Student Cheating Detection in Higher Education by Implementing Machine Learning and LSTM Techniques

Area of Science:

• Computer Science
• Educational Technology
• Data Science

Background:

• Academic integrity in online education is challenged by undetected cheating during exams.
• Traditional monitoring methods are insufficient, necessitating advanced detection techniques.
• Machine Learning (ML) offers potential for accurate identification of academic dishonesty.

Purpose of the Study:

• To propose and evaluate a novel ML-based method for detecting exam cheating incidents.
• To leverage the 7WiseUp behavior dataset for building predictive models.
• To enhance the accuracy of identifying potential academic dishonesty in online assessments.

Main Methods:

• Utilized the 7WiseUp behavior dataset, encompassing surveys, sensor data, and institutional records.
• Developed a predictive model using a long short-term memory (LSTM) neural network architecture.
• Employed Adam optimizer, dropout layers, and dense layers, with optimized hyperparameters and data preprocessing.

Main Results:

• The proposed ML model achieved a 90% accuracy rate in detecting potential exam cheating.
• This performance surpassed all previously reported methods in the literature.
• The enhanced accuracy is attributed to a sophisticated model architecture, optimized hyperparameters, and rigorous data preparation.

Conclusions:

• The novel ML approach demonstrates high efficacy in detecting academic dishonesty during online exams.
• Further research is needed to pinpoint the exact factors contributing to the model's superior performance.
• This method offers a promising solution for maintaining academic integrity in digital learning environments.