Time-Based Abnormality Identification in Uniform Datasets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality detection systems are unable to effectively identify abnormal data characteristics over a long period, particularly in fraud detection contexts, as they rely on single snapshots of data and fail to consider interactions between discrete and continuous data types, leading to missed fraudulent patterns.
Innovation Solution
The system employs vectorization processes to combine discrete and continuous data into a single vector, followed by dimensionality reduction using methods like hashing and auto-encoders, enabling long-term analysis and comparison of data patterns to detect abnormal behaviors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional single-snapshot data analysis is used, then system complexity is low, but the ability to detect long-term fraudulent patterns is poor
Solution Approach 1:
The patent segments data analysis into multiple time snapshots, analyzing data at different time points separately. This allows the system to capture temporal variations in fraudulent behavior while maintaining manageable complexity through modular processing of each time snapshot independently.
Solution Approach 2:
The patent adds the time dimension to data analysis by collecting and comparing data across multiple snapshots. This transforms single-point analysis into multi-dimensional temporal analysis, enabling detection of patterns that evolve over time without proportionally increasing system complexity.
2Reliability
If discrete and continuous data types are analyzed separately, then processing is simpler, but interactions between data types are missed
Solution Approach 1:
The patent merges discrete and continuous data types into a unified analysis framework. By combining these different data types and their interactions into single data records across multiple snapshots, the system detects patterns that involve relationships between discrete and continuous variables without managing them as completely separate processing streams.
3Reliability
If comprehensive temporal analysis is performed, then fraudulent patterns are better detected, but processing time increases
Solution Approach 1:
The patent applies partial action by selecting specific snapshots at key time points rather than continuously analyzing all data points. This approach captures essential temporal patterns for fraud detection while reducing overall processing time by focusing computational resources on critical moments in the data timeline.
Data Source
AI summary
Various embodiments provide systems and methods for detecting data abnormalities within data sets relating to a particular agent and comprising both discrete and continuous data features by encoding the one or more discrete features and generating a sequential feature vector representative of both the encoded discrete features and the continuous features, reducing the dimensionality of the generated sequential feature vector to generate a reduced dimension behavioral vector, and comparing the reduced dimension behavioral vector against other reduced dimension behavioral vectors.