Pt2520 Unit 6 Data Mining Project

##This data frames won't be actually put to use, but it is good to have an unaltered copy of the train and test sets for reference and exploratory analysis.

In percentage of receivables is based on an aging schedule. The aging schedule compares customer balance with how long they have been unpaid. The aging report is managed daily and the company manages the accounts by age and develops an expected loss. The percentages of loss can change during each period based on the length of time an account is outstanding, so there are adjusting entries that will have to be made to accurately record the losses.

Problem 2.6: In fitting a model to classify prospects as purchasers or non-purchasers, a certain company drew the training data from internal data that include demographic and purchase information. Future data to be classified will be lists purchased from other sources, with demographic (but not purchase) data included. It was found that “refund issued” was a useful

Where π1 and π2 are prior probabilities of good and bad credit populations, Q (2∣1) and Q (1∣2) compute the probability of making Type I errors, i.e., a customer with good credit is misclassified as a customer with bad credit, and Type II errors, i.e., a customer with bad credit is misclassified as a customer with good credit, and A (2∣1) as well as A (1∣2) are the corresponding misclassification costs of Type I and Type II errors.

Now, looking at the uncollected balances schedules as of the end of March and June, do these schedules properly measure customers’ payment

So it is also attractive for Jules Kroll to take this opportunity to enter the credit rating industry. In addition, the “issuer pays” model, which used by the big three rating agencies, lets companies shop around for the best ratings, putting pressure on the agencies to inflate their grades. As a result, it is very difficult to argue that they can adequately represent the users’ side. Therefore, Jules Kroll wants to use another model that can assign unbiased and reliable ratings.

Aside from determining credit card payment and interest rates one could also apply these skills to other real life situations such as loans for mortgages, cars, or even school. These concepts can be applied to any situation where money is owed and paid back in payments. This will aid in finding the payment plan that pays off the debt faster, resulting in money being saved.

Lending evaluations by Santander are based on credit background of the person (or company) who wants borrow. Through the development of a credit-approval system, Santander Consumer Finance increased an understanding into these clients on an online database which allowed for the use of real time analysis in determining interest rates for the business interactions.

The Leeds Building Society uses data mining on their 500,000 mortgage holders to determine which ones are most likely to go into arrears. They looked at all the information held on accounts that went into arrears in the past to form ‘high risk’ account profiles that can be applied to current and future accounts. They had some surprising

On the other hand, my interest in the incorporation of artificial intelligence and machine learning by financial institutions stems from my work as a Credit Risk Analyst for BECU. I recently joined the institution, and am working on a year-long project to integrate the Synergy Component System which will enable increased automation, improved attribution management, and reduced decision times for credit card and auto loans. Beyond testing plans and cases, I am most excited about designing decision matrices by analyzing member data such as credit quality, debt-to-income ratio, membership years, and trade lines to reach a credit decision. While I have had already experienced academic research, this will be my first taste of applied industry research.

How data mining can assist bankers in enhancing their businesses is illustrated in this example. Records include information such as age, sex, marital status, occupation, number of children, and etc. of the bank?s customers over the years are used in the mining process. First, an algorithm is used to identify characteristics that distinguish customers who took out a particular kind of loan from those who did not. Eventually, it develops ?rules? by which it can identify customers who are likely to be good candidates for such a loan. These rules are then used to identify such customers on the remainder of the database. Next, another algorithm is used to sort the database into cluster or groups of people with many similar attributes, with the hope that these might reveal interesting and unusual patterns. Finally, the patterns revealed by these clusters are then interpreted by the data miners, in collaboration with bank personnel.4

Britvic PLC was compelled in 2007 to increase with approximately 10 days the period in which credit customers pay their liabilities, so that the impact of economic recession over the organisation’s profit was reduced by maintaining the drop in the level of sales at an acceptable level. One year later, in 2008, an 8 days decrease in his ratio was recorded, followed by a 2 days increase in 2009. As reported by their financial statements, the debtors’ days ratio has been stable for the past two years, a sign of recovery after the economic recession, but not as low as in the economic boom, characteristic for the 2005-2006 period.

ID3 uses entropy to measure the values of each attribute. After that it derives rules through a repetitive decomposition process that minimizes the overall entropy. Messier and Hansen (1988) used ID3 to derive prediction rules from loan default and corporate bankruptcy cases. The loan default training sample contained 32 firms with 16 in each group (default or non-default). In the corporate bankruptcy case, the training sample contained 8 bankrupt and 15 non-bankrupt firms. For the holdout samples, the rules derived by ID3 correctly classified the bankrupt/non-bankrupt firms with perfect accuracy and 87.5% accuracy for the loan

As technology improves, the wide use of “hard information”, such as the borrower’s credit history, reduces informational asymmetries. Therefore, long-distance small business lending is easier (Frame, Srinivasan, \& Woosley, 2001; Petersen \& Rajan, 2002). However, even with the use of credit score data, collecting ``soft information" still helps local lenders control risks to avoid delinquency (DeYoung, Glennon, \& Nigro, 2008) and provides informational advances in offering more favorable rates (Agarwal \& Hauswald, 2010).

Support Vector Machine (SVM) is primarily a classier method that performs classification tasks by constructing hyperplanes in a multidimensional space that separates cases of different class labels. SVM supports both regression and classification tasks and can handle multiple continuous and categorical variables. For categorical variables a dummy variable is created with case values as either 0 or 1. Thus, a categorical dependent variable consisting of three levels, say (A, B, C), is represented by a set of three dummy variables: