Falkonry Technical Overview
This paper describes a general-purpose system and an associated approach for condition recognition from time series data, and it discusses how this system and approach can be applied to a wide array of operations management solution needs. There are many operational situations where sensor data is available and condition understanding is desired, but where constraints limit the application of machine learning or other techniques – constraints could include financial and time constraints, lack of behavioral models, lack of data science resources, or lack of labeled training sets. The system and approach described in this paper support pragmatic condition recognition in the presence of these constraints. The described system includes implementations of feature learning (with a focus on temporal pattern recognition), clustering, and classification techniques integrated to support unsupervised and semi-supervised learning for diverse sets of multivariate signal data. A continuous learning approach is employed where, initially, only unnamed recurring patterns might be recognized through unsupervised learning.
Please fill up the form below to download this whitepaper.