The Alivia Technology Blog

Three Common Misconceptions about AI and Big Data

Posted by Kleber Gallardo on Wed, May 3, 2017 @ 21:05 PM

As AI and Big Data continue to become commonplace in a range of industries, there are still some lingering concerns  about how effective its implementation may be in various organizations. While these doubts take a variety of forms, they essentially boil down to three primary concerns. Let’s take a look at the three concerns and the ways that the predictive analytic tool, Absolute Insight, address and overcome them.

Misconception #1 – The data is not reliable enough

Is the data reliable

There is no substitute for clean data. It’s required for effective decision making, strategy forming, and business execution. Unfortunately in legacy systems much of the data is incomplete and inaccurate – or in our terms – dirty. Currently, the reliance of agencies and organizations on legacy systems and the data they contain is still very high which hampers the adoption of big data application.

In analytic projects, accurately estimating the effort required to clean the legacy data is like trying to judge the mass of an iceberg hidden beneath the water: difficult if not impossible. Fortunately, one of the strong points of Absolute Insight is its ETL (Extract, Transform, Load) capabilities. Put simply, the tools needed to clean data simply, efficiently, and effectively are built into the application. Most importantly – in addition to streamlining the data cleaning process – once completed, the process can be re-used as data is refreshed and renewed for additional analyses.

Of course cleaning data isn’t enough. In addition to cleaning your data, Absolute Insight makes sense of your data –both for instant business decisions and for future planning. Volume of data isn’t an issue here either: the larger the data sources the better the models become.

Misconception #2: It's too complex to be useful by operations staff such as Fraud Unit Investigators and Auditors

Is Big Data and AI too confusing?

Big Data has traditionally been viewed as territory fit only for data scientists. The mathematical and technological skill set required to effectively use it has put it out of reach of many individuals who would benefit from it. Making it accessible to the average person is at the core of what makes Alivia Technology’s approach to predictive analytics so unique.

An easy way to think of Absolute Insight is that it’s your own personal data scientist. The goal at Alivia Technology is to free your staff from the need to master the programming and mathematical needs associated with predictive analytics. We supply the data analysis tools and training; your staff provides the subject matter expertise to best leverage the data.

Who are subject matter experts? These are people that currently exist within your organization who understand your business, your regulations, your challenges, and your goals. By arming them with the power of data, they will be in better position to take action immediately and efficiently compared to a data scientist who may lack the operational experience and savvy.

Misconception #3: AI isn’t financially feasible

Is Big Data too expensive?

AI can be expensive and user adoption low when a full blown solution gets “implemented”. It doesn’t have to be that way.

It is the partnership between the AI provider and your subject matter experts that allows for big data analytics to be delivered in an extremely cost effective manner.

Absolute Insight’s architecture utilizes lower cost (but highly powerful and fast) Apache Spark technology and state of the art visualizations to bring predictive analytics to the desktop of subject matter experts. By constructing libraries of templates and wizards, it allows users to employ the complex algorithms by merely selecting a course of action. By sidestepping the need to have every piece of data analyzed and disseminated by data scientists, this new approach saves money, while increasing speed and improving overall efficiency.

 

So there you have it. There are certainly barriers to predictive analytics, but fortunately with the right product, they are easily overcome.

Come learn more about how Absolute Insight is changing the way organizations approach analytics by signing up for a free demo!

Sign up for a free demo of Absolute Insight

Tags: Big Data, Predictive Analytics, Artificial Intelligence

5 Things Your Fraud Unit Needs To Do

Posted by Paul McLaughlin on Mon, Apr 17, 2017 @ 15:04 PM

In the recent blog The War on Healthcare Fraud, our colleague made the case that the Health plans and Government Health agencies will never win the war “one tip at a time.” Something dramatically different is needed.

The current investigation process is like a person looking for his car keys under a spotlight. When the policeman asks where do you think you lost them, the man points to over in the shadows. To which the policeman fires back: “Then why are you looking here?” The man looks up and says “That’s where the light is.”
5 things your fraud unit should be doing

While a spotlight is useful at illuminating what it’s pointing at, it needs to be moved manually to shine on other areas.

Here’s where Artificial Intelligence and Machine Learning come in: it shines the light on the shadows. New powerful tools are available to the fraud investigator that they never had access to before. These tools, such as Absolute Insight, offer a simplified approach to data sets, coupled with powerful analytics and data visualization for insights. However, to maximize the advantage of these powerful tools, the Fraud Unit needs to upgrade its business process, skills and best practices.


Here five things they should be doing now:

  1. Look at the entire data set to analyze the information.
    If you are only looking at samples, you’ll only find what’s under the spotlight. Big data encompasses the totality of information, rather than the bits and pieces featured in samples.

  2. Cross reference the data set with other databases.
    Big Data is a somewhat misleading term. Rather than thinking about it in terms of size, its strength is about the capacity to search, aggregate, and cross-reference large data sets. 

  3. Start Real-time monitoring of changes over time.
    Data isn’t static. Monitoring changes over time will allow you notice patterns, make predictions, and take action at a time when it will make the strongest impact.

  4. Focus on targeted audience (high risk) and audit them.
    While fraudulent behavior, waste and abuse takes on many forms, “soft” targets are the best places to start looking. Consolidate your efforts here to capture the high value audiences before shifting your efforts elsewhere. 

  5. Get alerts once changes occur.
    The atmosphere in which data lives in constantly changing. Investigators need to keep up with the changes, whether they’re regulatory, departmental, or practice based.


Fraud Units need to start incorporating these strategies. Along with the powerful capabilities offered by AI and Machine Learning, those responsible will also need to streamline the investigation process and harvest and monitor the benefits identified by predictive analytics.  

Come see how you can join the new wave of detection with a free webinar on Alivia Technology's cutting edge platform, Absolute Insight!

Sign up for a free webinar

Tags: fraud detection, special investigation unit, fraud unit

How to Detect Questionable Transactions Through Big Data

Posted by Paul McLaughlin on Mon, Apr 3, 2017 @ 16:04 PM

 

How to detect questionable transactions

By Paul McLaughlin

 

Not all transactions are created equal.

In normal circumstances, transactions are simply an instance of buying or selling something; a business deal that is agreed upon by all necessary parties. An easy way to think of legitimate transactions is to view them as being solid or having mass.

However, risky or questionable transactions also occur. Risky transactions are absent of some critical components which makes them less solid and thus have less mass. A risk bearing transaction may lack an approval, not comply with the business rules, or contain conflicting information – all of these conditions indicate a level of risk. This absence of one or more procedural safeguards makes the transaction different than complete and duly authorized transactions; and by our definition, less substantial.

Fraudulent transactions have the least substance. Imagine a legitimate transaction as a golf ball and a fraudulent one as a ping pong ball. The golf ball is going to withstand virtually any level of force, while the ping pong ball will collapse under a modest amount of pressure.

Absolute Insight, through leveraging big data, is designed to isolate and cluster those transactions lacking the proper approvals and safeguards, thus identifying the greatest amount of potential risk. The Absolute Insight engine evaluates all the transactions, scores them and clusters those that appear to lack the proper mass. Once clustered, the engine will bombard them with new and more focused algorithms. The goal is to increase the pressure on these transactions by identifying behavior patterns that defy the normal business environment and increase the level of observation on those transactions providers, or groups of providers.

This will put the results into the hands of skilled auditors and investigators for the final determination of systemic weakness or fraud. Eventually, the fraudulent transactions will – much like a ping pong ball struck with a golf club – implode.

Sign up for a free webinar on how to detect risky/fraudulent transactions! 

 

Tags: Big Data, Alivia, Questionable transactions, Risky transactions

Cognitive Intelligence: The Next Wave in Computing

Posted by Paul McLaughlin on Thu, Mar 30, 2017 @ 22:03 PM

By Kleber Gallardo and Paul McLaughlin

 

blog 1.png

 

Today more than ever the world runs on data. Not just governments and large corporations but small business, service industries, and educational institutions and individuals. Within this universe is a subset of users, decision makers, who need to analyze huge volumes of data, hundreds of millions even billions of transactions , with the requirement of identifying risk, risk laden behaviors, or conditions. There is so much data we need tools to find and identify these conditions and predictive models to help determine when and if to intervene. With the correct tools and strategies data you can make the data work for the benefit of individuals, private organizations, and government agencies.

The human brain cannot absorb all of the data that is available today and process it in a manner that will maximize the results of the decision making process. Today, each of us has a personal health care record that is enormous, over our lifetime it can grow to one-hundred gigabytes. How do physicians and other caregivers access this information? How do they analyze it? How do they know when to act on it? They will need analysis tools to help them uncover the patterns within the data.  

The Internet of Things[i] (IoT) is the network of physical objects or "things" embedded with electronicssoftwaresensors, and network connectivity, which enables these objects to collect and exchange data. In the case of healthcare this means that machine to machine data, data from appliances, medical images, and electronic notes from doctors can be collected, assimilated, analyzed, and present to the physicians and practitioners the summarized information they need to make decisions in real time. The objective of the new breed of analytic tools that use statistics, predictive analytics, and cognitive intelligence is to analyze, make inferences, and garner insights from data that is too vast for an individual to absorb.

The benefit from these types of tools extends well beyond just healthcare. Most importantly, these tools need to provide insights, identify anomalies, and predict future behaviors. In healthcare this new breed of analytics tools serve to analyze the vast amount that is available and directly provide business insight and actionable intelligence.  But this same technology can be employed to analyze sales results, identify the most successful salespeople and attempt to emulate them; identify fraudulent behaviors and disrupt them; or identify high risk behaviors and patterns and intercept them where appropriate. We are in a unique point of time, we are creating new, more powerful, more intuitive systems; systems capable of cognitive analytics.

Cognitive Intelligence

For decades we have been developing rules based approaches to solving problems. Deterministic models have been the tool that has been used to develop solutions. Deterministic models rely on the programmer’s knowledge of both the expected outcomes and the process. Processes, by and large, that are trails or paths through data using rules, if then else statements, and algebraic formulas to articulate the outcome (Petrocelli). The new strategy employs probabilistic models of creating solutions. A probabilistic model analyzes information, identifies trends, clusters, and behaviors that previously would have remained buried in the data. It does this by using statistical analysis that projects the probability of future behavior based on past history. The new approach is called cognitive computing. 

It utilizes an open approach, machine learning and sophisticated natural language processing.  Characteristic of cognitive applications, is the capability to understand language, apply logic, interpret, intuit and relate information, predict, evaluate, and make decisions. This provides a whole new group of computational solutions.

The Four Big Strengths of Cognitive Computing

  • Identifying and Isolating Unknown Patterns of Behavior

One strength of is that it can identify patterns of behavior, transactions, or trends within data that invisible to detection given the rarity of the transaction and the volume of data. Since most software is deterministic it assumes that there is an existing, recognized, and human designed blueprint, or schema. A schema is merely a model. Where deterministic solutions require that the model be identified and understood in advance, this new approach adapts itself to situations where the schema is not known in advance. “When faced with an unknown information, humans build new schema naturally while most software needs to have it spelled out ahead of time.” (Petrocelli).Cognitive computing has closed the gap between the two.”

  • Assimilate and Employ Vast Amounts of Data

Bounded rationality “assumes that people, while they may seek the best solution, normally settle for much less, because the decisions they confront typically demand greater information, time, processing capabilities than they possess. They settle for “bounded rationality or limited rationality in decisions.” (Chand).  It is not just the possession of data that creates bounded rationality, but the ability to consume, absorb, and make rational decisions based on the information available. Machine learning helps address the limits of bounded rationality.

Humans are limited to the amount of information they can synthesize and utilize during any decision event. Computers have no such limitation, and can actually absorb new information from additional sources, while fully engaged in a decision event. This new technology enhances the learning abilities of a human being with the processing power of big data technologies.

  • Data Driven and Probabilistic Recommendations

Decision making is greatly enhanced by machine learning’s ability to access and synthesize additional information, and to have an almost infinite appetite for new data. This approach also finds patterns and trends buried within the data and uses that information to further a solution to the problem at hand. Finding the patterns in the data is only one aspect of what makes this type of computing cognitive. Projecting future behaviors or results based on prior histories and incorporating that information into its recommended solution is what truly separates probabilistic solutions from deterministic ones.

  • Scalability

The unpredictable nature of data and specifically the amount of data required for a particular project or problem lends itself to a cloud solution. In the cloud users will have access to computing resources as well as addressing storage needs. There is a second type of scalability that is very important that of the human infrastructure required to maintain an ongoing cognitive computing initiative. The mathematics required to fully exploit the power of machine learning are beyond the comprehension and skill levels of most programmers. Those that can handle both the mathematics and the programming are very expensive. A cloud solution allows both areas of scalability to be addressed. A properly applied machine learning strategy frees up analysts to do what they do best; analyze and discover.

Who Can Benefit from Cognitive Computing

Judith Hurwitz notes in a recent article “…that a cognitive approach to advanced analytics will have a dramatic impact on hundreds of different market segments. When we have the ability to gain insights that is hidden and then apply learning to that data there is a potential to transform industries ranging from healthcare, to financial services, metropolitan area planning, security, and IT itself. At the heart of business transformation is the ability to make sense.” Better decisions result from better, timelier information, machine learning provides freedom from bounded rationality and the constraints of assimilating additional data on the fly. Industry, business, government and education are all being drowned in data, a machine learning strategy will allow them to harness the data rather than be overwhelmed by it.

Virtually any enterprise that needs to make data driven decisions will benefit from this exciting new capability. Absolute Insight is a product that Alivia Technology is developing to take advantage of these new technologies and apply it to healthcare, government and industry. 

 

BIBLIOGRAPHY

 

  1. Interview, Chand, Smitri. "YourArticleLibrary." n.d. Models of Decision Making Models. Internet. November 29, 2015.
  2. Judith Hurwitz. "Judith's Balancing Act." November 11, 2015. Hurwitz & Associates. November 29, 2015.
  3. Petrocelli, Tom. Neuralytix. February 24, 2014. November 28, 2015.

Tags: Big Data, Analytics, Alivia