K-means clustering algorithms used in the evaluation of online learners' behaviour

K-means clustering algorithm is used to analyse plenty of learners' behaviour data stored on the online learning platform. The learning behaviour data, basic information, and user types and factors affecting performance of online learning learners are firstly analysed and explored. Secondly, based on feature selection and optimisation algorithm of initial clustering centre, a K-means feature selection algorithm is proposed, and an equilibrium discriminant function is presented to balance the difference between the clusters and within the clusters. Finally, the clustering centre obtained by K-means feature selection algorithm is used as the centre of the neural network. The parameters, input and output variables of the prediction model are set. Based on the radical basis function (RBF) neural network structure, the performance prediction model is constructed, which dynamically updates to enable accurate performance predictions. The results show that the performance prediction model proposed has high prediction accuracy for online learners' performance.