Organizers: Tom Mller, Hana Rudov, and Zuzana Mllerov Outline - PowerPoint PPT Presentation

Organizers: Tomáš Müller, Hana Rudová, and Zuzana Müllerová

Outline Introduction 1 Competition problem and data instances 2 Presentations of finalists 3 Efstratios Rappos, Eric Thiémard, Stephan Robert 1 Jean-François Hêche, HEIG-VD, Switzerland Karim Er-rhaimini 2 Ministère de l’éducation nationale, France Alexandre Lemos, Pedro T Monteiro, Inês Lynce 3 INESC-ID / IST, Universidade de Lisboa, Portugal Dennis S. Holm, Rasmus Ø. Mikkelsen, Matias Sørensen, 4 Thomas R. Stidsen, Technical University of Denmark, Denmark Edon Gashi, Kadri Sylejmani 5 University of Prishtina, Kosovo Results of the competition 4 Now & future: benchmarks, PATAT 2021 5 Questions: chat in Zoom 6 International Timetabling Competition 2019 2

Competitions supported by PATAT Conference on the Practice and Theory of Automated Timetabling The First International Timetabling Competition (ITC 2002) The Second International Timetabling Competition (ITC 2007) The First International Nurse Rostering Competition (2010) The Third International Timetabling Competition (ITC 2011) Cross-domain Heuristic Search Challenge (CHeSC 2011) The Second International Nurse Rostering Competition (2014-2016) The Fourth International Timetabling Competition (ITC 2019) See http://patatconference.org/communityService.html International Timetabling Competition 2019 3

International Timetabling Competitions ITC 2002 events, rooms, students enrollment-based timetabling students in events cannot have any overlap ITC 2007 examination timetabling post enrolment-based course timetabling ITC 2002 extension curriculum-based course timetabling based on real-world instances from University of Udine ITC 2011 high-school timetabling real-world instances International Timetabling Competition 2019 4

ITC 2019: University course timetabling Enrollment-based timetabling students enroll in courses Course structure how to split course into events/classes students sectioning often needed Rooms with travel times and unavailabilities Events not meeting every week Distribution constraints on set of courses: 19 types DifferentTime, SameRoom, MinGap, MaxBreaks, ... Assignment of times and rooms to classes Optimization by minimizing penalties of time and room assignments, violated soft constraints, student conflicts ◦ T. Müller, H. Rudová, and Z. Müllerová, University course timetabling and International Timetabling Competition 2019. In PATAT 2018 – Proceedings of the 12th International Conference on the Practice and Theory of Automated Timetabling (PATAT 2018), pages 5–31, 2018. International Timetabling Competition 2019 5

Organization of the competition August 30, 2018: Announcement, sample data instances November 15, 2018: Early instances published February 1, 2019: 1st milestone June 1, 2019: 2nd milestone September 18, 2019: Middle instances published November 8, 2019: Late instances published November 18, 2019: Deadline January 1, 2020: Open source solvers publication September 2, 2020: Award ceremony August 2021: PATAT 2021 special track International Timetabling Competition 2019 6

Real-world data instances Data from the UniTime educational scheduling system Diverse characteristics 10 institutions from 5 continents Purdue University Masaryk University AGH University of Science and Technology Lahore University of Management Sciences ˙ Istanbul Kültür University Bethlehem University Universidad Yachay Tech Turkish-German University University of Nairobi Maryville University International Timetabling Competition 2019 7

Differences among data instances Size of the problem one school or faculty: 500 classes, 2,000 students, 50 rooms (part of) university: 8,800 classes, 38,000 students, 770 rooms Room utilization may happen: problem complexity difficult – not size Student course demands diverse pre-enrollments created based on curricula students by distribution constraints (e.g., SameAttendees) Course structure lecture, seminar, laboratory, ... student sectioning complex, e.g., introductory Biology for most freshmen simple: course = lecture International Timetabling Competition 2019 8

Differences among data instances Times Europe: class once a week USA: class several times a week at the same time and room even/odd weeks, some weeks distance learning, e.g., irregular classes on Fridays Travel times one building vs. campus Distribution constraints use of constraints for instructors (e.g., WorkDay, MaxBlock) International Timetabling Competition 2019 9

UniTime → ITC 2019 Data taken from the course timetabling module of UniTime anonymized and simplified format & model goal: remove features that do not make the problem easier to solve just easier to model Examples: removed/simplified features rooms a class may have multiple rooms, classes may share a room more complex room sharing distribution constraints: some removed or changed students reservations keeping some students together student conflicts may be prohibited instructors using same attendees in the competition instructors may require shorter travel times optimization: additional penalizations International Timetabling Competition 2019 10

Statistics about data instances Basics about courses, classes, rooms, and students Statistics about distribution constraints Statistics about domains Statistics about date and time patterns Statistics about utilization See https://www.itc2019.org/early-instances ... International Timetabling Competition 2019 11

Basic information about Early instances Basic information Students Instance Size Courses Classes fixed Rooms St. St.courses St.classes agh-fis-spr17 14.55 340 1,239 543 80 1,641 8.17 16.20 agh-ggis-spr17 5.82 272 1,852 332 44 2,116 6.98 29.92 bet-fal17 3.88 353 983 79 62 3,018 6.24 9.08 iku-fal17 12.60 1,206 2,641 530 214 0 – – mary-spr17 2.94 544 882 63 90 3,666 2.88 2.90 muni-fi-spr16 1.41 228 575 128 35 1,543 6.24 10.06 muni-fsps-spr17 1.48 226 561 191 44 865 7.76 11.60 muni-pdf-spr16c 15.92 1,089 2,526 1,132 70 2,938 8.72 17.35 pu-llr-spr17 4.69 687 1,001 318 75 27,018 3.02 3.40 tg-fal17 1.94 36 711 74 15 0 – – International Timetabling Competition 2019 12

Date and time patterns, utilization Date & time patterns Utilization per semester Instance Weeks Minutes Days Weeks Minutes Minutes Minutes per mtg. per class per class per class per room per student agh-fis-spr17 16 104.46 1.01 10.93 1,135.0 9,352.1 17,441.4 agh-ggis-spr17 16 124.46 1.00 4.18 450.2 16,111.4 12,428.7 bet-fal17 16 93.38 1.38 15.81 1,840.3 21,724.3 15,936.5 iku-fal17 14 123.70 1.00 14.00 1,734.4 19,060.9 – mary-spr17 16 141.88 1.51 13.83 2,606.1 20,069.3 7,252.5 muni-fi-spr16 15 121.45 1.00 10.89 1,188.3 17,295.7 12,593.5 muni-fsps-spr17 19 90.97 1.00 9.30 731.6 6,008.9 7,728.2 muni-pdf-spr16c 13 140.50 1.00 5.90 527.0 16,985.6 5,194.6 pu-llr-spr17 16 63.52 1.90 14.33 1,659.0 21,666.7 5,906.6 tg-fal17 14 132.53 1.00 1.16 159.8 7,308.4 – Minutes per class = |Weeks| * |Days a Week| * |Minutes per mtg.| International Timetabling Competition 2019 13

Current status November 18, 2019 5 teams participated in the final competition 7 (15) teams uploaded one or more solutions (including sample instances) 23 users successfully validated one or more solutions Now 13 (20) teams uploaded one or more solutions (including sample instances) 29 users successfully validated one or more solutions International Timetabling Competition 2019 14

Current status 263 registered users from 57 countries International Timetabling Competition 2019 15

Acknowledgements Conference on the Practice and Theory of Automated Timetabling (PATAT) ORTEC: optimization software and analytics solutions Apereo Foundation: supporting open-source software for higher education EURO working group on Automated Timetabling (EWG PATAT) UniTime educational scheduling system Faculty of Informatics, Masaryk University International Timetabling Competition 2019 16

Presentations of the finalists Efstratios Rappos, Eric Thiémard, 1 Stephan Robert, Jean-François Hêche HEIG-VD, Switzerland Karim Er-rhaimini 2 Ministère de l’éducation nationale, France Alexandre Lemos, Pedro T Monteiro, Inês Lynce 3 INESC-ID / IST, Universidade de Lisboa, Portugal Dennis S. Holm, Rasmus Ø. Mikkelsen, 4 Matias Sørensen, Thomas R. Stidsen MaCom / Technical University of Denmark, Denmark Edon Gashi, Kadri Sylejmani 5 University of Prishtina, Kosovo International Timetabling Competition 2019 17

Ordering of competitiors Ranking based on the computation of points in the F1 championship Points awarded per instance Position Early Middle Late 1. 10 15 25 2. 7 11 18 3. 5 8 15 4. 3 6 12 5. 2 4 10 6. 1 3 8 7. – 2 6 8. – 1 4 9. – – 2 10. – – 1 International Timetabling Competition 2019 18

1st milestone: Early instances Early instances published: November 15, 2018 1st milestone: February 1, 2019 Prices by PATAT: 300, 200, 100 EUR Winners 1st place (94 points) Edon Gashi, Kadri Sylejmani University of Prishtina, Kosovo 2nd place (69 points) Karim Er-rhaimini Ministère de l’éducation nationale, France 3rd place (32 points) Marlúcio A. Pires, Haroldo G. Santos, Túlio Ângelo M. Toffolo Universidade Federal de Ouro Preto, Brasil International Timetabling Competition 2019 19