Defaulters prediction on next cycle I have data of loan installment repayments by customers, it contains all regular details like loan amount, last installment paid amount, next installments, credit score, age, region etc. 
Along with all above details i have label on each customer as delinquent/non delinquent. But the delinq or non-delinq labeling has been done based on if customer paid for this month or not?
Now i am trying to predict that who will be the customers from the remaining data(who are non-deliquent this month) may turn delinquent next month.
My question is :
1) What is the best method to solve this problem?
2) To achieve best results, i can get the data image for past 12 Months(If time component to be used.)   
 A: In my opinion, your first port of call should be standard binary choice models. The 'silver bullet' in credit risk modelling is logistic regression. However, you can branch out and try generalised linear models with probit and complementary log-log link functions too. The benefits of these 'standard statistical methods' are:
1) Your model is easy to interpret. Model explainability is important in credit risk modelling. 
2) You don't need vast amounts of data, as is the case with machine learning models. 
3) Over sampling isn't required for standard statistical methods. In fact, over sampling will bias your model. 
My recommendation is to avoid machine learning methods.
Another challenge is variable selection. Unfortunately, step-wise selection procedures are common place in credit risk modelling. These should be avoided all together - see Frank Harrel's objections here.  A nice alternative is Bayesian variable selection, or even penalised regressions. Bayesian methods are nice since they let you elicit and incorporate expert knowledge about which variables are, and are not useful. 
