Business Impacts
~86.9%
model accuracy achieved
Customer Key Facts
- Country : Netherlands
- Industry : Automobile leasing and fleet management
Problem Context
The client, a prominent figure in the automobile leasing and fleet management industry, operates an internal system (MIMS) for gathering vehicle repair and maintenance information associated with each leased vehicle. However, the client faces a labor-intensive process where repair requests for vehicles are manually approved or declined based on historical data and cost, resulting in a time-consuming and resource-intensive procedure.
Challenges
- Lack of consistency in data capturing and data schema
- Multiple sources of data existed with complex dictionaries
- Large-scale data ingestion
- Integration and joining of 28 different tables to set up the master file
- Inclusion of exception conditions not captured in historical data in the form of policy rules
Technologies Used
Amazon Sagemaker
Amazon Elastic MapReduce (EMR)
Amazon Aurora
Amazon Cloudwatch
Amazon Simple Notification Service(SNS)
Amazon S3
AWS Lambda
AWS Cloudtrail
Apache Spark
Solution
Quantiphi developed a highly scalable and cost-effective system that utilizes a Machine Learning classification model, based on historical data, to automatically determine whether vehicle repair requests should be approved or declined. This approach significantly reduces the need for human intervention and substantially reduces the processing time for repair requests.
Results
- Vehicle repair requests automatically get classified into approved, rejected, or to be reviewed based on historical vehicle data
- Manual intervention reduced thus saving time and effort of Corporate Social Responsibilities (CSRs) to process them
