Description

There are several applications where demand to determine unexpected activities in a sequence of time-stamped observation data—for occurrence, may need to identify strange events in transactions at a website or in video of an airport tarmac. In this paper, start with a known set A of activities (together innocuous and dangerous) that we desire to observer. Though, in accumulation, we desire to categorize “unexplained” subsequences in an observation sequence that is poorly described (e.g., because they may contain occurrences of activities that have never been seen or anticipated before, i.e., they are not in A). We properly describe the probability that a sequence of explanations is inexplicable (totally or partially) w.r.t. A. We progress efficient algorithms to recognize the top-k completely and partially unexplained sequences w.r.t. A. These algorithms power propositions that allow us to quickness the search for totally/partially unexplained sequences. We define tests using real-world video and cyber-security data sets presenting that our method mechanism glowing in training in relationships of together running time and accuracy.