Subscribe to DSC Newsletter

Hi, 
I have a question which maybe simple as I am not very experienced... 
There is one target field and several attibutes in my dataset. I have created a model and calculated attibute importances. 
Now, my target value is changed over some time period and I would like to calculate/verify which attributes' changes had more contribution? 

Thanks

Views: 145

Reply to This

Replies to This Discussion

Hi Erdem,

create crosstabs (pivot tables) TARGET vs. ATTRIBUTES and compare them between periods. Crosstabs tell you how much has distribution & effect changed. 

If you have continuous variables, start with binning into 5 bins on training data. Then bin new data with the same boundaries. And only then compare crosstabs.

Good luck.

Jozo

Assuming no inter-dependency among the independent variables, checkout the correlations between target and each characteristic in both earlier time frame and current time frame. The variables that have drastic change in correlation over both the time frames could be contributing the most to the shift in target. For a more degree of accuracy, avoid correlations and draw scatter plots in both the time frames and compare the trends.

If your target is binary, plot the event rates for each attribute of all characteristics and compare the plots in both the time frames.

 

Hope this helps!

RSS

On Data Science Central

© 2019   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service