Reference no: EM132993597

This assignment begins our study of predictive modeling techniques, starting with two variants of decision trees, CARTs and Random Forests. You will first work through the mechanical aspects of programming these methods, then in Part 2 you will examine some of the theoretical and mathematical bases behind them.

Part 1: Operational Tasks

For this assignment, work with the loans_training and loans_test data sets. Use Python to solve each problem.
To demonstrate completion of this assignment, create a Word document with your working code, screenshots of program results, and written answers to questions. Upload your final Jupyter notebook and Word document to the LMS when complete.

1. Create a CART model using the training data set that predicts Approval using Debt to Income Ratio, FICO Score, and Request Amount. Visualize the decision tree. Describe the first few splits in the decision tree.

2. Develop a CART model using the test data set that uses the same target and predictor variables. Visualize the decision tree. Investigate the splits in the decision tree. Does the tree built using the test data match the tree built using the training data?

3. Build a C5.0 model using the training data set that predicts Approval using Debt to Income Ratio, FICO Score, and Request Amount. Specify a minimum of 1,000 cases per terminal node. Visualize the decision tree. Describe the first few splits in the decision tree.

4. How does your C5.0 model compare to your CART model for the loans_training data? Describe the similarities and differences.

5. Create a C5.0 model using the test data set that utilizes the same target variable, predictor variables, and minimum cases criterion. Visualize the decision tree. Does the tree built using the test data match the tree built using the training data?

6. Use random forests on the training data set to obtain the predicted value of Approval using the same predictor variables as in the CART and C5.0 models.

7. Use random forests on the test data set to obtain the predicted value of Approval in the test data set. Build a table comparing the predictions from the training and test data sets. How do they compare?

Part 2: Mathematical and Statistical Basis

1. Read Delen et al. (2013). Explain how the authors used principal component analysis (PCA) to decompose their data into linear components, similar to discriminant function analysis in MANOVA. How were the factors, eigenvectors, and the covariance matrix determined? What role did the squared error distance play in their conclusions?

2. Continuing with Delen et al. (2013), what role did the Chi-squared automatic interaction detector (CHAID) play in their decision tree algorithm? How does the CHAID technique compare to C5.0, CART (referred to as C&RT by the authors), and the quick, unbiased, efficient statistical tree (QUEST). How do the accuracies of these techniques compare, and how were they measured by the authors?

3. Read Meena et al. (2019). Critique the authors' implantation of the CART algorithm, including their association rules and the set-theoretic justifications they present in Sections 4.4, 4.5, 5, and 6. Do these mathematical bases properly support their computational results presented in section 7? Why or why not?

Include references to all theoretical concepts and works cited. Show all your steps with explanations. Explain major components of complex solutions, code, and any output. Include captions to tables, images, and diagrams. Use formal and detailed mathematical and scientific notation throughout the document

While APA style is not required for the body of this assignment, solid academic writing is expected, and documentation of sources should be presented using APA formatting guidelines, which can be found in the APA Style Guide, located in the Student Success Center.

Attachment:- Operational Task.rar