We play with one to-scorching security and have now_dummies toward categorical parameters on the software studies. On the nan-viewpoints, i have fun with Ycimpute collection and assume nan opinions from inside the numerical parameters . For outliers investigation, i use Regional Outlier Grounds (LOF) towards the software investigation. LOF detects and surpress outliers analysis.
For every most recent mortgage from the software investigation may have several past fund. For each previous application has you to definitely row that is recognized by the newest feature SK_ID_PREV.
I’ve each other drift and you can categorical variables. We implement score_dummies for categorical variables and you may aggregate to (mean, min, maximum, amount, and you can sum) to have float parameters.
The information off payment history to have earlier in the day loans at your home Credit. Discover that row each made payment and something row each missed fee.
With regards to the missing worth analyses, missing viewpoints are so short. So we don’t need to capture any step having shed thinking. We have each other float and you will categorical details. We implement rating_dummies getting categorical details and you can aggregate to (indicate, min, maximum, amount, and you may share) getting drift variables.
This info contains monthly balance pictures regarding previous playing cards you to new applicant obtained from your home Borrowing
It includes month-to-month investigation concerning early in the day loans within the Agency studies.