This was our first week as interns working on the Future Learn MOOC data visualisation project. During this time we became acquainted with the general goals of the research and met some of the people that are involved with it. However, the specific project requirements will be discussed with the other researchers over the course of the following weeks.
Most of our work for this initial week was comprised of reading the same set of papers related to MOOC data mining, analysis and organisation. The remainder of this post consists of our individual accounts of the research we have done.
Most of the papers that I read this week came from the initial list of recommended reading given to us by our supervisor. Following is a brief overview of the goals and findings of each of these papers:
- MOOCdb – the initial introduction of the already established standardised database schema for raw MOOC data; the original proposal of the paper is a standardised, cross-course, cross-platform database schema which will enable analysts to easily work on the data by developing scripts; the intention is to build a growing community by developing a repository of scripts; the concept proposes the sharing of data without exchanging it
- MOOCViz – presents the implementation of the analytics platform envisioned in the MOOCdb paper; the framework provides the means for users to contribute and exchange scripts and run visualisations
- Learner Interactions During Online MOOC Discussions – Ayse’s paper from the WAIS group; investigates the relation between high attrition rates and low levels of participation in online discussions; provides a novel model of measuring learners’ interaction amongst themselves and offers a method of predicting possible future interactions; dividing the predictions in categories and the means of calculating friendship strength are particularly interesting
- Monitoring MOOCs – a paper that reports the findings of a survey of 92 MOOC instructors on which information they find most useful for visualising student behaviour and performance; it provides good insight for the types of data and visualisation that would potentially be useful for our project; additionally, it is a very good reference source for papers dealing with different visualisation methods for MOOC data
- Visualizing patterns of student engagement and performance in MOOCs – investigates high attrition rates; its main goals are to develop more refined learning analytic techniques for MOOC data and to design meaningful visualisations of the output; to do so it classifies student types by using learning analytics of interaction and assessment and visualises patterns of student engagement and success across distinct MOOCs; employs a structured analysis approach where specific variables and analyses results are determined iteratively at increasingly finer levels of granularity; utilises different visualisation diagrams that will likely be of interest for our project
- Analyzing Learner Subpopulations in MOOCs – again, investigates attrition; previous paper took inspiration from this one for its analysis and visualisation approach; interesting method for classifying students by engagement; uses k-means clustering
The research I have conducted during this week has helped me to familiarise myself with the concept of MOOCs data visualisation and analysis and the challenges associated with them. More broadly, it has given me an insight into educational data mining and learning analytics. However, there is still an abundance of research that needs to be done. I have found that I am lacking in knowledge of statistics which prevents me from fully understanding some of the papers. In addition, there is a plethora of possible visualisation tools and methods available so becoming familiar with them and choosing the right ones in the available project time will prove to be challenging.
Apart from paper reading this week I also completed the first three weeks of the Doing Your Research Project MOOC to become acquainted with the structure of a typical MOOC on the Future Learn platform.
A great number of researches are trying to find a suitable way to help instructors of MOOCs understand and analyse the interactions and performance of students. Because of the enormous amount of students enrolling in MOOCs, it is a big challenge for scientists to use this data. In the paper “MOOCdb: Developing Data Standards for MOOC Data Science”, the authors propose MOOCdb to manage data. MOOCdb adopts various strategies to make people use data more efficiently. For example, standardize data, data from different resources will be formatted in same way finally. In addition, what information is important and how it helps instructors to analyse the interactions of students are problems as well. Some researches propose that students’ interactions with courses should be determined by their grades and duration. Other researches realize different interactive patterns will also affect students’ performance. In Ayse’s paper, she proposes a strength value which can be worked out and predict the friendship between two students. It is quite interesting opinion. Although I have seen various ideas so far, it seems that it is not quite sufficient for me to do our project. Next week I plan to read more papers and do more research in this field.