Customer Background

• Our client is a service provider of AI/ ML expertise and digitalization solutions to online insurance platforms.
• Mudah (https://www.mudah.my) is one of the largest online marketplace platforms in Southeast Asia. Apart from selling properties, electronics, travel/ tour services, home, and personal care products it also sells vehicles on the marketplace.
• Under the vehicles category, it focuses on used cars, motorcycles, other vehicles along with car accessories and parts.
• The end-user, a vehicle insurance company needed to automate insurance premium computation for used vehicles that are bought from marketplace platforms like Mudah.

Requirement

• Automate search and collection of data for used vehicles, from the marketplace platforms like Mudah
• Pre-process & classify the data by vehicle make, year, odometer, drive, fuel, manufacturer, cylinders, dimensions.
• Build artificial intelligence-based prediction engine for car price estimation
• Integrate the predicted price-output data set with the existing depreciation model used by the insurance company
• Store the output in the database for creating data visualizations

Solution

• Aress assigned a team comprising of a data scientist and developers to work on designing and deploying a scalable architecture to leverage an automated AI/ ML solution to address the requirement
• The team leveraged the latest data scraping methods to scrape data from the website online sources
• Appropriate data cleaning, tagging, and classification techniques were used to help categorization of information
• Various predictive models were built and optimized to predict the price of vehicles including linear regression, ridge regression, Lasso, random forest, KNN & XGBoost.
• The model yielding the highest accuracy (Random Forest) was deployed as a part of the prediction engine and the predicted price for each vehicle was sourced into the depreciation model and displayed in visualizations.

Business Benefits

• The solution was designed in such a manner that it helped extract data for thousands of vehicles at a pre-set frequency in an automated manner, thereby eliminating human intervention and resulting errors.
• Integrating the prediction engine outcomes with the depreciation model made it easy for the vehicle insurance company with a faster turnaround on premium quotes, improving customer satisfaction.
• Resultant improvement in operational efficiencies and decision-making, enhanced insurance company’s business prospects.

Process Overview