How It All Started

The Medical insurance company needed a solution that would help their auditors in making correct judgments on the claims issued by the medical aid providers. The main idea behind the creation of the system is to reduce the costs of claims auditing process performed by the auditors with the help of Machine Learning algorithms, therefore illuminating the risks or human mistakes. After being sufficiently trained, the algorithm is expected to optimize the day-to-day activities of the company’s insurance auditor. While calculating the required metrics, the system relies on AI technologies and the past data rather than making decisions based on variable computational rules.

Features

• The user is able to find a particular claim within a database
• The system allows users to compare claim valuations made by the auditor and the one of our ML algorithm
• The user may leave a comment on the results of particular claim valuation
• The system is able to: display claim’s fair paid amount, classify the claim as being fair/unfair, display error code and its classification as well as to decide whether the auditor’s action is required for a particular claim
• By using the system data with the accompanying documentation for an ML engine, an auditor will have an idea about the parameters influencing the system’s judgments of the particular claim. In case of any errors or a lack of parameters, an auditor will be able to leave a comment about the occurred errors for an ML Engineer. The latter option is required to improve the system’s prediction accuracy

Development Work

Our development team acts as a contractor to our customers and is responsible for creating the actual system.

The work is organized using the Agile development model and the Scrum framework. We have split the development into bi-weekly sprints with new features and product demos coming at the end of each stage. The client communicates with the team via Slack and Skype. We also use Git as a code repository.

The whole development process is subdivided into 4 main stages:

• The 1st stage which involves educating our Machine Learning algorithm. It involves training the ML algorithm to analyze the results of the auditor’s decisions based on the 6 Million claim items collected over the past 5 years
• At the 2nd stage of the development process, the system should be sufficiently trained to give auditors some useful information they’ll need to make correct judgments on the particular claim
• The 3rd stage should be characterized by our system being able to automatically give a correct decision on the claims received by the medical insurance company
• The 4th stage will be highlighted by the system being able to optimize the day-to-day activities of the company’s Medical insurance auditors

System Architecture

The system consists of the following UI elements:

• All claims table
• Claim details UI

Comments UI. The later is useful for Machine Learning engineer when there is a deviation between the auditor’s claim’ judgment and the one made by our system.

System Component Activities

• Big Data in Healthcare: load, combine and structure the available member’s data
• ETL: clean the database from the unnecessary data
• Process the data using the Supervised Learning, Neural Networks, Deep Learning and other algorithms

Workload

• Web App Developer
• Technical Lead
• Data Science engineer for ML engine prediction models
• Business Analyst
• QA specialist
• Project Manager