Model Performance Measures | Rank Order Table, KS, Gains Table & Lift Chart | k2analytics.co.in
ฝัง
- เผยแพร่เมื่อ 16 ม.ค. 2021
- Related Blog Link - 7 Important Model Performance Measures
1) Rank Order Table, Lift Chart & Gains Table
2) Kolmogorov-Smirnov test
3) Classification Accuracy
4) AUC-ROC
5) Concordance - Discordance
6) Gini Coefficient
7) Hosmer-Lemeshow Goodness of Fit
Blog Link: www.k2analytics.co.in/7-impor...
========================================================
Blog: k2analytics.in/blog
Resource: www.k2analytics.co.in/SpApp/r...
========================================================.
Facebook page: / ik2analytics
Instagram: / ik2analytics
Twitter: / ik2analytics
LinkedIn: / k2analytics
Meetup: www.meetup.com/ik2analytics
========================================================
About K2 Analytics:
We help students and working professionals build a career in Data Science / Machine Learning.
We provide training, consulting, and placement services.
For more details write to us at info@k2analytics.co.in - วิทยาศาสตร์และเทคโนโลยี
Awesome Video ...... Must see by all Data Science Enthusiasts
Thanks for this video. Really a great video on KS Statistics.
🌟🌟🌟Magnificent🌟🌟🌟 Very well explained. I don't think I am getting any problem in this again. Thanks Sir.
Easy to understand and really helpful.Informative and clear session.
Very well explained.
I'm creating marketing models in my org. right now and this is very helpful. Thanks.
Thanks for such a great content.
Informative and clear session. Easily understandable and well executed. Thanks for uploading Sir :)
The way you explained is just phenomenal. Will watch other videos too. Kudos 👏
Thanks
Best explanation ! Great explanation sir!
i really appreciate it.....u r really doing a good job..........thanku so much its very helpfull........
Very informative... Thank you
the best video
Thanks
Great session...really helpful :)
Thanks
very good session , nicely explained , but require the python codes , can we get the same ?
One more question you perform ks stats entire datasets......what about when we perform ks stats on traning and testing data if i perform what kind of information we will get
You have to perform KS separately on Train and Test sample. If the train-test KS numbers are comparable then there is no overfitting. If train KS is too high compared to test KS, then the model is overfit.
can you use any classification models
Target variable should be Binary
@@RajeshJakhotiaAIML thanks for reply when i use your code for decile analysis ,i got a decile table only 0 to 7...........but when i changed this code 10 to 12 here pd.qcut(dev.prob, 12, labels=False,duplicates='drop')
so i got decile table 0 to 9 why this is happeinng
@@abhishekprakash9803 Assume the values are 1,2,2,3,4,4,5,5,6,7
and I want to cut it in 10 deciles. I will get only 7 buckets instead of 10 as the number of distinct values are only 7.
@@RajeshJakhotiaAIML thanku so much for reply sir, actually i perform ks stats for my classification models ,i got a table just same as whateever u got.........i hav so much confusion, i just wanated to share my table in wtsup...so, can i get yout wtsup number if u don't mind........
Why probability is belongs to 1( responder class) only? Why not 0?
Probability is always mentioned for the Event of Interest (Favourable Event). If there are two teams A & B playing a match. You are interested in Team A winning the match, then, you will naturally express the probability in terms of Team A winning the match (and not Team A losing the match)
@@RajeshJakhotiaAIML but how model know our class of interest?
Sir, request if you can upload r codes?
Dear Souradeep,
Ok. Will do it soon.
See the video from 1:00 Hour onwards... and type the R code.
th-cam.com/video/23SQJ7tNFZE/w-d-xo.html
Thank you sir