Online Proctoring Incident-Tree Regression for Aberrant Behavior Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional ML-based methods for online proctoring fail to provide a holistic analysis of candidate behavior during examinations, missing the sequence of incidents that could indicate aberrant behavior, thus making online examinations vulnerable to cheating.
Innovation Solution
A regression-based method and system that utilizes an incident tree construction technique to analyze a sequence of incidents, computing weighted sums and path scores to predict aberrant behavior, incorporating machine learning for dynamic adjustments and comparisons with historical data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional ML-based methods compare pre-defined parameters with extracted user parameters, then the system can determine violations of assessment criteria, but the system fails to provide holistic analysis of candidate behavior sequences
Solution Approach 1:
The patent segments the analysis into multiple dimensions by creating separate detection modules for different types of aberrant behaviors (facial expressions, eye movements, keyboard patterns, mouse movements). Each module specializes in detecting specific behavioral patterns, allowing comprehensive analysis without overwhelming complexity in a single system.
Solution Approach 2:
The patent adds temporal dimension to the analysis by examining sequences of incidents rather than isolated events. The system analyzes the chronological order and patterns of multiple incidents to detect aberrant behaviors, transforming static parameter comparison into dynamic sequence analysis.
2Adaptability or versatility
If the system analyzes all possible parameters to predict aberrant behavior, then holistic analysis is achieved, but the computational complexity and processing time increase significantly
Solution Approach 1:
The patent applies different analysis strategies to different types of data with different levels of importance. Critical behaviors like facial expressions and eye movements receive more intensive analysis with multiple detection algorithms, while less critical parameters use simpler detection methods, optimizing resource allocation across diverse data types.
Solution Approach 2:
The system dynamically adjusts detection thresholds and sensitivity levels based on the examination context and candidate behavior patterns. By changing parameters adaptively rather than using fixed thresholds, the system achieves versatile detection across different scenarios without requiring proportional increases in system complexity.
3Reliability
If real-time detection of aberrant behavior is implemented, then immediate alerts can be generated to prevent malpractice, but the system requires sophisticated processing capabilities
Solution Approach 1:
The system performs preliminary analysis by continuously monitoring and storing sequences of incidents during the examination. Rather than analyzing all data at once, the system prepares incident sequences in advance and applies detection algorithms to pre-processed data, enabling real-time detection without requiring peak processing power during critical decision moments.
Solution Approach 2:
The patent introduces an intermediary layer that translates complex multi-dimensional behavior data into simplified incident sequences and scores. This intermediary processing layer converts raw sensor data from multiple sources into standardized incident representations, reducing the complexity burden on the final detection and decision-making systems.
Data Source
AI summary
Examinations are used to determine the ability of a candidate such as a student or prospective practitioner as it pertains to proficiency in a particular subject or skill set. Many standardized tests are now administered online, and online examinations are difficult to proctor. One solution to overcome the above challenge is to predict the behaviour of the candidate using Machine Learning (ML). However, conventional ML based methods fail to consider sequence of incidents pertaining to a candidate during online proctoring for assessing the behavior of the candidate. To overcome the above challenges, embodiments herein provide a method and system for regression based prediction of aberrant behavior in online proctoring. The present disclosure generally concerns with exam proctoring and more specifically focus on detecting of aberrant behaviors in advance during examination by analyzing a plurality of incidents pertaining to a candidate using incident tree based regression analysis.