Feature Dimensionality Reduction for Visualization and Clustering on Learning Process Data

Interactive digital learning media has been introduced widely to educational institutions in the past few years. It is developed with the primary goal of supporting students' learning activity delivered by sophisticated technology and gathering their learning process data to acquire some learning patterns. These patterns can be used to improve learning quality. However, due to the complexity and high dimensionality of the data, information mining often faces several obstacles. Recently, several methods of useful information extractions from complex data have been proposed; a potential one involves data visualization and clustering, which enables some degree of insight regarding the data. This paper introduces a data visualization and clustering of students' learning process data in Monsakun, an interactive digital learning media that focuses on exercising arithmetic word problems by using the integration of mathematical sentences. Principal Component Analysis (PCA) is implemented for visualization view, while the K-means algorithm is adapted for clustering purpose to represent a different point of view in presenting and understanding student behavioral data in the learning process. This paper reports a preliminary effort of the usage of both PCA and K-means as an analytical manner. It shows that data understanding by using visual analysis from high-dimensional data to a low-dimensional one is quite promising.