Georgia Institute Of Technology - CS ISYE6501XISYE 6501 Course Project American Honda Motor Co., Inc

Document Content and Description Below

ISYE 6501 – Course Project American Honda Motor Co., Inc Case Background: Just like other automotive distributors, American Honda works with a network of dealerships that offers warrantied repai... r work. This can be a significant cost for the company; therefore, American Honda uses analytics to make sure the warranty claims are accurate upon submission. To decrease warranty expense, Honda wants to create a proprietary process to surface suspicious warranty claims. Before this innovative system, it would take the staff members a week out of each month to manually identify and scrutinize claims. The wastage of time and human resources would cost more to the company. With this proprietary process, American Honda would reduce labor costs and have faster claims analysis. The company also implemented SAS* forecast server to assist with business planning to forecast usage of parts and services for future needs. This new system will not only decrease warranty expense and forecast future needs, but it will also improve customer satisfaction by using analytics to quickly evaluate customer survey data and gain insight. Case Summary: In this report, I will outline the analytics techniques and models I learned from this course that could possibly be used in the solution of this business problem. The three goals of this case are the followings: (1) Examining warranty data to make maintenance efficient (2) Predict the volume of customers coming into the dealerships (3) Use service data to forecast future needs to improve customer satisfaction Details of Models Could be Used: Identifying suspicious warranty claim accurately Step1 Given: Historical customer data and historical warranty claim data, including predictive factors such as date of purchase, Customer addresses (multiple claims leading to the same address), duplicate claims(Y/N; A service center submits duplicate claim for the same VIN), Prereplacement of parts(Y/N; file a warranty claim to repair part that has not yet failed), false date(Y/N; submit a back-dated warranty claim to fall within the warranty period), false defect(Y/N; submit a claim for a covered defect when the actual defect is not covered by the manufacturer warranty), claim amount, and claim frequency, etc. Response variable would be either suspicious or non-suspicious claim (binary; Y/N). Use: Variable selection methods such as Elastic Net variable selection or Stepwise selection To: Identify which of these predictive factors have the greatest impact in classifying the warranty claims into suspicious or non-suspicious [Show More]

Last updated: 1 year ago

Preview 1 out of 3 pages