Exploratory learning analytics methods from three case studies

Abstract

Brief outlines of exploratory analysis methods (analysis designed to develop hypotheses) from three research projects illustrate the size, scope, variety and increased resolution that are becoming increasingly available at the unit of analysis for research in the learning sciences. The tools and methods applied in these studies are briefly outlined, which enable researchers to deal with complexity in time and event structures involving complex data in learning analytics projects. In particular, the transformation of data involving both reduction methods and pattern aggregation into motifs were found to be crucial for data interpretation. The article describes data mining with a self-organizing map, involving unsupervised machine learning and symbolic regression and combining exploratory analysis methods to achieve causal explanations.