NEC Labs America’s Time Series Data Research Drives Space Systems Innovation
With decreasing hardware costs and increasing demand for autonomic management, many of today’s physical systems are equipped with an extensive network of sensors, generating a considerable amount of time series data daily.
Contributor: Haifeng Chen, Department Head, Data Science & System Security
A highly valuable source of information, time series data is used by businesses and governments to measure and analyze change over time in complex systems. Organizations must consolidate, integrate and organize a vast amount of time series data from multiple sources to generate insights and business value.
When processed and analyzed correctly, time series data helps organizations understand the underlying causes of trends or patterns over time. Examples include seasonal retail shopping patterns, identifying and predicting economic trends, signal processing for oil and gas exploration, understanding long-term weather patterns, and predictive maintenance on complex machinery ranging from automobiles, trains, airplanes, and manufacturing and IT systems.
Due to the high system complexity, however, those time series contain heterogeneous dependencies among different parts across the system and are mixed with noise and operational patterns. It is challenging to correctly discover the system’s operational status and healthiness from the collected data. There are typically four different variations of time series data analysis:
- Seasonal – short-term changes that occur periodically over the course of a year
- Trend – patterns over more extended periods
- Cyclical – regularly occurring fluctuations around a trend
- Random – changes to variables related to unknown factors
NEC Labs America researchers are working to transform the time series data into insight around a system’s productivity, reliability and safety. We discover the “invariants” from massive time series to reduce the system complexity, and leverage those invariants for system management across a wide range of business use cases. We’re developing deep learning-based time series retrieval to digest and compress historical time series for explaining current observations.
We also leverage the attention mechanism and the prototype learning to discover hidden patterns from time series.
Solutions developed by NEC Labs America are differentiated from competitive offerings by their ability to model dependencies at scale as well as proven success in mission-critical operations for many organizations. For example, we’re developing new approaches to help clients, such as Lockheed Martin, better analyze the telemetry data from complex systems to improve space craft production. At Lockheed Martin, our cutting-edge research is being used to gain holistic views of space systems using AI. This allows the company to drive proactive anomaly detection during the design, development, production, and testing of spacecraft – even before applications are in mission operations. Applying big data, AI and ML capabilities will enable significant cost savings on system production while improving mission assurance.
The dynamics with times series data are constantly changing. NEC Labs America will continue to advance technologies to optimize an increasing number of mission-critical workloads.