Case No 4 The HDFC Bank is using neural net technology in data mining software to develop more accuracy in marketing and pricing financial products, such as home loans etc. HDFC Bank offers a variety of tailored product packages by adjusting fees, interest rates, and features. The result is a staggering number of potential strategies for reaching profitable customers. Selecting through the vast number of combinations requires the ability to identify very fine opportunity segments. Data extracted from the date warehouse was analysed by neural net based data mining software to discover hidden patterns. For example, the software discovered that a certain set of customers were 15 times more likely to purchase a high margin lending product. The bank also wanted to determine the sequence of events leading to purchasing. They fed the parameters to the Discovery software from HYPER parallel and built a model for finding other customers. This model proved to be so accurate that it discovered people already in the process of applying and being approved for the lending products. Using this profile, a final list of quality prospects for solicitation was prepared. The resulting direct marketing response rates have dramatically exceeded past results. The bank also machine learning to identify the loan applicant may be defaulters. The system analyse all the applicant data and then make different categories of applicant as good or bad in terms of loan repayment probability. Machine learning applies associating rules to associate the characteristics of past loan applicant and who have defaulted. However machine learning is not without risk. As it became evident during infamous financial crisis of 2008 when some banking executives deliberately deceived risk management systems in order to adjust capital-on-hand requirements. This deception let firms load upon risky debt, while carrying less cash for covering losses. If manager deceive their systems by entering bad data then resulting models are going to be worthless. Exactly this is what happened in this case. Faulty estimates from bad data cause the banks facing very high exposure to risk. Eventually when debt defaults occurred it resulted in failure of several banks, leading to the worst financial crisis since the Great Depression. In the banking collapse of 2008, and we also saw computer- driven trading funds plummeted in the face of another unexpected event the burst of the housing bubble. Answer the following questions: Both questions carries 5 marks: Minimum No. of Words per Question - 150 4a. Evaluate how data mining application has helped the HDFC Bank. What type of machine learning is used in predicting the probable defaulters? Give reason. 4b. Examine the problems with machine learning Point out the critical skills required by data mining team to avoid the problems. Justify