Undergraduate Team:

Ella Fanning

Steven Chan �Savannah Davidson

Brent Lane �Regina Ratkovich

​

Graduate Team:

Hannah Smith

Beau Groom

UTK Classroom Assignment �ISE Senior Design Project

Team Members

Classroom Upgrade Committee

George Drinnon - Executive Director of Undergraduate Programs for Haslam College of Business

RJ Hinde - Vice Provost for Academic Affairs

Sarah Mobley - Department of Civil and Environmental Engineering Lecturer

Brian Coldren - Assistant Vice Provost for Academic Initiatives and University Registrar

Leonard Clemons - Associate Vice Provost for Student Success

Janet Schmitt - Assistant Vice Provost for Enrollment Management

Chuck Collins - Executive Associate Dean of the College of Arts and Sciences

Jennifer Hardy - Deputy Registrar

Problem Description

Project Charter

• Key stakeholders’ common vision
• 17 total undergraduate/ graduate team members & stakeholders

Problem Statement

• Current Method
• Not robust against last minute changes
• Does not minimize percent excess capacity
• Problems
• Causes cascading classroom assignment changes
• Caused some classes to be forced online
• Caused students to be less invested
• Less of a relationship with professor
• Causing retention rates to decrease

Current State Value Stream Map

Impact of Project

• Optimize classroom allocation process
• Maximize usage of campus resources
• Socio-economic impacts
• Reasonable implementation time for the registrar
• Reduce average professor walking distance

Approach

Objectives & Constraints

Objectives:

• Cascading classroom changes ≤ 5
• Minimize average percent of seats left empty
• Minimize walking distance of professors

Constraints:

• Cannot exceed classroom capacities (Health and Safety)
• Meet individual class requirements (Economic)
• Meet professor requests (Political)
• Short implementation time (Manufacturability)

IE Tools

• Gantt Chart
• Current State VSM

• Optimization Modeling
• Process Improvement/Lean Techniques
• Sensitivity Analysis
• Standardized Process

Spring Semester:

Fall Semester:

Approach

1.3.1 Rough Draft of Bipartite Graph

1.3.2 Rough Draft of Model

1.4 Initial Testing on Model

1.5 Testing and Refining Model

1.4.1 Create User Dashboard

1.4.3 Integrate Code into Dashboard

• Create code to read input files
• Filter and sort input data
• Feed data into Two-Phase Model and generate solutions
• Output optimal assignments in Excel

Classroom Assignment Project

Gantt Chart

Model Creation

Input Data

• University-owned rooms can be used by any department
• We can ignore department-owned rooms
• Some departments have priority in specific rooms
• Filter classroom and course data
• Reformat data
• Combining clustered/cross-listed courses
• Separate courses into standard time blocks

Introduction to Bipartite Graphs

• Consists of two sets, U and V, which are disjoint, and a set of arcs connecting the two sets
• Each node in U is connected to a node in V

​

Bipartite Graph Matching

• Matching: set of arcs that do not share any common nodes
• Maximum matching is the maximum number of arcs that do not share a common node
• Types of matchings:
• Minimum-weight matching
• Maximum-weight matching
• Maximum-cardinality matching

Model Phase 1

• Bipartite Graph Matching using course priorities
• Nodes
• Left = courses
• Right = campus-hosted classrooms
• Edges
• Course capacity < campus-hosted classroom capacity
• Weights according to constraints added to select optimal solution

Impact of Phase I

• 823 courses to possibly assign during Phase I
• 613 courses were assigned to classrooms in which they have priority during the first phase of optimization
• Eliminates over 5 million arcs across all days and thus decreases solve time and model complexity

Model Phase 2

• Network Flow Model
• Nodes for leftover rooms in each time block
• Professors travel through the potential paths
• Weights use distance and excess capacity

8:00-8:50

9:10-10:00

10:20-11:10

Results Metrics

Assignment and Capacity Changes

• Percent Change in Classroom Assignments
• 88.02%
• Percent Change in Building Assignments
• 64.73%

​

• Average Percent Excess Capacity per Course
• 22.97% Originally
• 12.17% Optimally

Optimal Results

Average Walking Distance per Day for Spring 2022

Weights vs Distance

Weights vs Capacity

Run Time

Original Lead Time: 113 Days

Cascading Classroom Changes

• No longer a major issue due to the run time of the model
• No need to change one at a time
• New optimal in only 3 hours
• 10% excess space reduces the frequency
• No need for Monte Carlo

Licensing

• Free academic Gurobi license for model creation
• Each license for internal use is $9,000/year
• Each license if commercialized is $13,200/year
• Submitted information to UTRF

Dashboard View

Designed by: Regina

Proof We Deserve to Graduate

Constraints

• Cannot exceed classroom capacities (Health and Safety)
• Meet individual class requirements (Economic)
• Meet professor requests (Political)
• Short implementation time (Manufacturability)

IE Tools

• Gantt Chart
• Current State VSM
• Optimization Modeling
• Process Improvement/Lean Techniques
• Sensitivity Analysis
• Standardized Process

​

Questions?

Example (professor schedule)

Office in COM. Originally in COM building all day, now has to travel 2.4 miles every MWF

Office in COM. Only has one class. Used to only travel 0.6 miles once a week, now travels 2.4 miles a week.

There are other professors with optimal walking distances around 2.4 miles, majority of whom had to travel farther in previous years.

331 increase. 527 decrease. 559 don’t change

Example of Improvement

Office in MCT. Original walking distance is 3.6mi per day, now 1mi per day

Office in COM. Original walking distance is 2.4mi a day. Now 0mi per day

Office in MCT. Originally walk 2.4mi a day. Now 0mi a day