For instance, it may be necessary to spot odd occurrences in online transactions or in footage of an airport tarmac in a number of applications where it is necessary to recognise unusual behaviours in succession of timestamped observation data. In this essay, we'll begin by focusing on a well-known group A of behaviours (both harmless and harmful combined) which we want to observe. Even if we want to label "unexplained" subsequences in a poorly characterised observation sequence (For instance, they can include instances of actions that have never been witnessed or expected, therefore they do not fall under category A).We correctly define the likelihood that a series of explanations is unexplainable (completely or partially) in terms of A. To find the top k entirely and partially unexplained sequences with respect to A, we advance effective methods. We can rapidly accelerate the search for completely or partially unexplained sequences using the strength of these algorithms' proposals. We construct tests utilising data from cyber-security and real-world video sets, demonstrating how our technique mechanism shines in training in terms of the correlation between running time and accuracy

.