Case Study:$$German Credit Risk Analysis:$$

$$

Context:

To minimize loss from the bank$$s perspective, the bank needs a decision rule regarding whom to approve the loan and whom not to$.$ An applicant$$s demographic and socio$-$economic profiles are considered by loan managers before a decision is taken regarding his$/$her loan application.

In this dataset, each entry represents a person who takes credit from a bank. Each person is classified as a good or bad credit risk according to the set of attributes.

$$Objective:

The objective is to build$$predictive models on this data to help the bank take a decision on whether to approve a loan to a prospective applicant.

$$

Considerations:

$$

If a potential customer is misclassified as being at Risk, the bank will not give the loan to that person. This will be a loss of opportunity to earn interest on the potential loan.

If a potential customer is misclassified as NOT being at Risk, they may be given the loan but may default later. This will be a loss of resources.

$$

Based on the above logic, you need to decide whether to look for maximizing Recall, or Precision, or f$1-$score. Please give your reasons for choosing a particular metric and use the metric chosen by you to evaluate the performance of the models.

$$Attribute Information:

The data contains characteristics of the people

Age $($Numeric: Age in years$)$ Sex $($Categories: male, female$)$ Job $($Categories : unskilled and non$-$resident,$$ unskilled and resident, skilled, highly skilled$)$ Housing $($Categories: own, rent, or free$)$ Saving accounts $($Categories: little, moderate, quite rich, rich$)$ Checking account $($Categories: little, moderate, rich$)$ Credit amount $($Numeric: Amount of credit in DM $-$ Deutsche Mark$)$ Duration $($Numeric: Duration for which the credit is given in months$)$ Purpose $($Categories: car, furniture$/$equipment$,$ radio$/$TV$,$ domestic appliances, repairs, education, business, vacation$/$others$)$ Risk $(0-$ Person is not at risk, $1-$ Person is at risk$($defaulter$))$

The data set $$German$\_$Credit.csv$$ can be downloaded from Data sets folder in CANVAS

$$

$$

Tasks and rubric:

$$

$1.$ Explore: $3$ points

Examine the data set and carry out EDA $($particularly showing how the other variables may be related to the target variable Risk $($through barplot$/$lineplot$/$boxplot etc.$),$ to derive initial insights

$2.$ Data preparation: $2$ points

Check for missing values, convert string $($object$)$ variables to category. Separate the predictor and target variable. Create dummy variables as needed $($the final data set should have all variables as numeric$)$

Note: If a variable is binary, dummy variables are not needed. However, if the two values of a binary variable e$.$g$.$ gender is coded as $$Male$$ and $$Female$,$ these should either be converted to $0$ and $1,$ or dummy variables made. For categorical variables with more than two categories, dummy variables MUST be made.

$3.$ Model building: $5$ points

Split the data in train and test sets, using a $75$:$25$ split. Build a Decision Tree model, and a Random Forest model. Compare the performance on metrics: F$1$score, Precision and Recall.

$4.$ Tuning and evaluation: $5$ points

Improve the performance of these models by tuning the hyperparameters $($use GridSearchCV$).$ You can also try to use different class weights.

Compare the performance of all four models on training and test data set. Based on your criteria, choose the best model

$5.$ Insights: $5$ points

Determine the feature importance in your chosen model

List out the business insights, based on your EDA and chosen model

$$

Caution: Tuning the Random Forest is computationally intensive. Therefore, do not specify a very large Hyperparameter space in GridSeachCV

Guidelines for submitting:

$$ Annotate your Jupyter Notebook, to explain your procedures, comments and conclusions

After completion, run the Jupyter notebook from start to finish