FRAMEWORK FOR COMBATING EXAMINATION MALPRACTICE IN NIGERIAN SECONDARY SCHOOLS: A SENTIMENT ANALYSIS AND MACHINE LEARNING APPROACH

Malpractice in examination in Nigeria and the Sub-Saharan African region as a whole is a multifaceted
educational crisis that has its origins on the systemic and societal level. Increased attention to the problem
was on the application of the traditional approach, but this paper suggests a proactive model based on the
utilization of the artificial intelligence (AI) and sentiment analysis. It is structured to identify psychologi-
cal antecedents to misconducts, which are frustration, distrust, and intent to cheat, through an analysis of
sentiment-infused data among the relevant stakeholders in the high schools. The data is gathered by the
use of audio interviews, transcribed, and annotated into three sentiments, and before it is preprocessed with
techniques like Term Frequency-Inverse Document Frequency (TF-IDF) vectorization. To prevent the pos-
sibility of class imbalance, the framework uses SMOTE prior to training and testing of the four machine
learning classifiers, whose results include; Logistic Regression, Random Forest, Multinomial Naive Bayes
and Support Vector Machine (SVM). A set of metrics is used to assess performance, and these metrics are
precision, recall, F1-score, ROC-AUC, MCC, Cohen Kappa, and balanced accuracy. The proposed system
architecture will recognize the underlying sentiment patterns in relation to malpractice, which will form
the basis of the evidence-based, data-driven interventions. This framework is a new and holistic way of
protecting academic integrity and promoting sustainable educational reforms in secondary education system
in Nigeria by replacing a reactive to a predictive approach. Key Words: Artificial Intelligence, Educa-
tional data mining, Machine learning, Sentiment analysis, Term Frequency-Inverse Document Frequency,
SMOTE, Predictive framework.