这是一个CS的Problem Set作业代写案例
Feel free to ask about algorithm-related problems. Please learn to debug your code using Google, GitHub, Stackoverflow, and the package manual.
This problem set aims to provide some experience applying econometric methods in risk analysis. The data set “credit_risk.csv” is available to you online. Your main task is establishing models that predict the default label using all other covariates.
A table of variable explanations is here:
1. Simple Logistic Model
Run logistic regression: regress default label on income and past bad credit. Summarize your result. Obtain prediction values in the regression above. Compute and plot the ROC curve. Compute AUC value. Explain your main results.
2. Full Logistic Model
Now you are free to use any variables and transformations you wish. Try to obtain a model with as high AUC as possible. Report the result from logistic regression, AUC value, and the ROC plot as your main result (similar to what you did in the previous simple task).
Hint: You might want to try to use interaction terms or non-linear transformations (polynomials, log transformations, dummy variables, etc.).
3. SVM/Random Forest
You might wonder whether non-linearity can help. Try SVM or Random Forests method. You can select either one. Then, report the key parameters of your model, the AUC value, and the ROC plot as your main result.
4. Out-of-Sample Test
You might worry that your results may be over-fitting. Consequently, you decide to split the sample into two parts, use one part to train a model, and test the model in the test sample.
Take the first 10,000 samples as your training sample, and use the rest as the test. Run a logistic regression/SVM/Random Forests in the training sample based on your model in Section 2 and report your model. Then, apply the model to the test sample to create an out- of-sample prediction score for each observation in the test sample. Report out-of-sample AUC and ROC.
You may also try to resemble different models to pursue a higher out-of-sample AUC.
Scores of this assignment is heavily related to your out-of-sample AUC score for this question and the description of your optimization logic.