Low R2 value in my regression problem

I am doing an uplift model in which i have my target variable -3 , 0 and 3 . And i have a multiple features like 100 in which there is not a good relationship between the target and the features but i have done a regression model . I’am beginner i would like to know why in doing that i have found a R2 value so close to 0 and when i print the model.intercept_ i have found 3 value of intercept !!! I would like to see the result of the model how it is …

In the other hand , i see that this is problem of classification because the target variable is not continuous and we would like to predict a probability so the best solution is doing a classification but also i would like to know why with régression i have found this result and how can i interpret it ?

I understand you should try building a logistic regression model with dependent variable (Y) having values (-3,0,3), actually it will be a multi-logistic Model. For evaluation of logistic model, you should try looking at the decile report, confusion matrix etc and not relay on R2.

To get better relationship between target and feature, try some transformation on the features (e.g. sin, tan, log, exp etc). Since linear regression will only try to capture the linear relationship, transformation or interacting two variables are found useful

To answer why you got 3 intercept, it would be great if you could share the details algorithm and parameters you are using.

I have done a different classification model and it works fine but i would like to understand why i have found this result with RandomForestRegressor and XgbRegressor i would like to interpret the result of the Low value of R2 . in regression my model all the time predict a value near to 0 and the mean is like 0.004 so the model it predict like the mean that’s why i get a R2 low value but why it is predicting always value close to 0 i undersampling The class 0 to 800 sample to get a balanced dataset but always it is predicting value close to 0 .

here is the visualisation of the actual value vs the predicted value


© Copyright 2013-2019 Analytics Vidhya