Assignment 4

Minicase

Background

You work for BostonFinance, and, for the purposes of this minicase, you will be playing two roles:

a. Credit Development Manager (CDM)

b. Member of the Data Science Team (DST)

As CDM, you are trying to grow the credit business of your unit and through your network you have been able attract some potential clients (6 potential clients) who are looking for a line of credit (LOC).

You have some data on these potential clients that you share with your DST and ask your DST to give you a couple of frameworks to help you make your final decision on whether you should approve a line of credit for the 6 potential clients.

The DST, in response, to the CDM’s request, has a team meeting where it’s decided that 2 modeling frameworks (Multiple Regression and K-NN) will be employed and two sets of decisions emanating from the two frameworks will be offered to the CDM. As a member of the DST, you volunteer to conduct the analyses and submit a brief report to the CDM.

Specifically, as a member of the DST, you do the following:

1. You obtain data that is similar to that shared by the CDM corresponding to current clients of BostonFinance (the data you obtain is the data-set, “CREDIT APPROVAL DATA SET ONLY”)

2. You run a Multiple Linear Regression model with Decision as the dependent variable with the goal of finding the one or two most important drivers of Decision (you follow exactly the process as you did with building and running the Final Cut Predictive Model of Assessed values)

3. After determining the one or two most important drivers of Decision, you then make a judgement of who should be approved for the LOC and who should not be approved

4. You also build and run a K-NN model with Decision as your output variable, partitioning 60% of the data into the training set and 40% into the validation set, normalizing the input data, using a cutoff value of 0.5, allowing K to vary from 1 through 10, and then scoring the new data with the best K that is chosen by the K NN Data Mining algorithm).

5. You prepare two tables (Table 1 and Table 2) showing the models of Decision from the Multiple Linear Regression and the K-NN modeling and submit them to the CDM

In your role as CDM you review the results from the two tables and ask the DST to run the K NN algorithm again but with K=10 and a Cutoff Value of 0.0001, holding everything else the same (using the same partitioning – 60% training data and 40% validation data and normalizing the input data. [See the video, “Choosing K and the Cutoff Value: The Role of Business Managers,” for one reason why, you, as a CDM makes this request. In Module 5, the reason behind this request will be discussed more fully].

Task

Your task, as a member of the Data Science Team, is to prepare a Report in Microsoft WORD which includes the following:

Table 1 (Scoring 6 potential clients based on Multiple Linear Regression results – enter your Decision based on these Multiple Linear Regression results in the Decision Column)

Table 2 (Scoring 6 potential clients based on K-NN modeling allowing K to take values from 1 through 10 and with a Cutoff Value 0.5. Enter your Decision based on your K-NN analysis in the Decision Column)

Table 3 (Scoring 6 potential clients based on K-NN modeling with K=10 and Cutoff Value of 0.0001. Enter your Decision based on your K-NN analysis in the Decision Column)