your trained model to predict the house price, your average prediction would be below the average house price you had earlier. Hope that helps :)

@@dylanlee777 thanks so much!

Those are cut-off points that decile the data.

@@dylanlee777 Thank you!

github.com/Dylanlee777/5-Min-Machine-Learning/blob/master/7%20PDP.ipynb