Learning Analytics for Smart Classroom System in a University Campus

Abstract

AbstractLearning Analytics (LA) allows a student to see his progress and allows an instructor to see comparative progress of the different learners involved in the class. LA is thus being incorporated in different educational settings such as in smart classrooms where students’ behaviour and performance are observed and analyzed. This chapter presents LA in a smart classroom using predictive models to assess formative assessment, attendance, and behavioural performance of students. The smart classroom is modelled by an automated attendance using RFID tags and an activity detection system using camera and sound sensor to record attendance and detect different behaviours of students. Moreover, an attendance model was implemented to analyze the students’ attendance record and their behavioural performance from the activity detection system using a predictive model. In addition, an automated quiz system was implemented in a web application to assess students along with its model to predict their formative assessment performance. To perform prediction, the three models have been trained using different machine learning algorithms where the most accurate models are deployed in the web application. The attendance model was trained using five different multiclass classification algorithms namely, decision tree (100%), logistic regression (78.7%), Naïve Bayes (83.2%), KNN (98.4%) and Adaboost M1 (73.5%). Similarly, the formative performance model was trained using decision tree (99.1%), logistic regression (80.8%), Naïve Bayes (92.5%), KNN (94.6%), and Adaboost M1 (87.1%). The behaviour model is evaluated using four multiclass machine learning algorithms namely, decision tree (100%), logistic regression (91%), Naïve Bayes (95%), and KNN (100%). Since both decision tree and KNN have the same accuracy, fivefold cross-validation technique is used to differentiate their accuracy. Moreover, the chosen prediction models are evaluated by comparing the accuracy along with the f1-score, precision, and recall. The smart classroom system consists of additional functionalities like addition or removal of a quiz, ranking of students, quiz history, and graphical display of performance for both students and lecturers to see.KeywordsInternet-of-ThingsSmart classroomSensorsLearning analyticsMachine learning