Behavioral Deviation Detection Using Prediction Interval Profiles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in accurately identifying deviated or outlying behavior in individuals through predictive data analysis, particularly in monitoring behavioral activities with low frequency occurrences.
Innovation Solution
The method involves generating a prediction interval profile using sensor data from historical time periods, which includes predicted count intervals and predicted timing intervals for each behavioral activity. This profile is used to classify behavioral activities as normal or deviated by comparing observed activity counts and timings with the predicted intervals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional behavioral monitoring methods are used, then all sensor data is stored and analyzed, but data storage space and computing resources are wasted
Solution Approach 1:
The system performs preliminary actions by generating prediction interval profiles from historical sensor data before actual behavioral monitoring. These profiles contain predicted count intervals and timing intervals for various behavioral activities, establishing a baseline for comparison that eliminates the need to store and analyze all raw historical sensor data
Solution Approach 2:
The system extracts only the essential information needed for behavioral deviation detection by converting historical sensor data into compact prediction interval profiles. These profiles contain only the critical statistical parameters (count intervals and timing intervals) required for anomaly detection, separating necessary information from unnecessary raw data
2Measurement precision
If all historical sensor data is stored for analysis, then comprehensive behavior patterns can be analyzed, but computing resources are consumed
Solution Approach 1:
The system extracts only the essential statistical features from historical sensor data to create prediction interval profiles, which contain predicted count intervals and timing intervals. This extraction process converts large volumes of raw sensor data into compact statistical representations that require minimal computing resources for comparison and anomaly detection
Solution Approach 2:
The system performs preliminary processing of historical sensor data to generate prediction interval profiles before actual behavioral monitoring begins. This preliminary action pre-computes the statistical parameters needed for deviation detection, eliminating the need for computationally intensive real-time analysis of raw historical data
3Measurement precision
If prediction interval profiles are generated using relevant historical time periods, then classification accuracy is maintained, but data storage is reduced
Solution Approach 1:
The system applies local quality by selecting and generating prediction interval profiles from specific historical time periods that are locally relevant to the prediction target. Instead of using all historical data uniformly, the system identifies and utilizes only the locally optimal historical periods that best match the characteristics of the prediction time period, thereby maintaining classification accuracy with reduced data quantity
Data Source
AI summary
Various embodiments provide methods, apparatus, systems, computing entities, and/or the like, for identifying deviated behavior of an individual indicated by outlying activity counts and outlying activity timings classified by a prediction interval profile generated from historical behavior data. In an embodiment, an example method comprises accessing sensor data describing behavioral activities of an individual during historical time periods and generating a prediction interval profile for the behavioral activities comprising a predicted count interval within a prediction time period for at least some behavioral activities. The method further includes receiving sensor data for the prediction time period and extracting an observed activity count and an observed activity timing for each behavioral activity. The method then includes classifying a particular behavioral activity as a behavioral deviation if the observed activity count does not satisfy the predicted count interval and/or if the observed activity timing does not satisfy the predicted timing interval.