Data Mining to Predict 
Student Success 
in a Large Class 
Perry Samson 
College of Engineering 
University of Michiganr 
samson@umich.edu 
and Hybrid @pjsamson
What Affects Student Learning? 
Activities 
Grades
What Affects Student Learning? 
Intrinsic Influences 
• Before class 
• During class 
• After class 
Extrinsic Influences 
• History 
• Motivation 
• Wellness
MONITORING STUDENT PARTICIPATION 
Type notes here 
Type Notes 
synchronized to 
slides 
Pose/review 
questions 
Browse 
slides 
Annotate 
slides 
Bookmark 
slides 
Note confusion 
Learning Analytics: Data Collection
MONITORING STUDENT PARTICIPATION 
Learning 
Analytics 
Database 
Questions 
Posed 
Confusion 
Activity 
Answers 
Notes 
Attendance 
Bookmarks 
Lecture 
Capture 
Learning Analytics: Data Collection
MONITORING STUDENT PARTICIPATION 
LMS 
University 
db 
SIS 
Other 
Vendors 
Learning 
Analytics 
Database 
Questions 
Posed 
Confusion 
Activity 
Answers 
Notes 
Attendance 
Bookmarks 
Lecture 
Capture 
Learning Analytics: Data Collection
Question #1 
What Aspects of Student 
Participation Affect 
Student Outcomes?
Learning Analytics: Extrinsic Affects 
Lesson #1 
Student outcomes 
are modestly related 
to the students’ 
motivations. 
100 
90 
80 
70 
60 
50 
Low 
Interest 
in 
Some 
Interest 
Medium 
Interest 
Exam 1 
Exam 2 
Exam 3 
High 
Interest 
Truly 
Excited 
by 
Exam Grades 
Level of Interest
Learning Analytics: Extrinsic Affects 
“Where on this wellness chart 
would you put yourself today?”
100 
90 
80 
70 
60 
50 
Average Exam Grades 
Learning Analytics: Extrinsic Affects 
Physical State Emotional State 
Exam 1 
Exam 2 
Exam 3 
Average Physical State 
100 
90 
80 
70 
60 
50 
Avergae Exam Grade 
Exam 1 
Exam 2 
Exam 3 
Average Emotional State 
Lesson #2 
Student outcomes are somewhat related to student wellness 
.
Learning Analytics: Intrinsic Effects 
160 
140 
120 
100 
80 
60 
40 
20 
0 
Number of Students 
In the classroom From my school residence From elsewhere Missing
In a Hybrid Class Outcomes are Unrelated to Location 
Lesson #3 
Student outcomes 
are unrelated to 
whether students 
participate face-to-face 
or remotely. 
100 
90 
80 
70 
60 
50 
0 1-2 3-5 >5 
Average Exam Grade 
Number of Times Participating 
Remotely 
Exam 1 
Exam 2 
Exam 1 
Learning Analytics: Intrinsic Effects
Learning Analytics: Intrinsic Effects 
Outcomes ARE Related to In-Class Participation 
Lesson #4 
Student outcomes 
are related to their 
level of 
participation* in 
class.
Learning Analytics: Intrinsic Effects 
Outcomes ARE Related to In-Class Performance 
Lesson #5 
Student outcomes 
are related to 
their performance 
in class.
Learning Analytics: Intrinsic Effects
Learning Analytics: Intrinsic Effects 
Lesson #6 
Student outcomes 
are related to the 
number of slides 
for which they take 
notes. 
100 
95 
90 
85 
80 
75 
70 
65 
Exam Grades 
Exam 1 
Exam 2 
Exam 3 
Number of Slides Containing Notes
Question #2 
DO HIGHER GPA 
STUDENTS BEHAVE 
DIFFERENTLY THAN 
LOWER GPA STUDENTS?
Participation vs. GPA 
Learning Outcomes are Related to Incoming GPA 
100 
90 
80 
70 
60 
50 
40 
1.0 2.0 3.0 4.0 
Exam Grades 
GPA 
EXAM1 
EXAM2 
EXAM3
Participation vs. GPA 
Participation as a Function of Incoming GPA 
Fraction of Questions Attempted 
Fraction Correct of Gradeable Questions 
Ratio of Percent Correct to Percent Attempted 
1.0 
0.8 
0.6 
0.4 
0.2 
0.0 
Avgerage Fraction 
Incoming GPA 
Lesson #7 
Students with 
lower incoming 
GPA answered 
about 70% of 
questions in 
class versus 
85% for higher 
GPA students.
Participation vs. GPA 
Participation as a Function of Incoming GPA 
Lesson #8 
Students with 
lower incoming 
GPA took notes 
on 5X fewer 
slides. 
3600 
3200 
2800 
2400 
2000 
1600 
1200 
800 
400 
0 
180 
160 
140 
120 
100 
80 
60 
40 
20 
0 
Avg. Num. Slides 
w/Notes 
Avg. Words 
Typed per Class 
<3.0 3.0-3.6 >3.6 
Number of Words Typed 
Number of Slides w/Notes 
Grade Point Average
Participation vs. GPA 
Participation as a Function of Incoming GPA 
20 
18 
16 
14 
12 
10 
8 
6 
4 
2 
0 
Avergae Days Participated by Location 
Participated in Classroom 
Participated Remotely 
Missed class 
Incoming GPA 
Lesson #9 
Students with 
lower were far 
more likely to 
participate 
remotely than 
higher GPA 
students.
Thank you 
http://www.sageonstage.com 
@pjsamson

Data Mining to Predict Student Success @A_L_T #altc

  • 1.
    Data Mining toPredict Student Success in a Large Class Perry Samson College of Engineering University of Michiganr samson@umich.edu and Hybrid @pjsamson
  • 2.
    What Affects StudentLearning? Activities Grades
  • 3.
    What Affects StudentLearning? Intrinsic Influences • Before class • During class • After class Extrinsic Influences • History • Motivation • Wellness
  • 4.
    MONITORING STUDENT PARTICIPATION Type notes here Type Notes synchronized to slides Pose/review questions Browse slides Annotate slides Bookmark slides Note confusion Learning Analytics: Data Collection
  • 5.
    MONITORING STUDENT PARTICIPATION Learning Analytics Database Questions Posed Confusion Activity Answers Notes Attendance Bookmarks Lecture Capture Learning Analytics: Data Collection
  • 6.
    MONITORING STUDENT PARTICIPATION LMS University db SIS Other Vendors Learning Analytics Database Questions Posed Confusion Activity Answers Notes Attendance Bookmarks Lecture Capture Learning Analytics: Data Collection
  • 7.
    Question #1 WhatAspects of Student Participation Affect Student Outcomes?
  • 8.
    Learning Analytics: ExtrinsicAffects Lesson #1 Student outcomes are modestly related to the students’ motivations. 100 90 80 70 60 50 Low Interest in Some Interest Medium Interest Exam 1 Exam 2 Exam 3 High Interest Truly Excited by Exam Grades Level of Interest
  • 9.
    Learning Analytics: ExtrinsicAffects “Where on this wellness chart would you put yourself today?”
  • 10.
    100 90 80 70 60 50 Average Exam Grades Learning Analytics: Extrinsic Affects Physical State Emotional State Exam 1 Exam 2 Exam 3 Average Physical State 100 90 80 70 60 50 Avergae Exam Grade Exam 1 Exam 2 Exam 3 Average Emotional State Lesson #2 Student outcomes are somewhat related to student wellness .
  • 11.
    Learning Analytics: IntrinsicEffects 160 140 120 100 80 60 40 20 0 Number of Students In the classroom From my school residence From elsewhere Missing
  • 12.
    In a HybridClass Outcomes are Unrelated to Location Lesson #3 Student outcomes are unrelated to whether students participate face-to-face or remotely. 100 90 80 70 60 50 0 1-2 3-5 >5 Average Exam Grade Number of Times Participating Remotely Exam 1 Exam 2 Exam 1 Learning Analytics: Intrinsic Effects
  • 13.
    Learning Analytics: IntrinsicEffects Outcomes ARE Related to In-Class Participation Lesson #4 Student outcomes are related to their level of participation* in class.
  • 14.
    Learning Analytics: IntrinsicEffects Outcomes ARE Related to In-Class Performance Lesson #5 Student outcomes are related to their performance in class.
  • 15.
  • 16.
    Learning Analytics: IntrinsicEffects Lesson #6 Student outcomes are related to the number of slides for which they take notes. 100 95 90 85 80 75 70 65 Exam Grades Exam 1 Exam 2 Exam 3 Number of Slides Containing Notes
  • 17.
    Question #2 DOHIGHER GPA STUDENTS BEHAVE DIFFERENTLY THAN LOWER GPA STUDENTS?
  • 18.
    Participation vs. GPA Learning Outcomes are Related to Incoming GPA 100 90 80 70 60 50 40 1.0 2.0 3.0 4.0 Exam Grades GPA EXAM1 EXAM2 EXAM3
  • 19.
    Participation vs. GPA Participation as a Function of Incoming GPA Fraction of Questions Attempted Fraction Correct of Gradeable Questions Ratio of Percent Correct to Percent Attempted 1.0 0.8 0.6 0.4 0.2 0.0 Avgerage Fraction Incoming GPA Lesson #7 Students with lower incoming GPA answered about 70% of questions in class versus 85% for higher GPA students.
  • 20.
    Participation vs. GPA Participation as a Function of Incoming GPA Lesson #8 Students with lower incoming GPA took notes on 5X fewer slides. 3600 3200 2800 2400 2000 1600 1200 800 400 0 180 160 140 120 100 80 60 40 20 0 Avg. Num. Slides w/Notes Avg. Words Typed per Class <3.0 3.0-3.6 >3.6 Number of Words Typed Number of Slides w/Notes Grade Point Average
  • 21.
    Participation vs. GPA Participation as a Function of Incoming GPA 20 18 16 14 12 10 8 6 4 2 0 Avergae Days Participated by Location Participated in Classroom Participated Remotely Missed class Incoming GPA Lesson #9 Students with lower were far more likely to participate remotely than higher GPA students.
  • 22.