This is the first in Jay's series of “innovation chats” on how artificial intelligence is changing the face of government services. In this series, we will explore how chatbots, blockchain, and other artificial intelligence methods like machine learning will be used to prevent waste, fraud and abuse, order supplies, perform IT operations management and many, many functions performed manually today.
Machine Learning (ML) is a type of artificial intelligence that allows software applications to predict outcomes by using statistical analysis without explicitly designed coding algorithms. Using statistical analysis, ML applications ingest data and performs statistical analysis on the data to predict an outcome or a range of outcomes. Processes used in machine learning often use data mining and predictive analysis to look for patterns and then adjusting the applications actions accordingly. Waze uses ML to cipher through volumes of traffic data to continuously predict the fastest route to your destination. Facebook uses ML to personalize each user’s experience by analyzing user behavior’s (like when a user stops to read, clicks, shares or likes a friend’s post), the application will prioritize that friend’s post earlier in the News Feed.
Program Integrity (PI) departments are responsible for determine waste, fraud, and abuse (WFA), most often WFA is identified only after the claims have been paid, and significant effort is spent trying to recover paid claims (often called “pay and chase” approach.) By using ML, PI departments can implement a continuously improving “cost avoidance” approach by performing techniques like “risk scoring” on providers that will allow proactive mitigations on providers with high or rising scores, perform “link analysis” that identifies unusual or hidden relationships amongst providers, or between providers and beneficiaries, or “trend analysis” that can assist in Medicaid in developing policies that will lead to cost reductions, based on trends.
Another, growing use of AI in general, and ML specifically is around IT operations management (called AIOps). As performance monitoring, security monitoring, service desk incidents, and automation tools are deployed, ML can continuously review the various data points, and make corrective and preventative actions to increase availability and secure the infrastructure. This deployment of ML can also provide insights for manual and automatic capacity planning.
How does this affect you? ML requires data, lots of it! This big data can be sourced from many different sources including applications, mobile devices, and various forms of data sources. This means there is a real need for application integration, data harmonization, extraction, transformation, and loading (ETL) into big data sources that require proper data modeling. Lastly, the predicted outcomes will need to be reported, displayed, or acted upon in some way, shape or form, which may require process improvement. However, this information in the wrong hands can be detrimental to mission objectives, so security compliance is essential.
What are some of the leading ML technologies? There are a host of cloud ML services available like Amazon Machine Learning, Azure ML (part of the Cortana Intelligence Suite), Google Cloud Prediction API, and BigML which all have APIs (application programmer interfaces) that come with SDKs (software development kits) for Java, .Net, Python, PHP, and Ruby. Additionally, companies like SoftwareAG offer streaming analytics that provides real-time event processing on fast-moving data from mobile devices, Internet of Things (IoT) or internal transaction systems, via its Apama Streaming Analytics platform. And, yes, there are open source offerings too.
DSFederal’s Innovation Council is focused on emerging technologies, including ML, and will continue to research and seek the most compelling technologies; and to discover how to use those technologies to power transformative solutions.