Analytics@TP Pre resen ented ed by: : Michael Yap 2018-09-28 - - PowerPoint PPT Presentation
Analytics@TP Pre resen ented ed by: : Michael Yap 2018-09-28 Agenda Our Analytics Journey Capability Development Challenges Sample of Data Products Student Analytics Learning Analytics Graduate
2018-09-28
Pre resen ented ed by: : Michael Yap
2018-09-18 2
2018-09-18 3
Gartner Analytics Ascendancy Model
4
Raw Data & Reports
Data Information Knowledge Insight Action
What can we analyse ?
5
Talent Mgmt Manpower Optimisation HR Attrition Capability Development
Staff Finance Procurement Estates & Facilities IT Industry Relation Customer Satisfaction Students
Attendance Acad Performance Financial Assistance/ Awards Admission Learning Engagement CCA Graduation Internship / OCP Attrition
Alumni
TP Analytics Roadmap
6
Finance Analytics Utility Analytics Industry Partner Analytics
Outreach Analytics
Alumni Analytics
Social media Analytics
Data Scientist Training
Student Support Analytics Learning Analytics Procurement Analytics Graduate Analytics Student Analytics HR Analytics IT Resource Analytics
Awareness Training Visual Analytics Training Data Analyst Training
Early Learning Analytics Project
tool for School Directors and Course Manager to gain academic insights – on course and subject performance
7
2018-09-18 8
Capability Development
9
Start fr from basi asic : : Awareness
lassroom le lear arnin ing
Learning
On-the-job-train inin ing
ication
(CoP)
Sharin ing
2018-09-18 10
Challenges
Data sources and Quality
Structured vs s unstructured data
Data consis istency
Data qual alit ity is is im important in in producin ing meanin ingful l resu sults
11
Familiarisation with tools
New an anal alyt ytic ical l tools an and systems
Different tools ls for dif ifferent role les - Bac ackend, fr frontend, Admin inistration
Competency Building
Learnin ing Curve
Lack of
skill illed personnel l in in busi siness an anal alyt ytic ics
ith dom
in experts an and IT IT ap appli lication teams
Challenges
12
System Performance
loadin ing an and resp sponse tim time
Drill ill down, drill rill th through
Data si size doe
Access Control
Different fr from tr transaction system
self lf-servic ice
Change Management
Different min indset – data-driven decis ision makin ing
Strategic vs s op
Descriptive to
analyt lytics
2018-09-18 13
Student Analyt lytics
Conceptual Architecture
14
Data Sources
Performance
Demographics
Management System
Employment
Data Management
Extract, Transform Load Data Marts Descriptive Analytics Predictive Analytics
Analytics Visualisation
Learning Analytics Student Analytics Procurement Analytics Graduate Analytics Top Performer / At-Risk students
Business Users
Oracle, MS SQL Files SQL Server Integration Services SAS Data Integrator SAS Visual Analytics SAS Enterprise Miner
Insight Foresight
MS SQL
Periodic refresh
15
16
Sch A Sch B Sch C Sch D Sch E Sch F
AAA BBB CCC DDD EEE FFF GGG HHH
Graduation/Attrition Report
17
Admission Category
18
Entry Qualification
19
20
Predictive models were built
2018-09-18 21
Learning Analyt ytics
22
LMS Content Access
23
Filters Students’ Access Patterns
Student Workload Distribution
24
2018-09-18 25
Graduate Analyt ytics
26
To distil key drivers for graduates’ outlook to enable personalised interventions Predict Graduates’ Outlook
Will the graduate be economically active? working in field related to studies? engaged in further study? etc.
Distil underlying Key Drivers
Demographic Entry Qualification Academic Performance Financial etc.
Enable Intervention
(every semester or as needed)
Generate propensity of students at the end of each semester for ‘personalised’ interventions
Demographics
Entry Qualification
Category Group
Qualification
Academic Performance
CDS Subjects passed / failed
Financial
Non-academic
Disciplinary / Attendance
Record
Utilise data in TP source systems to study students’ behaviour comprehensively
Input Considerations
27
Model Building
28
Data Loading
unnecessary variables
Data Partition
validation data
Score Code Export
Model*
(Decision Tree)
model(s) with the best balanced
Data Scoring
with the validation data
All Students
Job Market Leakage: 32% Others: 68%
Decision Tree
widely used by organisations for its intuitiveness and business interpretability
Job Market Leakage: 26% Others: 74% Job Market Leakage: 40% Others: 60% Job Market Leakage : 36% Others: 64% Job Market Leakage : 21% Others: 79% School A, … School B, … School X Y Gender Job Market Leakage : 32% Others: 68% Job Market Leakage : 49% Others: 51% < 17.5 17.5 O-Level Raw Aggregate Job Market Leakage : 35% Others: 65% Job Market Leakage : 21% Others: 79% <3 >=3 GPA
High-risk groups Lower-risk groups
For Illustration: Job Market Leakage Model (extract)
Key Modelling Methodology
29
[For illustration only]
2018-09-18 30
Procurement Analyt ytics
32
too stringent or geared towards a particular brand of item?
Spending patterns ?
33
stakeholders to review the data so that necessary intervention can be considered at different stages of the procurement process.
can also be used for audit function.
2018-09-18 34
– Teaching Effectiveness, Subject Review, Course Review
– E.g. Online Student Evaluation of Teaching (OnSET) collects about 100,000 free-text comments annually
Leverage on technology to analyse free-text comments, so as to gain insights on themes and sentiments
Objective
36
37
quantifiable results
Dictionary’
38
Text Analytics
Q11: Write down something that your lecturer has done especially well
Female students reflect more positively on the learning experience. Key insights for further research:
2018-09-18 39
40
Car Plate Number Recognition
41
Using CCTVs and video analytics for people counting.
2018-09-18 42
2018-09-18 43
2018-09-18 44
Our journey continues…
2018-09-18 45