Task $1$:

Import the raw data $($CC$\_$Default.csv$)$ into your Jupyter notebook.

$1\mathrm{.}1$ Check if the data is loaded correctly by printing a few observations. Check the total number of observations and variables.

$1\mathrm{.}2$ Provide the descriptive statistics and manipulate data.

a$.$ Check for missing values if any.

b$.$ Plot the univariate distribution.

c$.$ Convert the relevant variables such as payment variables $($Pay$0-$Pay$6$ and customer related variables$)$ to categorical variables as appropriate.

$1\mathrm{.}3$ Find the variables that are correlated and the variables that might help in finding the defaulters next month using a few plots. The plots should provide insights on the following:

a$.$ The independent variable that should help identify those who will default from the next month$$s credit card payment

b$.$ The relation between dependent and independent variables

c$.$ The correlations among the variables, etc.

$1\mathrm{.}4$ Provide your insights into the variables and their relationship based on your analysis in Task $1\mathrm{.}3$ in a markdown cell in your Jupyter notebook.

Task $2$:

Import $$Train$.$csv$$ into your Jupyter notebook.

$2\mathrm{.}1$ Check the total number of observations and print a few records. Please note that the variable conversion in the raw data, similar to Task $1\mathrm{.}2$ should be applied

Hint: Convert the relevant variables such as payment variables, Pay$0-$Pay$6,$ and customer related variables $($demographic$)$ to categorical variables as appropriate.

$2\mathrm{.}2$ Fit a logistic regression after making the dataset balanced.

Hint: Use class weight parameter.

$2\mathrm{.}3$ Remove the variable$($s$)$ that would cause multicollinearity. Explicitly state the variable$($s$)$ that you are dropping in a markdown cell in your Jupyter notebook.

Hint: To remove a variable, use the drop function.

Import $$Test$.$csv$$ into your Jupyter notebook.

$2\mathrm{.}4$ Test the model on the test dataset. Please note that the variable conversion in the raw data, similar to Task $1\mathrm{.}2$ should be applied.

Hint: Convert the relevant variables such as payment variables, Pay$0-$Pay$6,$ and customer related variables $($demographic$)$ to categorical variables as appropriate.

$2\mathrm{.}5$ Plot the confusion matrix.

$2\mathrm{.}6$ Provide your insights on accuracy, precision and F$1$ Score in a markdown cell in your Jupyter notebook.

Task $3$:

$3\mathrm{.}1$ Fit a random forest model on $$Train$.$csv$$ with a random state of $1,500$ epochs, a maximum depth of $3$ and a maximum feature of $3.$

$3\mathrm{.}2$ Evaluate the confusion matrix, F$1$ scores and accuracy. Compare the random forest model with the logistic regression from Task $2.$ State your observations in a markdown cell in your Jupyter notebook.

Task $4$:

$4\mathrm{.}1$ Fit support vector machine $($SVM$)$ algorithms on $$Train$.$csv$$ with the following parameters: gamma $=0\mathrm{.}025$ ; C$=3.$

$4\mathrm{.}2$ Provide the confusion matrix, F$1$ scores and accuracy in a markdown cell in your Jupyter notebook.

Task $5$:

$5\mathrm{.}1$ Fit an ANN model $($sequential$)$ with $16$ input neurons and add two hidden layers with $8$ neurons each.

$5\mathrm{.}2$ Use $$relu$$ activation and adam optimiser. Use the normal kernel initialiser. Run it for $100$ epochs on $$train$.$csv$$ with a batch size of $15.$

$5\mathrm{.}3$ Provide the confusion matrix, F$1$ scores and accuracy on the test dataset in a markdown cell in your Jupyter notebook.

Task $6$:

Explain which model you will use based on the evaluation metrics on the test dataset among all the models from Task $2$ to Task $5($logistic regression, random forest, SVM and ANN$)$ and explain why. Put your answer in a markdown cell in your Jupyter notebook.

About the data

The $$CC$\_$Default.csv$$ "Train.csv$"$ "Test.csv$"$ dataset contains a total of $25$ variables which are the following:

ID A numerical value assigned to each credit card customer

LIMIT$\_$BAL: The remaining credit a customer can use

LIMIT$\_$BAL $=($credit limit $-$ used up amount$)$

SEX $1=$ male ; $2=$ female

EDUCATION A customer$$s educational attainment:

$1=$ graduate school

$2=$ university

$3=$ high school

$4=$ others

$5=$ unknown

$6=$ unknown

MARRIAGE A customer$$s marital status:

$0=$ unknown

$1=$ married

$2=$ single

$3=$ others

AGE A customer$$s age in years

PAY$\_0$ Repayment status in September $2005$:

$0$ or less: Paid duly

$1$ or greater $=$ the number indicates the number of months the payment was delayed

PAY$\_2$

Repayment status in August $2005$

$0$ or less: Paid duly

$1$ or greater $=$

PAY$\_3$ Repayment status in July $2005$

$0$ or less: Paid duly

$1$ or greater $=$

PAY$\_4$ Repayment status in June $2005$

$0$ or less: Paid duly

$1$ or greater $=$

PAY$\_5$ Repayment status in May $2005$

$0$ or less: Paid duly

$1$ or greater $=$

PAY$\_6$ Repayment status in April $2005$

$0$ or less: Paid duly

$1$ or greater $=$

BILL$\_$AMT$1$

BILL$\_$AMT$2$

BILL$\_$AMT$3$

BILL$\_$AMT$4$

BILL$\_$AMT$5$

BILL$\_$AMT$6$

PAY$\_$AMT$1$

PAY$\_$AMT$2$

PAY$\_$AMT$3$

PAY$\_$AMT$4$

PAY$\_$AMT$5$

PAY$\_$AMT$6$

default.payment.next.month

Shows customers who defaulted on their payments on the following month: $1=$ yes $0=$ no