digital & data the power of learning analytics june 2016 · jisc uk learning analytics the...
TRANSCRIPT
June 2016The key role of relational, student & other data sources in developing Learning Analytics
Digital & Data – The power of Learning Analytics
Jisc UK Learning Analytics
The Business Case: Improve student retention, achievement & employability
Method: The application of big data techniques such as machine based learning and
data mining to help learners and institutions meet their goals.
Research & Development: Statistical analysis of historical and current data derived from
the learning process to create models that allow for predictions that can be used to
improve learning outcomes. Models are developed by “mining” large amounts of data to
find hidden patterns that correlate to specific outcomes.
Key Outputs: Secure & rich multi-tenanted data warehouse, provision of early alerts to
members of staff - enabling early interventions to students, and further BI from data;
Enabling Services: Intuitive bilingual software tools and apps to automate, facilitate &
engage both student and staff to realise LA benefits;
About the
student
(UDD)
Activity data
(xAPI)
Current Progress – June 2016• Discovery readiness reviews completed/ ongoing
• Deployment of our LA Data Access Agreement (DAA)
• Review latest student data specification (UDD v1.2.4) -
https://github.com/jiscdev/analytics-udd
• Ongoing Technical Trials: set-up Learner Records Warehouse(s), install VLE
plugin(s), perform ETL for UDD student data output
• Predictive Model development – service pilots start Q4 2016
• BlackBoard Learn – activity data building block development/ evaluation
• Student App – Beta v1.0 due for release June/ July 2016 (iOS/ Android)
• Student App evaluations – currently being formulated for 3 HEIs
Progress in numbers…• Expressions of interest: 85
• Engaged in activity: 24
• Discovery to Sept 16: agreed (20), completed (11), reported (6)
• Over 4 million activity (engagement) records collected in real-time
• Moodle Historic Data Transformation: 41 million records
transformed from Moodle log files to xAPI
• Blackboard Data Transformation: 12 million blackboard records
transformed from Blackboard log files to xAPI
UDD – Unified Student Data Definition (v1.2.4)
• Student data will be sourced to enable all LA services – historical & live data collections
• Personal and Professional (academic progression) data, per individual student (HE & FE)
• Peace of mind through all-encompassing Jisc DAA
• Structure similar to HESA - a near-subset of the student return (ER, fields, mappings)
• HEDIIP/ HESA Data Futures – align to evolving structures for future data landscape;
• Interventions need to work properly – therefore be timely
• Correct (quality) data needs to be captured as frequently as possible
• Systems & business processes + data-input efficiency required
Let’s talk about data…• Historical data (12-18 months minimum) for a Learning Analytics implementation
• Predictive Model Creation (12-36months – ideal!) & Pilot
• (Portable) Predictive Model Validation (12months minimum) & Pilot
• Engagement & activity (xAPI) data – VLE (Bb, Moodle), LMS, AMS and more…
• Moodle & Bb xAPI live plug-ins for activity, Moodle (log extract) & Bb (Plugin) for xAPI historical
• No (or not good enough!) historical data?
› Confirm/ refine retention and data recycling measures
› Guidance on data quality, reduce gaps etc.
› Use other predictive models derived from the project – validate now, create later
› Evaluate, influence and co-design other LA service offerings – SSP, Student App, dashboards…
• Individual assignment (coursework) and examination marks required per module – with dates!
• Data transformation: in-house or using off-the-shelf software tool (Kettle, Alteryx…)
A Granular View of Data Sources
Student Insight Overview
Data collection and
Data mining
Create models
Predict and understand
Patterns
RelationshipsTrends
Behaviours
Collect
SIS integration
Student activity data
1
Identify
Outcome influences
Risk prediction
Student tagging
2 Awareness
Monitor cohorts and tags
Student engagement
Understand risk influences
3 Act
Proactively manage progress
Record decisions and actions
Manage student interventions
4
Improve
Assess intervention effectiveness Student feedback
5
Student Insight – 5 Stage Process
Press Coverage – Times Higher (Feb 2016)
Press Coverage – The Independent (June 2016)
Press Coverage – Scottish Herald (May 2016)
Get involved!
• Do you foresee any issues or problems with near-LIVE student data capture
or timely data extraction for Learning Analytics?
• Are there any resource implications for you, in making Learning Analytics
happen at your institution?
• Can student assessment information be easily extracted from your system(s)
per module/ student? Is the information time indexed?
• What other sources/ good indicators of student engagement or activity exist?
• Turnitin, specific attendance monitoring systems…
Demonstration
AgendaRisk prediction
Academic performance
risk prediction
Course withdrawal
risk prediction
Monitor groups
and individual
AgendaIdentify
Tag students
at risk
Monitor risk by
student cohort
See what factors affect
outcomes
AgendaOutcome analysis
See what factors affect
outcomes
AgendaUnderstand risk influences
Prediction trend
View what influenced
the prediction
Compare to student
cohorts
AgendaRecord decisions and actions
Log decisions
and actions
Maintain history of decisions
made
AgendaManage student interventions
Inform student support
team
Cases allocated
to student support
staff
Track cases through to completion
AgendaAssess intervention effectiveness
View impact of
actions made
Take into account student
feedback
Thank you
Rob Wyn Jones – Consultant – JISC Special Projects
Craig Petch – Tribal – Consultant, Student Insight