英国爱丁堡大学留学生指导作业-University of Edinburgh-Credit Scoring and Da

University of Edinburgh
Credit Scoring and Data Mining
Workshop 2: Statistical techniques for building scorecards
1. Look at the attached print out of a statistical package which is trying to build a scorecard. Which characteristics would you have in the scorecard and what values would you give them. What does the other data tell you about the validity of the scorecard.
2.Consider the following data on age. Find the best split in a classification tree using the following criteria
i) Kolmogorov-Smirnov，welcome to web：ii) basic impurity index
iii) Gini impurity index
iv) chi square
Example of splitting on age
age group <21 22-24 25-27 28-31 32-34 35-39
no good 4 7 7 9 6 4
no bad 2 3 5 3 3 7
age group 40-44 45-51 52-59 60+
no good 4 5 5 9
no bad 6 5 5 1
3. Prove that if the distribution of goods and bads is normal then the resultant population satisfies the logistic regression assumption that
Log ( p(x|g)/p(x|B) ) = w.x. + w0