In this predictive analytics example, this application is a specific example of the remote device monitoring solution developed by Striim, which alerts technicians to possible machine outages and cash shortages. Key ATM components are monitored in realtime, including: CPU, printer, card reader, temperature, and cash balances. The App also acquires streaming data about current ATM transactions and historical data to compare past behavior. With visibility into data streams you can predict ATM component failures and service your machine as needed, rather than when scheduled.
To learn more about the Striim platform, visit our platform overview page.
Hello, my name is Dmitri. Now I’m an app developer. To our back in this video I would like to show how sample ATM map works within Striim platform collection platform gives you an ability to process complex events in real time, correlate multiple data sources using a SQL-like language and deploy them in scalable environment. ATM application uses a combination of predictive analytics and rules based on knowledge of data. To alert the DTM technician about possible machine outage. It does it by monitoring key ATM components, CPU, Printer card, reader and temperature. The application also reads data about ATM transactions and historical data about its behavior. In the past, a ATM obligation utilizes intern features, set of web action platform including reading stream data via many adapters such as CSV or XML, JSON or CDC, event processing, Engine Analytical Library and then simple deployment model and user interface.
Speaker 1: Specifically for this uh, application. My approach has been to collect data for three sources, issues, history, real then trans transactions performed on each ATM. That would comprise of ID, time and amount spent and component matrix id, time and value of a reported by the component. It would then train model in real time, on interval of last 20 minutes of data and then generate linear coefficients that show trend for each ATM component. It will then run a application on the current date and predict component status trends using generated coefficients and real time. At the same time it’ll find out if a number of transactions qualifies the TM activity traffic to be a hot, warm or cold. And then it’ll intersect the findings and send alerts to user interface. A UI consists of three pages summary that shows overview of all locations being monitored. So in here you can see that we’re monitoring five locations and 2180 tms.
Speaker 1: 20 of them are called, one is warm, there are no hot. And then you can see most critical components, status predictions for the next 20 minutes. And then components, overall status and then details and the table on the right. And then on the left you can see a map with locations. So there is location, uh, all the way in Alaska and then in central United States. And their statuses are changing in real time. If you over one of them, you can see it’s ATM ids. In this case 64, it’s in Tucson, Arizona and it has a critical or warm, in this case everything is normal. So a second ago it had a critical temperature. Now it’s critical again, but whatever reason, after that, if you click and the into location summary, you will see all the tms for this location. In this case, this is in Tucson again, Arizona. This is where it is. You can zoom in into this location and see where exactly it’s located. And then you can click down to view the uh, component details and then hourly activity details for this specific KTM. In this picture you can see a CPU and then predictions for CPU in next 20 minutes. And then, printer component. This a user interface doesn’t show the other two components, but it’s not necessary for this exercise. So, uh, on the, on the right you can see totals and then transaction counts.
Speaker 2: Sure.
Speaker 1: Yeah. So by using the same approach should be easy to modify the app to monitor another type of device such as slug machines. So in conclusion, I would like to say that went back and provides a great end to end application platform that is easy to install. It’s a either unzip or on tar and then just start. And then each node will just, each additional node will just join the cluster in that they allow, it will allow scalability out of the box. And then, uh, it’s easy to implement applications with the sip simple declarative SQL like language. And it’s also easy to reuse or modify the app to repurpose it for some other usage.