Predicting the Effectiveness of Scientific Inquiry in Educational Technology in Moroccan Secondary Schools: A KNN-Based Analysis through Observations and Interviews

Abstract

The main objective of this study is to assess the extent to which machine learning models can predict the effectiveness of Scientific Inquiry Approaches (SIA) in learning and teaching within Moroccan secondary education. Data from 1,392 students, collected through classroom observations and interviews during the period 2021–2023, were used to develop a predictive analytics framework. The dataset was meticulously preprocessed by applying missing value treatment and normalization techniques to ensure robustness. The K-Nearest Neighbor (KNN) algorithm was implemented using the scikit-learn Python library. Model performance was evaluated using multiple metrics, including accuracy, precision, recall, F1-Score, specificity, false positive rate, Receiver Operating Characteristic (ROC) analysis, and Area Under the Curve (AUC). The results demonstrate strong predictive performance, with an AUC of 0.8875 for interview-based data and 0.9309 for observation-based data, corresponding to prediction accuracies of 90.2% and 94.5%, respectively. These findings indicate that machine learning is an effective tool for predicting the success of SIA teaching approaches. By integrating complementary data sources, this study provides novel evidence from the Moroccan educational context regarding prediction reliability. The findings have important implications for improving instructional practices, supporting data-driven decisionmaking, and informing educational policy. Future research may further validate the proposed framework by exploring additional machine learning algorithms and broader datasets.