© 2010 IBM Corporation
Syracuse University Classroom Utilization Study Final Presentation
August 13, 2010
IBM/SU Confidential 2
Executive Summary Background
– The current class schedule is a product of incremental changes over the years.
– The room assignment and scheduling process is completed at both the university-level and at individual colleges and department.
Project Objectives
– Increase SU’s understanding of classroom usage and trends
– Enhance SU’s classroom utilization approach with consideration of scheduling and facility constraints
Approach
– Gathered and analyzed past classroom data, analyzed the current state of classroom utilization.
– Using ILOG CPLEX, developed an optimization model that determines the optimal course-to-classroom mapping.
Key Findings
– Provided SU a complete and in-depth picture of its classroom inventory and how it is used today
– Without adding new classrooms or making any changes in class times, provided an approach for SU to better assign course to classrooms to improved faculty and student experience
• For example, a 50% improvement in seat utilization and a 30% improvement in location mapping.
– “Stress test” results (squeezing in as many classes as possible) showed that SU’s current classroom inventory can support a maximum of 18% additional usable class times.
IBM/SU Confidential 3
Background The current class schedule is a product of incremental changes over the years.
– The schedule contains implicit rules and preferences, including course conflicts, technology requirements, faculty preferences.
The room assignment and scheduling process is completed at both the university-level and at individual colleges and department.
Class utilization can be Improved by
– Taking a comprehensive approach, considering all courses and rooms at the same time
– Establish rules and explicit requirements that provides clear guidance
– Pooling of captive and registrar classrooms to enable a more effective use of existing resources
Potential benefits of improved classroom utilization
– A balance of optimizing the overall university classroom utilization while meeting individual department’s needs
– Avoid unnecessary costs to construct new buildings or outfit existing rooms to meet future class needs
– An improved student and faculty experience
– Increase in flexibility to accommodate future classes and/or repair needs
IBM/SU Confidential 4
Project Objectives
Syracuse University is interested in enhancing its classroom utilization approach, including
– Increase SU’s understanding of current classroom usage and trends
– Establish a comprehensive and systematic analytical approach to assign courses to rooms, with consideration of scheduling and facility constraints
– Understand the potential of current facility level and identify areas of improvements.
IBM/SU Confidential 5
Pre-Analysis
IBM/SU Confidential 6
Pre-Analysis: Objective and Approach
To get a better understanding of the classroom inventory and how it is used today.
Gathered and analyzed past classroom schedule and inventory data. – Data Cleansing: accounted for cross-listings and incomplete classroom information – Data Source: All Registrar classrooms and the corresponding classes held during Fall
2009
Based on the Fall 2009 class schedule, the analysis: 1. Report classroom utilization per (i) days of the week and time of day, (ii) classroom
types, and (iii) department 2. Report seat fill rates per (i) days of the week and time of day, (ii) classroom types, and
(iii) department 3. Show vacancy times for each classroom 4. Identify classes that meet outside of the typical class times 5. Identify academic colleges and subjects that have a disproportion percentage of
classes that meet during prime time.
IBM/SU Confidential 7
Period Utilization: All Registrar Classrooms MWF
• Max # of Classrooms = 165 • Max # of Class Hours per Time Block is approximated = 165*80min ~ 220 hour • Note that 55-min classes are included as part of the 80-minute interval when the time periods overlap, for example, classes held on Monday from 12:45-1:40pm are included in the 12:45-2:05 slot
# of Class Hours per Time Blocks for Monday, Wedensday & Friday
0
20
40
60
80
100
120
140
160
180
200
8:00-9:20am(80min) or 8:25-9:30am (55min)
9:30-10:25am(55min)
10:35-11:30am(55min)
11:45-12:35pm(55min)
12:45-2:05pm(80min) or 12:45-1:40pm (55min)
2:15-3:35pm(80min)
3:45-5:05pm(80min)
5:15-6:35pm(80min)
# of
Cla
ss H
ours
MON WED FRI
IBM/SU Confidential 8
Period Utilization: All Registrar Classrooms TTh
Observations: Classroom utilization varies greatly across different days of the week. Descending order of utilization: Tue, Thur, Wed, Mon, Fri. Very few Friday classes in the last three blocks (12:45-2:05pm, 2:15-3:35pm, 3:45-5:05pm) Rooms are highly utilized in both AM and PM on Tuesdays & Thursdays. Rooms are more utilized in PM than AM on Mondays & Wednesdays.
# of Class Hours per Time Blocks for Tuesday & Thursday
0
20
40
60
80
100
120
140
160
180
200
8:00-9:20am(80min)
9:30-10:50am(80min)
11:00-12:30pm(80min)
12:30-1:50pm(80min)
2:00-3:20pm(80min)
3:30-4:50pm(80min)
5:00-6:20pm(80min)
# of
Cla
ss H
ours
TUE THUR
IBM/SU Confidential 9
Technology Type UtilizationNo Technology 42%V/D Projection 46%V/D Projection with inboard computer and document camera 60%Advanced Installation possible multiple projectors 56%Cluster 32%Overall Average 54%
Period Utilization: Per Different Classroom Types Monday to Friday (8am to Various End Times*)
Room Capacity Utilization1-20 42%11-35 51%35-50 58%51-75 63%76-100 53%101-200 54%>200 44%Overall Average 54%
Room Type UtilizationClass/Multi-use 27%Cluster 32%Lecture Hall/Aud 51%Seminar Room 52%Tablet Arm Classroom 55%Overall Average 54%
Note: * Various End Times: MW 6:35pm TTh 6:20pm F 5:05pm Cells are highlighted when period utilization > 60% Travel time between classes are not included in this analysis, which equals to roughly 10% additional utilization.
IBM/SU Confidential 10
Optimization Model: Fall 2009 Classroom Assignment
IBM/SU Confidential 11
Approach
Using ILOG CPLEX, IBM developed an optimization model that determines the optimal course-to-classroom mapping. The model considered all the relevant operational constraints and faculty/administrator
preferences simultaneously and automatically
OUTPUTS INPUTS CLASSROOM ASSIGNMENT MODEL
Optimal Room Assignment
Class Requirements
Faculty & Student Preferences
University Priorities
Stress Test: Determined the Maximum Additional Course Capacity
Registrar Rooms
IBM/SU Confidential 12
• Period Taught •Day of the week •Start and end times
•Enrollment Capacity •Technology Index
• Period Taught •Day of the week •Start and end times
•Enrollment Capacity •Technology Index
• Period Taught •Day of the week •Start and end times
•Enrollment Capacity •Technology Index
Graphical Illustration
•Department or Building •Period Taught
•Day of the week •Start and end times
•Enrollment Capacity •Technology Index •Fall 2009 Room Assignment
Course ID
Course ID
Course ID
Course ID
• Period Taught •Day of the week •Start and end times
•Enrollment Capacity •Technology Index
• Period Taught •Day of the week •Start and end times
•Enrollment Capacity •Technology Index
• Period Taught •Day of the week •Start and end times
•Enrollment Capacity •Technology Index
• Building ID •Classroom Capacity •Technology Index
•Period •Day of the week •Start and end times
Classroom ID
Classroom ID
Classroom ID
Classroom ID
Qty of 2358 courses Examples: •Course ID 16759, Writing 105, Section M220 •Course ID 16765, Writing 105, Section M240 •Course ID 19003, History 111, Section M009
Qty of 165 Classrooms Examples: •Bldg MC14, Rm AG202 •Bldg EO32, Rm SHAW029G
Assign
Implied Data
IBM/SU Confidential 13
Business Rules
The mathematical model assigns courses to classrooms such that
1. Each course is assigned to at most one classroom and the assigned classroom meets the course's must-have requirements, including
• Technology, seating arrangement and capacity (seat-fill rate <= 1.0)
2. Each classroom has no more than one class at any instance (no double-booking)
The #1 model objective is to ensure all courses are assigned and meet the above two requirements.
When the #1 priority is met, the model will maximize the nice-to-have features, in descending order of priorities:
1. Proximity to the department location
2. Achieving a seat-fill rate between 0.7 to 1.0 (maximum seat-fill rate = enrollment cap / room cap. Since enrollment tends to be 10% less than enrollment cap, this characteristics will result in actual seat-fill rate will around 0.6 to 0.9)
3. Keeping the same room assignment as prior history The Tie Breaker
IBM/SU Confidential 14
Course Assignment Tradeoffs
200
300
400
500
600
700
800
900
1000
1100
1200
600 700 800 900 1000 1100 1200 1300
Courses Held in Non-Department Buildings
Cour
ses
with
Util
izatio
n <
70%
D = Cost for Different Building
U = Cost for Utilization < 70%
D and U varied from 10 to 1000
Cost for Moving = 1
D/U = 2 D/U > 10
D/U = 1
U/D = 2 U/D > 10
Baseline Condition
Suggested Cost Tradeoff
IBM/SU Confidential 15
Course Assignment Results
Based on discussions with the Working Group, we adopted the following priority:
Using the plot on previous slide, we adopt the condition in which (Cost for Utilization < 70%) / (Cost for Different Building) = 2. Specifically:
– Cost of move = 1, Cost of different building = 100, Cost of not meeting seat utilization = 200, Cost of Not Assigning a Course = 10,000
This condition leads to the following improvement over the baseline case, including a 65.6% reduction in classes that do not utilize at least 70% of seating:
#3. Avoid Move #1. Meet Seat Util Target #2. Stay in Dept Bldg
IBM/SU Confidential 16
Optimization Model: Stress Test
IBM/SU Confidential 17
Stress-Test: Objective and Approach
Objective: Evaluate the degree to which SU can gain extra classroom capacity by simply re-assigning the course-to-room assignment (without changing the actual time-schedule)
Approach: Estimate the number of additional courses we can fit in the current set-up
1. Generate “dummy courses”: one for each class block per classroom • Each dummy course can only be assigned to one of the classrooms.
Hence, there are 165 x 16 = 2,640 dummy classes • This ensures that if there is any possibility of “squeezing” in a class
under the current set up, there is a dummy class to be placed in that slot.
2. Include “dummy courses” in the model, give them 2nd priority (after the real courses).
3. Based on model result, determine how many additional courses the current set-up can accommodate.
IBM/SU Confidential 18
Stress-Test: A Graphical Illustration (part 1)
8am 9am 10am 11am 12pm 1pm 2pm 3pm 4pm 5pm 6pmRoom 1 Class 1 Class 2Room 2 Class 2 Class 3Room 3 Class 4 Class 5
8am 9am 10am 11am 12pm 1pm 2pm 3pm 4pm 5pm 6pmRoom 1 Class 1 Class 2Room 2 Class 2 Class 3 New Class 1Room 3 Class 4 New Class 2 Class 5
Current Class Assignment
Improvement 1: Finding open slots without re-assigning rooms
IBM/SU Confidential 19
Stress-Test: A Graphical Illustration (part 2)
8am 9am 10am 11am 12pm 1pm 2pm 3pm 4pm 5pm 6pmRoom 1 Class 1 Class 2Room 2 Class 2 Class 3Room 3 Class 4 Class 5
8am 9am 10am 11am 12pm 1pm 2pm 3pm 4pm 5pm 6pmRoom 1 Dummy a Dummy b Dummy cRoom 2 Dummy d Dummy e Dummy fRoom 3 Dummy g Dummy h Dummy i
Current Class Assignment
Improvement 2: Finding open slots with room re-assignments
Dummy Classes
ILOG Model
+
8am 9am 10am 11am 12pm 1pm 2pm 3pm 4pm 5pm 6pmRoom 1 Class 1 Dummy b Dummy cRoom 2 Class 2 Dummy e Class 5Room 3 Class 4 Class 3 Class 2
IBM/SU Confidential 20
Stress-Test: Major Findings
Class Counts
Class Type Mon/Wed/Fri Tue/ThurCurrent 1,252 1,109Additional (dummy) 338 222Total 1,590 1,331
% Increase in Classes (Maximum possible) 21.3% 16.7%
We showed that SU’s current classroom inventory can support a maximum of 21.3% additional usable class times on MWF and 16.7% on TTh, according to the SU standard class blocks, – Current classes include all the classes that meet on any one of the week days. – “Dummy” classes are set up such that it meets during the standard class block.
IBM/SU Confidential 21
Stress-Test: # of Current and “Dummy” Classes on MWF
** Additional Classes are designed to meet during the exact class blocks in all 3 days (Mon, Wed & Fri)
Mon/Wed/Fri # of Current and Additional Class Hours
0
20
40
60
80
100
120
140
160
180
8:00-9:20am
9:30-10:25am
10:35-11:30am
11:40-12:35pm
12:45-2:05pm
2:15-3:35pm
3:45-5:05pm
5:15-6:35pm
Monday Wednesday Friday Additional*
IBM/SU Confidential 22
Stress-Test: # of Current and “Dummy” Classes on TTh
** Additional Classes are designed to meet during the exact class blocks in both Tuesday & Thursday.
Tue/Thur # of Current and Additional Class Hours
0
20
40
60
80
100
120
140
160
180
8:00-9:20am
9:30-10:50am
11:00-12:20pm
12:30-1:50pm
2:00-3:20pm
3:30-4:50pm
5:00-6:20pm
Tuesday Thursday Additional*
IBM/SU Confidential 23
Effect on “Nice-to-Have” Features
# Courses that Moved
# Courses in
Different Building
# Courses with
Utilization < 70%
Baseline
- 1209 1155
Basic Analysis
1245 1112 397
Stress Test
1271 1124 406
With the additional number of “Dummy” classes, there is a minor effect on the current class (the actual classes) assignment in terms of the “nice-to-have” features:
The addition of dummy courses had a relatively small effect on the “nice-to-have feature” for the existing courses.
IBM/SU Confidential 24
Prime-Time Analysis
IBM/SU Confidential 25
Prime-Time Analysis
School Prime Time Limit NotesUniversity of Arizona MWF 9-3 & TTh 9:30-3:30 start times <70%University of California-Berkeley 9am to 3pm <70%Brandeis University 10am to 3pm <50%
Duke University 2 to 4 <50% for <300-level classes
Exception: Dept with less than 10 <300-level courses; Labs/Discussion Session excluded
Brown UniversityIndiana University - Bloomington 9:05 am to 2:29 pm <55%University of Maryland - College Park 9 to 3 <80% <70% in MWF, <45% in TThUniversity of Minnestota - Twin Cities 9 to 3 <60%Ohio State University - Columbus 9:30 to 2:30 <15% in any single hour
University of Southern California 8 to 3Between 16% and 23% in any single hour
University of Virginia
Objectives – To understand the concentration of classes in prime time among different
colleges and a selected set of subjects. – To inform potential policy decisions on prime time scheduling requirement
Background: Prime-time definition & scheduling requirements among various universities:
IBM/SU Confidential 26
Prime Time Analysis per Academic Colleges
Colleges During Prime Time (1) Total % in Prime Time (2)
A&S 1,004 1,557 64%ARCH 13 20 65%ECS 72 125 58%
EDUC 30 97 31%HUEC 57 107 53%
IST 20 41 49%MGMT 114 200 57%UCNC 0 4 0%VPA 35 65 54%Total 1,345 2,216 61%
# of Classes
(1) Classes are considered to be during prime time if it's start time falls within one of the following:Mon/Wed 9:30am to 3:35pmTue/Thur 9:30am to 3:20pm
Fri 9:30am to 12:35pm(2) Cells are highlighted in yellow if the % of classes in Prime Time > 55% (note that the 21 out of 38 of the standard class blocks are within prime time, ~55%)
IBM/SU Confidential 27
Prime Time Analysis for Selected-Subjects
Colleges Subjects During Prime Time (1) Total % in Prime Time (2)
A&S MAT 115 172 67%A&S PPA 14 30 47%A&S WRT 90 151 60%ECS ECS 24 40 60%ECS ELE 17 25 68%
EDUC EDU 8 28 29%EDUC EED 4 10 40%HUEC NSD 12 24 50%HUEC SWK 11 19 58%MGMT ACC 21 34 62%MGMT EEE 11 25 44%
PC COM 21 48 44%PC TRF 18 33 55%
# of Classes
(1) Classes are considered to be during prime time if it's start time falls within one of the following:Mon/Wed 9:30am to 3:35pmTue/Thur 9:30am to 3:20pm
Fri 9:30am to 12:35pm(2) Cells are highlighted in yellow if the % of classes in Prime Time > 55% (note that the 21 out of 38 of the standard class blocks are within prime time, ~55%)
IBM/SU Confidential 28
Summary
IBM/SU Confidential 29
Benefits of Improved Classroom Utilization A balance of optimizing the overall university classroom utilization while meeting
individual department’s needs
Avoid unnecessary costs to construct new buildings or outfit existing rooms to meet future class needs
An improved student and faculty experience
Increase in flexibility to accommodate future classes and/or repair needs
IBM/SU Confidential 30
What We’ve Accomplished in this Study
Provided SU a complete and in-depth picture of its classroom inventory and how it is used today, for example,
– Obtained a detailed occupancy view of individual rooms – Mid-size classrooms and those equipped with technology are well utilized, but not as
fully utilized as prior expectations
The optimization model showed that, without adding new classrooms or making any changes in class times, this project provided an approach for SU to better assign course to classrooms to achieve:
– Improved faculty and student experience, e.g., 50% improvement in seat utilization and a 30% improvement in location mapping
– Increased the number of classroom availability, e.g., up to 18% increase of usable class times based on stress test results (which is intended to squeeze in as many classes as possible).
Identified the academic colleges/subjects that have a disproportion percentage of classes
that meet during prime time.
IBM/SU Confidential 31
Recommendations
IBM/SU Confidential 32
Classroom Utilization can be improved by
Taking a comprehensive approach, considering all courses and rooms at the same time
Establish rules and explicit requirements that provides clear guidance
Pooling of captive and registrar classrooms to enable a more effective use of existing resources
Consider both classroom assignment and class time scheduling
Initiated in This Study
Potential Next Steps
IBM/SU Confidential 33
Recommended Extension Options – Associated with the Current Room Assignment Model
2) Reporting Tool to Automate the
Classroom Utilization Report
1) Business User Interface
to the Optimization Model
3) Captive Classroom Study
to include captive classrooms in the pool
Room Assignment Model to optimize room assignment
COMPLETED
The following describe slides describe various extension components related to the current room assignment model; and how they relate to each other.
The inclusion and timing of each component can be adjusted according to SU’s preference.
IBM/SU Confidential 34
1) Business User Interface for the Current Room Assignment Model
Benefits: Facilitate repeatable usage of the room assignment model by designated SU business users to evaluate different policy decisions and access impact.
Summary: To simplify the usage of the optimization model by automating the data input, pre-processing and output process, such that users only need to interface with an MS Access front-end (or equivalent).
Approach: SQL will be used to conduct the data pre-processing in MS Access while VBA will be used to connect MS Access and ILOG for model input/output.
OUTPUTS INPUTS ROOM ASSIGNMENT MODEL
Optimal Room Assignment
Class Requirements
Faculty & Student Preferences
University Priorities Stress Test
Optimization Model in ILOG
RAW DATA REPORTS User Input/Output in MS Access
Pre-Processing Output Analyzer
Registrar Rooms
IBM/SU Confidential 35
2) Reporting Tool Associated with the Current Room Assignment Model
Benefits: Provide an efficient, comprehensive and consistent view of SU’s classroom inventory and usage for both current and future scenarios
Summary: Formalize and automate the generation of the classroom utilization reports, which will follow the same design and format as the ones provided in the pre-modeling step (8 reports total).
Approach: Leverage the experience gained in the “pre-analysis” phase of the recent work, develop a reporting tool in the desktop environment that streamlines the creation of the classroom utilization reports.
IBM/SU Confidential 36
3) Captive Classroom Study
Benefits: Evaluate and quantify the potential benefit of pooling the Registrar and captive classrooms in room assignment.
Summary: Extend the current classroom optimization model to include the captive rooms and the corresponding classes.
Approach: Gather the data on captive rooms and corresponding classes; incorporate it into the current classroom optimization model and determine the optimal assignment strategy. OUTPUTS INPUTS ROOM ASSIGNMENT
MODEL Optimal Room
Assignment
Class Requirements for all classes
Faculty & Student Preferences
University Priorities
Stress Test: Determined the Maximum Additional Course Capacity
Registrar Rooms
Captive Rooms
Revised New Footnote:
IBM/SU Confidential 37
Recommended Extension Options – Extending to Class Time Scheduling
4b) Business User Interface to Optimization Model
(integrated with the user interface of the room assignment model)
4a) Class Time Scheduling Model
To extend current assignment model to optimize class time
The following describe recommended extension components associated with Class Time Scheduling.
The inclusion and timing of each component can be adjusted according to SU’s preference.
IBM/SU Confidential 38
4a) Class-Time Scheduling Model
Benefits: Quantify the potential benefit of an optimal class schedule and provide a data-driven approach to revise class schedules that best meet SU’s academic mission.
Summary: Extend the current model to also optimize class time scheduling.
Approach: Construct a prototype in a similar manner to the recent classroom assignment study, during which we will focus on capturing the business rules, developing a working class-time optimization model and develop initial estimate on potential savings/benefits for various scenarios.
OUTPUTS INPUTS ROOM ASSIGNMENT & SCHEDULING
MODEL Optimal Room
Assignment
Class Requirements
Faculty & Student Preferences
University Priorities Stress Test:
Determined the Maximum Additional Course Capacity
Registrar Rooms
Optimal Class Times
Revised New Footnote:
IBM/SU Confidential 39
4b) Business User Interface for the Class Time Scheduling Model
Benefits: Facilitate repeatable usage of the class time scheduling model by designated SU business users to evaluate different policy decisions and access impact.
Summary: Build upon the business user interface for the Room Assignment model, added in the components for the class time scheduling, such that users only need to interface with a simple front-end MS Access front-end (or equivalent)
Approach: SQL will be used to conduct the data pre-processing in MS Access while VBA will be used to connect MS Access and ILOG for model input/output.
OUTPUTS INPUTS ROOM ASSIGNMENT & SCHEDULING MODEL Optimal Room
Assignment Class Requirements
Faculty & Student Preferences
University Priorities Stress Test
Optimization Model in ILOG
RAW DATA REPORTS User Input/Output in MS Access
Pre-Processing Output Analyzer
Registrar Rooms Optimal Class
Times