Learning analytics systems collect, measure and report data about learners to support instructors to understand students’ learning and make informed interventions to optimize learning. Existing technologies mainly make use of log data collected from the learning environment and aggregate statistics such as no. of access to learning resources, learning activity attempts and performances. Those statistics will be used as features of students’ learning to derive learning patterns or train predictive models. While the features related to resource accesses and performances are insightful, the data that captures the process of how students solve the problems in learning activities could provide teachers more insight on students’ mastery of the knowledge and problem-solving strategies. In our learning analytics system for step-based learning activities, we collect, analyze and visualize students’ problem-solving steps and patterns. By correlating the analysis result with resource accesses and performance, teachers, as well as students themselves, can have a more informed intervention/reaction on how to improve learning and achieve learning outcome.
Figure 1 below shows the block diagram of the proposed system which consists of a database for raw logs, data processing pipeline, database for computed features and visualization system. The database persists logs of students’ actions on the online learning platform; the raw log data will be processed by the data processing pipeline to extract features such as problem-solving steps, patterns, strategies as well as statistics on resource accesses and performances. The analyzed result will be stored and feed to visualization system for teachers and students to analyze the learning.
The detailed process flow of the data processing pipeline is presented in Figure 2. The raw log data is first to be processed as learning actions and other learning data and stored in the temporal model; afterwards, learning algorithms and aggregations will be carried out based on the processing pipeline to enrich the temporal model and build up summary model.
The technology could be used for analytics and business intelligence on sequence-based tasks and processes. Adopting the technology, features in the sequences could be extracted, analysed and presented.
Currently, for setting up analytics functions, companies could either making use of basic data processing and visualization libraries to build the whole analytics system from scratch, enabling the total customization of the tools; or, making use of existing systems such as Tableau and QlikView to realize the required functions on the system, lacking of the ability for free customization. Our technology tried to provide a solution in between the two approaches if the company would like to analyze sequential processes.
The technology provides a generic analytics framework. Adopting the framework, customers could setup analytics and business intelligence functions easily on top of their existing technology infrastructure.