Georgia Institute Of Technology - ISYE 6501ISyE 6501 Introduction to Analytics Modeling Course Project

Document Content and Description Below

ISyE 6501 Introduction to Analytics Modeling Course Project Finding your best customers with machine learning https://www.sas.com/en_us/customers/seacoast-bank.html Why I picked this project? - Th... e reason I picked this project is because with one of my past projects I have done similar analytics similar with bunch of personal customer data such as financial records, housing, mortgages, family history and more. I was in charge of basic operations such as finding outliers, finding correlations and good substitutions of missing values. I wanted to deep dive and share some of my ideas on advanced analytics modeling with similar project. Seascost bank: - As banking services move online, it’s critical for banks to get a clear view of their most loyal customers to answer questions like: What’s the lifetime value of a customer? What’s driving that profitability? And what’s the best opportunity to increase that value? - Banks have a lot of customer data, given their role as financial intermediaries. The challenge is tapping into it to understand customer value and, ultimately, find new ways to better serve, retain and acquire customers. After reading the introduction, I listed some of issues the Seascost bank was having: 1. Lots of data 2. Estimating a customer’s potential and how much the bank should invest in reaching and serving that customer to receive maximum ROI? 1. Lots of data - Find extreme outliers and remove if they look like incorrect values. For example, if the mortgage is excessive more than what the house is worth - Fill in the missing value based on converting other values in the categorical values or using mean. For example, subsititute the income with mean of the other customers who have similar features based on other factors such as occupation and mortgage. - Then run Principal Component Analysis to is to reduce the dimensionality of a data set consisting of many variables correlated with each other, either heavily or lightly, while retaining the variation present in the dataset [Show More]

Last updated: 1 year ago

Preview 1 out of 2 pages