Read this White Paper


Statistics show that each year, fraudulent activities cause an estimated amount of $30 billion loss for U.S. property and casualty insurers. To be able to prevent these losses, a new approach of fraud detection needs to be embraced.

In this paper, we show how TAZI’s Claims Fraud Detection Solution can help SIU members spend less time on clean claims and concentrate more on the suspects. Our solution is based on a patented Continuous and Explainable, No-code, Automated Machine Learning platform. We lay out how our solution detects new patterns of fraud. We also demonstrate how it is possible to discover new micro segments in those patterns by receiving explanations.

Provide your email to read this full White Paper.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.


Investigating suspicious claims has become harder due to new fraud methods emerging via new technologies. Fraudsters keep changing their techniques and insurance companies are having a hard time keeping up to recognize those new patterns. 

According to an article from Insurance Information Institute, The FBI estimates that the total cost of insurance fraud (excluding health insurance) is more than $40 billion per year and this costs the average U.S. family between $400 and $700 per year. Considering the estimated cost and the percentage of fraud being around 10% of total claims, fraud detection is an important and challenging operation. Each year, SIU members are only able to review 10% of suspicious claims, which results in them catching around 1% of fraudulent claims.

With TAZI, it becomes possible to help SIU members investigate the right segment of suspicious claims, prioritize accordingly, and result in doubling the catch rate.

Since fraudulent activity patterns keep changing over time, capabilities of SIU to recognize and investigate them can stay short. In order to keep track of continuously changing fraudulent activity patterns, an assistance with wider capability needs to be added to rules-driven processes. This is where continuous learning takes over. With standard batch training, models learn the data you provide each time you retrain the model. In order to stay up to date in terms of recognizing new fraud patterns, machine learning models simply need to learn sample by sample, continuously.

In this paper, we share the detailed benefits of TAZI’s Continuous and Explainable No-Code Automated ML solution for Claim Fraud Detection.

Claims Fraud Problem Statement:

Fraud losses are on the rise in the U.S., the U.K. and many other countries. This increases the operational costs for insurers accordingly. In order to maintain their profits up to a certain point, insurers need to add extra amounts to customer premiums. To turn the tide, insurers should develop their fraud detection mechanisms.

For Insurance Claims, there are two types of fraud: professional and average. Professional frauds are submitted by organizational crime units and are very hard to catch. Average fraud, on the other hand, is just performed by people who want to earn some extra money from their insurance company. Common frauds include inflation of claims, misrepresenting facts on an insurance application, and submitting claims for injuries or damage that never occurred. 

In general, the three most encountered challenges are detecting continuously changing fraud patterns, reducing false alarms, and learning root causes of fraudulent transactions. These challenges are extremely hard to overcome by using traditional machine learning because: they can not be updated frequently enough; you can not give feedback to them which is the only way to reduce the false alarms; these traditional machine learning models do not explain themselves so you stay behind on learning the root causes of frauds.

Figure 1: Automated Churn Reduction Approach

In general, the three most encountered challenges are detecting continuously changing fraud patterns, reducing false alarms, and learning root causes of fraudulent transactions. These challenges are extremely hard to overcome by using traditional machine learning because: they can not be updated frequently enough; you can not give feedback to them which is the only way to reduce the false alarms; these traditional machine learning models do not explain themselves so you stay behind on learning the root causes of frauds.

TAZI Claims Fraud Detection Solution: 

TAZI utilizes the Fraud Detection Workflow shown in Figure 2. Once our AutoML platform is connected to the data source of the company, the AI model keeps updating itself so that you can be sure about considering newly emerging types of fraud as you investigate. TAZI Profiler makes sure that provided data is useful for ML applications and if not, gives recommendations. After this enrichment process, customer’s data becomes the best version of itself in terms of usability.

Figure 2: TAZI’s Fraud Detection Workflow

With this approach, TAZI’s Claim Fraud Solution helps SIU members spot emerging fraud trends by viewing dashboards. Our solution’s dashboards are designed to help you receive the most beneficial information extracted from our ML models. These dashboards, being customizable, are able to help SIU members view all hidden insights from company data and plan the investigation process accordingly. Below, Figure 3 shows a part of the dashboard, visualizing the suspect indicators for fraudulent activity. It clearly shows the special conditions indicating suspectful behavior of claimants, such as having less than 3 days between the claimants policy start and claim occurrence. These types of conditions have been extracted from company data and visualized to help better understand.

Figure 3: Claim Suspect Indicators on TAZI’s Dashboard.

In our approach, the second phase is simplifying the identification of fraud microsegments. We provide our customers with a sunburst diagram representing the explanation of our models. This particular element is crucial while enabling domain experts to benefit from our solutions. Our claim fraud detection model explanations provide insights on fraud within specific microsegments. A microsegment can be a class of customers that have reported more than 6 claims to date or are using the same type of car. 

On the left of Figure 4, the interface shows the most important features sorted by their effect on the decision of fraud detection. The sunburst diagram divides the claims by separating them to microsegments, making it easier to view the similar claims together. This interface helps SIU members while targeting the investigation of claims and lets them give feedback both in terms of instance and model.

Figure 4: Explainable AI interface showing predicted fraud microsegments (in red) and highlighting a particular microsegment.

This figure shows a microsegment created by several different conditions. This particular microsegment represents the claims which indicate nonsuspect for conditions like having little number of claims to date and claim parties not being familiar. On the other hand, at the end this microsegment indicates suspect due to conditions like having a high amount of claims to date and the claim time being night. This type of analysis becomes possible with ease thanks to our explanation model.

The appropriate actions after detecting the fraudulent microsegments may depend on their definition. The action may be to cancel the policy, close the claims or to investigate further in accordance with the outputs provided.

TAZI can generate automated targeted intervention strategies by each microsegment. These interventions can be prioritized depending on the claim amounts.


In addition, especially when there are abrupt changes through newly emerging fraud techniques, economic conditions, and risk levels, there might also be emerging micro segments with small amounts of data where traditional machine learning models are not able to detect these new patterns. For these situations, the TAZI self-updating models can be fine-tuned by business users who are able to detect newly emerging patterns for improved performance.

With TAZI, the SIU members are now able to drill down into the microsegment to view specifics around the fraudulent activities along with the most likely fraud reason and the intervention strategy which suits the claim best. While doing all these, TAZI assures to protect the insurer’s customer centric culture by its false positive reducing system. TAZI’s explainable technology also helps our customers make sure they are relying on unbiased ML with their decisions. With an increased SIU referral rate and reduced false positives, TAZI solves the problem of fraud at your institution.


  • If you have enough and clean data to detect fraud?
  • How can you build your own evolving fraud detection models within 10-30 days?
  • How can you start saving up on the correct prediction of fraud within 1-2 months?
  • How can you become 7 times more likely to catch fraud at top 1% high scores?
  • How can you up-skill your business and data teams to adopt machine learning?
  • Any other questions?

Contact us at:


Artificial intelligence (AI) is a source of both huge excitement and apprehension, transforming enterprise operations today. It is more intelligent as it unlocks new sources of value creation and becomes a critical driver of competitive advantage by helping companies achieve new levels of performance at greater scale, growth, and speed than ever before, making it the biggest commercial opportunity in today’s fast changing economy.

TAZI is a leading global Automated Machine Learning product/solutions provider with offices in San Francisco. TAZI is a Gartner Cool Vendor in Core AI Technologies (May 2019) and is considered as "The Next Generation of Automated Machine Learning” by Data Science Central.


Founded in 2015, TAZI has a single mission which is to help businesses to directly benefit from Automated Machine Learning by using TAZI as a superpower, shaping the future of their organizations while realizing direct benefits like cost reduction, increasing efficiency, enhanced (dynamic) business insight, new business (uncovered), and business automation.


Through its understandable continuous machine learning from data and humans, TAZI is supporting companies in banking, insurance, retail, and telco industries in making smarter, more intelligent business decisions. 

TAZI solutions are based on a most compelling architecture that combines the experiences of 23 patents granted in AI and real-time systems, proven at different global implementations. 

Some unique differentiators of TAZI products are:

  • Business users can automatically configure custom ML models based on their KPI and the available data. TAZI's Profiler accelerates this process through data understanding and automated cleaning, feature transformation, engineering, and selection capabilities.
  • TAZI models learn continuously, and are suitable for today's dynamic, real-time data environments.
  • TAZI models are GDPR compliant (no black-box models). They provide an
  • explanation in the business domain's terminology for every result they produce.
  • TAZI supports multiple (heterogeneous) data sources, i.e.,.: external, batch, streaming, and others.
  • TAZI can learn both from human domain experts and from data, which speeds up accuracy improvement.
  • TAZI’s hyper parameter optimization feature reduces human time spent for model configuration. TAZI products contain algorithms that are developed and coded to be lean, efficient, and scalable.

Get Started Today
Tazi Hub User Interface