verification using stream processing, machine learning and text analytics
False alarms triggered by security sensors incur high costs for all parties involved. According to police reports, a large majority of alarms are false. Recent advances in machine learning can enable automatically classifying alarms. However, building a scalable alarm verification system is a challenge, since the system needs to: (1) process thousands of alarms in real-time, (2) classify false alarms with high accuracy and (3) perform historic data analysis to enable better insights into the results for human operators. This requires a mix of machine learning, stream and batch processing – technologies which are typically optimized independently. We combine all three into a single, real-world application. This paper describes the implementation and evaluation of an alarm verification system we developed jointly with Sitasys, the market leader in alarm transmission in central Europe. Our system can process around 30K alarms per second with a verification accuracy of above 90%.
Sima, A. C., Stockinger, K., Affolter, K., Braschler, M., Monte, P., & Kaiser, L. (2018). A hybrid approach for alarm verification using stream processing, machine learning and text analytics. In EDBT 2018, Vienna, Austria, 26-29 March 2018. ACM. https://doi.org/10.21256/zhaw-3487