Anomaly Detection
Fraud Detection for Healthcare
Fraud detection in healthcare is an important yet difficult problem. We focus on a concrete problem of probabilistic outlier detection from a feature set designed for pharmacy claims. Although the re-ported results are specific to pharmacy claims, this approach can be applied widely. We are currently extending the solution to fraud screening of more general medical claims and fraud detection in other verticals.
The Challenge:
Despite the substantial incentives, the fraud detection problem is still far from being solved. Several challenges need to be addressed. Data is inherently big and complex. Data analytics solutions need to handle extreme size data sets, with large amounts of data, billions of lines of records, and millions of patients and providers. The diversity of medical data demands a coherent system to handle multiple modality data(clinical, diagnosis data, claims data, etc).
Fraudulent actions turn into a set of seemingly normal claims. It needs a lot of intelligence and effort to tackle the fraud detection problem.
The Fraud Detection starts with a Framework; not an end product:
Fraud detection framework comprises of five components:
1. Fraud Rule Generation
2. Feature Construction
3. Risk Score Computation
4. Outlier Identification
5. Reporting and Visualization
Suite of Claim Screening Capabilities:
Outlier Identification
Unsupervised learning method to screen entities (e.g., pharmacies) with behavior that is drastically different from other similar pharmacies.
Relational Analysis
Relational Analysis. Analysis of relationships between entities (for instance, between doctors and pharmacies, or pharmacies and patients) to identify possible concrete collusion.
Temporal Sequence Analysis
Analysis of medical sequences (such as diagnosis, treatment, and medicine) to detect unusual patterns. An unusual sequence might be billing for an unnecessary or non-existent service, or due to identity misuse.
Geo-spatial Analysis
Geo-spatial analysis. To identify improbable drug fill or procedures.
Services -> Insights
Our unique approach is not only what differentiates us, but also what makes us successful. We provide a broad range of machine learning and artificial intelligence services and solutions to help organizations drive change, achieve their vision and optimize performance and productivity.
In as little as a six week engagement with our team, your team will have concrete machine learning training and implementations.
Understanding the Problem:
Finding Value in Data
The key to finding value in any scenario:
Understanding what data is being collected, how it’s Transformed
What are the insights needed and what’s possible
Cost to Value
Are you using the right platform at the right cost?
Are you storing the data necessary for insights
Are you reporting solutions providing analytics you need
Expertise
Data Science require heavy disciplines and expertise in:
Mathematics | Programming
Background in data management
Experience in solving business problems