DATA CLUSTERING USING WEIGHTED K-MEANS IN N-DIMENSIONAL SPACE FOR STUDENT PERFORMANCE ANALYSIS

The paper proposes a modified k-means algorithm for multidimensional space with feature weighting. The method enables clustering of student academic records considering subject priority. A mechanism for calculating weighted Euclidean distance between performance vectors is described. Experimental verification was performed on a sample of 10000 students with four clusters: "excellent", "good", "satisfactory" and "unsatisfactory". The results demonstrate 17% improvement in interpretation accuracy compared to classical approach when using weight coefficients.