Recommendation of Activity Sequences during Distributed Events by - - PowerPoint PPT Presentation
Recommendation of Activity Sequences during Distributed Events by Diana Nurbakova Supervisors : Prof. Sylvie Calabretto, Prof. Jrme Gensel, Dr. La Laporte Examiners : Prof. Patrice Bellot (reporter) Prof. Anne Boyer (reporter) Prof.
13 December 2018 by Diana Nurbakova Supervisors:
Examiners:
2
3
4
5
Scenario 1: Overlapping events at Comic-Con w.r.t. 15 min long timeslots
Very high density of events Average number of simultaneous events = 37 Max number of simultaneous events = 112 ▼ Need for Recommender System
6
7
Scenario 2: Program of activities on board of Disney Fantasy cruise
8
Scenario 2: Program of activities on board of Disney Fantasy cruise
Desired Sequence of Activities
9
Use of a Recommender System
Scenario 2: Program of activities on board of Disney Fantasy cruise
Desired Sequence of Activities
10
Provide an integrated support for users to create a personalised itinerary
distributed events
Conceptual Direction: Practical Direction: Analyse the field of recommendation
better understanding, conceptual definition and modelling of the field Provide an integrated framework to address the defined problem
11
Conclusions & Perspectives Objectives & Research Questions RSSI : Recommendation of Sequences of Spatial Items ANASTASIA: approach for solving RSSI Fantasy_db& DEvIR: datasets for RSSI and evaluation of ANASTASIA
12
12 min 13 min 10 min 3 min 5 min
You are here! Conclusions & Perspectives Objectives & Research Questions RSSI : Recommendation
Items ANASTASIA: approach for solving RSSI Fantasy_db & DEvIR: datasets for RSSI and evaluation of ANASTASIA
12 min 13 min 10 min 3 min 5 min
13
You are here! User Items Interaction log
time
Ordered list
items
item_id user_id value ...
14
G. Adomavicius and A. Tuzhilin. Toward the next generation of recommender systems: A survey of the state-of- the-art and possible extensions. IEEE Trans. on Knowl. And Data Eng., 17(6):734–749, June 2005
User Items Interaction log
time
Ordered list
items
item_id user_id value ...
?
Users Items
15
G. Adomavicius and A. Tuzhilin. Toward the next generation of recommender systems: A survey of the state-of- the-art and possible extensions. IEEE Trans. on Knowl. And Data Eng., 17(6):734–749, June 2005
User Items Interaction log
time
Ordered list
items
item_id user_id value ...
?
Users Items
Matrix completion problem (find missing values)
16
G. Adomavicius and A. Tuzhilin. Toward the next generation of recommender systems: A survey of the state-of- the-art and possible extensions. IEEE Trans. on Knowl. And Data Eng., 17(6):734–749, June 2005
17
Activity on board of Disney Fantasy Identifier, 𝑗𝑒 0001 Name, 𝑜
Sailing Away
Location, 𝑚 Deck Stage Start Time, 𝑢𝑡 20.06.2015, 16h30 End Time, 𝑢𝑓 20.06.2015, 17h15 Duration, 𝜀 45 min Categories, 𝒟 Characters, Fun for All Ages Description, d It’s time to go Sailing Away! Join Mickey and Minnie along with Tinker Bell and the rest of the gang as they welcome you aboard the Disney Fantasy.
18
Activity on board of Disney Fantasy Identifier, 𝑗𝑒 0001 Name, 𝑜
Sailing Away
Location, 𝑚 Deck Stage Start Time, 𝑢𝑡 20.06.2015, 16h30 End Time, 𝑢𝑓 20.06.2015, 17h15 Duration, 𝜀 45 min Categories, 𝒟 Characters, Fun for All Ages Description, d It’s time to go Sailing Away! Join Mickey and Minnie along with Tinker Bell and the rest of the gang as they welcome you aboard the Disney Fantasy.
19
Activity on board of Disney Fantasy Identifier, 𝑗𝑒 0001 Name, 𝑜
Sailing Away
Location, 𝑚 Deck Stage Start Time, 𝑢𝑡 20.06.2015, 16h30 End Time, 𝑢𝑓 20.06.2015, 17h15 Duration, 𝜀 45 min Categories, 𝒟 Characters, Fun for All Ages Description, d It’s time to go Sailing Away! Join Mickey and Minnie along with Tinker Bell and the rest of the gang as they welcome you aboard the Disney Fantasy.
20
An activity 𝑏 represents an action of a certain duration that a user 𝑣 can attend or take at some geographically located point in a particular time window (e.g., conference session, concert during the festival, activity/entertainment during the holidays).
Activity, 𝑏 = (id, 𝑜, 𝑚, 𝑢𝑡, 𝑢𝑓, 𝜀, 𝑑, 𝑒)
21
Locations Single Items:
Y. Zheng. Location-Based Social Networks: Users, pages 243–276. Springer New York, NY, 2011.
22
Locations POIs Single Items:
H. Yin, X. Zhou, Y. Shao, H. Wang, and S. Sadiq. Joint modeling of user check-in behaviors for point-of-interest recommendation. In Proc. of the 24th ACM CIKM ’15, pages 1631–1640, 2015. Z. Yu, H. Xu, Z. Yang, and B. Guo. Personalized travel package with multi-point-of-interest recommendation based on crowdsourced user footprints. IEEE Trans Hum Mach Syst., 46(1):151–158, 2016. I.R. Brilhante, J.A.F. de Macêdo, F.M. Nardini, R. Perego, and C. Renso. On planning sightseeing tours with tripbuilder. Inf. Process. Manage., 51(2):1–15, 2015.
23
Locations POIs
𝑢1 𝑢2
Events Single Items:
X. Ji, Z Qiao, M. Xu, P. Zhang, C. Zhou, and L. Guo. Online event recommendation for event-based social networks. In Proceedings of the 24th WWW ’15 Companion, pages 45–46, 2015. K. Kaneiwa, M. Iwazume, and K. Fukuda. An upper ontology for event classifications and relations. In Proc. of the 20th Australian Joint Conference on Advances in Artificial Intelligence, AI’07, pages 394–403, 2007. E. Minkov, B. Charrow, J. Ledlie, S. Teller, and T. Jaakkola. Collaborative future eventrecommendation. In Proc. of the 19th ACM CIKM ’10, pages 819–828, 2010.
24
Locations POIs
𝑢1 𝑢2
Events/Activity Single Items:
25
Locations POIs
𝑢1 𝑢2
Events/Activity Single Items:
t
Trajectory Sequential Items:
Y. Zheng. Trajectory data mining: An overview. ACM Trans. Intell. Syst. Technol., 6(3):29:1–29:41, 2015.
26
Locations POIs
𝑢1 𝑢2
Events/Activity Single Items:
t
Trajectory Itinerary (activity sequence)
𝑢1 𝑢2
Sequential Items:
27
POI
[Yin 2015]
Event
[Minkov 2010, Macedo 2015]
Trip
[Brilhante 2015, Yu 2016, Zheng 2015]
Activity Sequence during Distributed Events
List (sequence) recommendation Unique Items Unique Visit Item Limited Availability Travel Time Future oriented (lack
H. Yin, X. Zhou, Y. Shao, H. Wang, and S. Sadiq. Joint modeling of user check-in behaviors for point-of-interest
E. Minkov, B. Charrow, J. Ledlie, S. Teller, and T. Jaakkola. Collaborative future eventrecommendation. In Proc. of the 19th ACM CIKM ’10, pages 819–828, 2010. A.Q. Macedo, L.B. Marinho, and R.L.T. Santos. Context-aware event recommendation in event-based social
I.R. Brilhante, J.A.F. de Macêdo, F.M. Nardini, R. Perego, and C. Renso. On planning sightseeing tours with
Z. Yu, H. Xu, Z. Yang, and B. Guo. Personalized travel package with multi-point-of-interest recommendation based on crowdsourced user footprints. IEEE Trans Hum Mach Syst., 46(1):151–158, 2016. Y. Zheng. Trajectory data mining: An overview. ACM Trans. Intell. Syst. Technol., 6(3):29:1–29:41, 2015.
User Interaction
time
item_id user_id action_type timestamp ...
28
M. Quadrana, P. Cremonesi, and D. Jannach. Sequence-aware recommender systems. ACM Comput. Surv., 51(4):66:1–66:36, July 2018.
User Spatial Items Interaction
time
Chronologically
Target time
Demographics
Emotions & Cognition
Social
TechnologyContext Constraints
item_id user_id action_type timestamp ...
29
M. Quadrana, P. Cremonesi, and D. Jannach. Sequence-aware recommender systems. ACM Comput. Surv., 51(4):66:1–66:36, July 2018.
Let 𝑉 be a set of users and 𝐵 be a set of spatial items. Let 𝑢 be the point in time for which the recommendation is sought. We denote as 𝐵𝑢 ⊂ 𝐵 a set of candidate spatial items that are available at time 𝑢. Let 𝒬(𝐵𝑢) be the powerset of 𝐵𝑢, and 𝑂 be its power. A candidate sequence 𝜊 = (𝑏 1 → ⋯ → 𝑏 𝑡 → ⋯ → 𝑏(𝑙)), where 𝑏(𝑘) ∈ 𝐵𝑢 and 1 ≤ 𝑡 ≤ 𝑙 ≤ 𝑂 is then an element of the set of all permutations of the length (𝑙) of 𝒬(𝐵𝑢), i.e. 𝜊 ∈ 𝑇𝑙(𝒬(𝐵𝑢)). We denote the latter set as Ξ = 𝑇𝑙(𝒬(𝐵𝑢)). The problem of recommendation of sequences of spatial items (RSSI) consists in finding the sequence 𝜊∗ ∈ Ξ for the target user 𝑣 ∈ 𝑉 at target time 𝑢, s.t.
𝜊 𝑣, 𝑢 = argmax𝜊∈Ξ 𝜏 𝑣, 𝜊 ,
where 𝜏(𝑣, 𝜊), 𝜏: 𝑉 × Ξ → 𝔒 is the satisfaction function that returns a satisfaction score for a user 𝑣 ∈ 𝑉 with respect to a sequence.
30
31
Activity availability Constraint
𝑢𝑡 𝑏 𝑗 ≤ 𝑡𝑢𝑏𝑠𝑢 𝑏 𝑗 ≤ 𝑢𝑓 𝑏 𝑗
32
Activity availability Constraint Activity Completion Constraint
𝑢𝑡 𝑏 𝑗 ≤ 𝑡𝑢𝑏𝑠𝑢 𝑏 𝑗 ≤ 𝑢𝑓 𝑏 𝑗 𝑡𝑢𝑏𝑠𝑢 𝑏 𝑗 + 𝜀 𝑏 𝑗 ≤ 𝑢𝑓 𝑏 𝑗
33
Activity availability Constraint Activity Completion Constraint Time Budget Constraint
𝑢𝑡 𝑏 𝑗 ≤ 𝑡𝑢𝑏𝑠𝑢 𝑏 𝑗 ≤ 𝑢𝑓 𝑏 𝑗 𝑡𝑢𝑏𝑠𝑢 𝑏 𝑗 + 𝜀 𝑏 𝑗 ≤ 𝑢𝑓 𝑏 𝑗
34
Activity availability Constraint Activity Completion Constraint Time Budget Constraint Start & Destination Constraint
𝑢𝑡 𝑏 𝑗 ≤ 𝑡𝑢𝑏𝑠𝑢 𝑏 𝑗 ≤ 𝑢𝑓 𝑏 𝑗 𝑡𝑢𝑏𝑠𝑢 𝑏 𝑗 + 𝜀 𝑏 𝑗 ≤ 𝑢𝑓 𝑏 𝑗 𝑏 0 , 𝑏 𝑡+𝑙 ∈ 𝜊
35
Two-Step Methods Sequence Learning
36
Two-Step Methods Sequence Learning
37
I.R. Brilhante, J.A.F. de Macêdo, F.M. Nardini, R. Perego, and C. Renso. On planning sightseeing tours with tripbuilder. Inf. Process. Manage., 51(2):1–15, 2015. Z. Yu, H. Xu, Z. Yang, and B. Guo. Personalized travel package with multi-point-of-interest recommendation based on crowdsourced user
Z. Friggstad, S. Gollapudi, K. Kollias, T. Sarlos, C. Swamy, and A. Tomkins. Orienteering algorithms for generating travel itineraries. In Proceedings of the 11th ACM WSDM ’18, pages 180–188, 2018. S. Rani, K.N. Kholidah, and S.N. Huda. A development of travel itinerary planning application using traveling salesman problem and k-means clustering approach. In Proceedings of the 2018 7th ICSCA 2018, pages 327–331, 2018.
Two-Step Methods Sequence Learning
Accounting for constraints; Easy understanding Computational complexity; Simplified assumption of resulting satisfaction of a sequence as sum of its parts
38
Two-Step Methods Sequence Learning
Accounting for constraints; Easy understanding Computational complexity; Simplified assumption of resulting satisfaction of a sequence as sum of its parts
𝑄
𝑘→𝑙
𝑄
𝑗→𝑘
▪ Markov Models ▪ Neural Networks
39
J. Chen, X. Li, W.K. Cheung, K. Li. Effective successive POI recommendation inferred with individual behavior and group preference. Neurocomputing, 210: 174–184, 2016. F. Figueiredo, B. Ribeiro, J.M. Almeida, and C. Faloutsos. Tribeflow: Mining & predicting user trajectories. In Proceedings of the 25th WWW ’16, 2016 J. He, X. Li, and L. Liao. Category-aware next point-of-interestrecommendationvia listwisebayesianpersonalized ranking. In Proceedings of the 26th IJCAI 2017 S.H. Hashemiand J. Kamps. Where to go next?: Exploiting behavioral user models in smart environments. In Proceedings of the 25th UMAP ’17
[Hashemi 2017] [Chen 2016, Figueiredo 2016, He 2017]
Accounting for constraints; Easy understanding Computational complexity; Simplified assumption of resulting satisfaction of a sequence as sum of its parts
Two-Step Methods Sequence Learning
Consideration of sequential nature of human behaviour; Capturing of dependencies between items No accounting for constraints
𝑄
𝑘→𝑙
𝑄
𝑗→𝑘
▪ Markov Models ▪ Neural Networks
40
Two-Step Methods Sequence Learning
𝑄
𝑘→𝑙
𝑄
𝑗→𝑘
▪ Markov Models ▪ Neural Networks
41
Create a hybrid approach by incorporating sequence learning techniques into two-step method
Conclusions & Perspectives Objectives & Research Questions RSSI : Recommendation
Items ANASTASIA: approach for solving RSSI Fantasy_db & DEvIR: datasets for RSSI and evaluation of ANASTASIA
12 min 13 min 10 min 3 min 5 min
42
You are here! 43
During Distributed Events. Procedia Computer Science 108, ICCS’17, Supplement C (2017), 2069 – 2078
A.Q. Macedo, L.B. Marinho, and R.L.T. Santos. Context-aware event recommendation in event-based social networks. In Proc. of the 9th ACM RecSys’15, pages 123–130. 2015. Sang, J., Mei, T., & Xu, C. (2015). Activity Sensor: Check-In Usage Mining for Local Recommendation. ACM Trans. Intel. Syst..6, pp. 41:1--41:24. ACM. Zhang, J.-D., & Chow, C.-Y. (2015). Spatiotemporal Sequential Influence Modeling for Location Recommendations: A Gravity-based Approach . ACM Trans. Intel. Syst. 7 (1), pp. 11:1--11:25. ACM. Vansteenwegen, P., Souffriau, W., Vanden Berghe, G., & Van Oudheusden, D. (2009). Iterated local search for the team orienteering problem with time windows. Comput. Oper. Res., 36(12),
Content-Based Approach
44
Content-Based Approach
45
Content-Based Approach
46
Categorical Influence
Main categories Other categories
1 1 1
𝑏
ℐ(𝑏 ∈ 𝐷)
Ƹ 𝑠
𝑑𝑏𝑢 𝑏, 𝑣 = 𝑣𝑑𝑏𝑢 ∙ ℐ(𝑏 ∈ 𝐷)
Weights:
𝑣𝑑𝑏𝑢 = agg𝑏∈𝐵 𝑣 1 1 + 𝛽 𝜐 𝑏 × Ԧ 𝑏𝑑𝑏𝑢
User Profile:
47
Textual Influence
𝑏
𝑢𝑔 − 𝑗𝑒𝑔
Ƹ 𝑠
𝑑𝑐 𝑏, 𝑣 = 𝛽𝑣 ∙ cos 𝑉𝑞𝑝𝑡, Ԧ
𝑏 − 𝛾𝑣 ∙ cos 𝑉𝑜𝑓, Ԧ 𝑏
𝑉𝑜𝑓 = agg𝑏∉𝐵 𝑣 1 1 + 𝛽 𝜐 𝑏 × Ԧ 𝑏𝑢𝑔𝑗𝑒𝑔 User Negative Profile: User Positive Profile: 𝑉𝑞𝑝𝑡 = agg𝑏∈𝐵 𝑣 1 1 + 𝛽 𝜐 𝑏 × Ԧ 𝑏𝑢𝑔𝑗𝑒𝑔
48
Temporal Influence
𝑏
1 1
1 × 96
𝑢𝑡 𝑏 = 16ℎ20; 𝑢𝑓 𝑏 = 16ℎ40
1 1 1
1 × 96
1 1
𝑏1
1
𝑏2
1
𝑏3
1 × 96
User Profile:
49
Categorical Influence
Ƹ 𝑠𝑑𝑏𝑢 𝑏, 𝑣 = 𝑣𝑑𝑏𝑢 ∙ ℐ(𝑏 ∈ 𝐷)
Weights:
𝑣𝑑𝑏𝑢 = agg𝑏∈𝐵 𝑣 1 1 + 𝛽 𝜐 𝑏 × Ԧ 𝑏𝑑𝑏𝑢
User Profile:
Textual Influence
𝑉𝑜𝑓 = agg𝑏∉𝐵 𝑣 1 1 + 𝛽 𝜐 𝑏 × Ԧ 𝑏𝑢𝑔𝑗𝑒𝑔
User Negative Profile: User Positive Profile:
𝑉𝑞𝑝𝑡 = agg𝑏∈𝐵 𝑣 1 1 + 𝛽 𝜐 𝑏 × Ԧ 𝑏𝑢𝑔𝑗𝑒𝑔
Ƹ 𝑠𝑑𝑐 𝑏, 𝑣 = 𝛽𝑣 ∙ cos 𝑉𝑞𝑝𝑡, Ԧ 𝑏 − 𝛾𝑣 ∙ cos 𝑉𝑜𝑓, Ԧ 𝑏
Temporal Influence
Ƹ 𝑠𝑢𝑗𝑛𝑓 𝑏, 𝑣 = ቐ 1, 𝑗𝑔 𝑢𝑏 ∩ 𝑢𝑣 ≠ ∅ 0.5, 𝑗𝑔 𝑢𝑏 ∩ 𝑢𝑣−1 ∪ 𝑢𝑣+1 ≠ ∅ 0.1, 𝑝𝑢ℎ𝑓𝑠𝑥𝑗𝑡𝑓
50
Textual Influence
User Negative Profile: User Positive Profile:
Temporal Influence
51
Categorical Influence
Weights: User Profile:
Option 1: Option 2:
𝑠
𝑑𝑏𝑢 - categorical score
𝑠
𝑢𝑓𝑦𝑢 - textual score
𝑠
𝑢𝑗𝑛𝑓 - temporal score
Ƹ 𝑠
𝑑𝑏𝑢, Ƹ
𝑠
𝑢𝑓𝑦𝑢, Ƹ
𝑠
𝑢𝑗𝑛𝑓
Hybrid Score
𝑠
ℎ𝑧𝑐 𝑏𝑗, 𝑣 = 𝛿𝑣 ⋅ Ƹ
𝑠
𝑑𝑏𝑢 𝑏𝑗, 𝑣 + 𝜀𝑣 ⋅ Ƹ
𝑠
𝑢𝑓𝑦𝑢 𝑏𝑗, 𝑣
⋅ Ƹ 𝑠
𝑢𝑗𝑛𝑓 𝑏𝑗, 𝑣
Computational Strategies
52
Strategy 1 Strategy 2
53
54
Crafts: Door Hangers → Pictionary Challenge → Singles’ Lunch → Goofy → The Comedy and Hypnosis of Ricky Kalmon
Sequence Extraction
Zhang, J.-D., & Chow, C.-Y. (2015). Spatiotemporal Sequential Influence Modeling for Location Recommendations: A Gravity-based Approach . ACM Trans. Intel. Syst. 7 (1), pp. 11:1--11:25. ACM.
55
Crafts: Door Hangers → Pictionary Challenge → Singles’ Lunch → Goofy → The Comedy and Hypnosis of Ricky Kalmon
Sequence Extraction
*Adaptation of L2TG [Zhang 2015]
Construction of Activity- Activity Transition Graph
Zhang, J.-D., & Chow, C.-Y. (2015). Spatiotemporal Sequential Influence Modeling for Location Recommendations: A Gravity-based Approach . ACM Trans. Intel. Syst. 7 (1), pp. 11:1--11:25. ACM.
56
Crafts: Door Hangers → Pictionary Challenge → Singles’ Lunch → Goofy → The Comedy and Hypnosis of Ricky Kalmon
Sequence Extraction
*Adaptation of L2TG [Zhang 2015]
Construction of Activity- Activity Transition Graph Construction of Category-Category Transition Graph
Zhang, J.-D., & Chow, C.-Y. (2015). Spatiotemporal Sequential Influence Modeling for Location Recommendations: A Gravity-based Approach . ACM Trans. Intel. Syst. 7 (1), pp. 11:1--11:25. ACM.
Subsequence of activities Subsequence of categories {Crafts: Door Hangers Pictionary Challenge Singles’ Lunch Goofy The Comedy and Hypnosis of Ricky Kalmon} {Fun for All Ages Fun for All Ages Adults Characters Fun for All Ages}
57
Crafts: Door Hangers → Pictionary Challenge → Singles’ Lunch → Goofy → The Comedy and Hypnosis of Ricky Kalmon
Sequence Extraction
*Adaptation of L2TG [Zhang 2015]
Construction of Activity- Activity Transition Graph Construction of Category-Category Transition Graph
Zhang, J.-D., & Chow, C.-Y. (2015). Spatiotemporal Sequential Influence Modeling for Location Recommendations: A Gravity-based Approach . ACM Trans. Intel. Syst. 7 (1), pp. 11:1--11:25. ACM.
Subsequence of activities Subsequence of categories {Crafts: Door Hangers Pictionary Challenge Singles’ Lunch Goofy The Comedy and Hypnosis of Ricky Kalmon} {Fun for All Ages Fun for All Ages Adults Characters Fun for All Ages}
58
Crafts: Door Hangers → Pictionary Challenge → Singles’ Lunch → Goofy → The Comedy and Hypnosis of Ricky Kalmon
Sequence Extraction
𝑄𝑈 𝐺𝑣𝑜 𝑔𝑝𝑠 𝐵𝑚𝑚 𝐵𝑓𝑡 → 𝐺𝑣𝑜 𝑔𝑝𝑠 𝐵𝑚𝑚 𝐵𝑓𝑡 = 1 2 𝑄𝑈 𝐺𝑣𝑜 𝑔𝑝𝑠 𝐵𝑚𝑚 𝐵𝑓𝑡 → 𝐵𝑒𝑣𝑚𝑢𝑡 = 1 1 = 1 𝑄𝑈 𝐵𝑒𝑣𝑚𝑢𝑡 → 𝐷ℎ𝑏𝑠𝑏𝑑𝑢𝑓𝑠𝑡 = 1 1 = 1 𝑄𝑈 𝐷ℎ𝑏𝑠𝑏𝑑𝑢𝑓𝑠𝑡 → 𝐺𝑣𝑜 𝑔𝑝𝑠 𝐵𝑚𝑚 𝐵𝑓𝑡 = 1/2
Estimation of Transition Probability
*Adaptation of L2TG [Zhang 2015]
Construction of Activity- Activity Transition Graph Construction of Category-Category Transition Graph
Zhang, J.-D., & Chow, C.-Y. (2015). Spatiotemporal Sequential Influence Modeling for Location Recommendations: A Gravity-based Approach . ACM Trans. Intel. Syst. 7 (1), pp. 11:1--11:25. ACM.
Subsequence of activities Subsequence of categories {Crafts: Door Hangers Pictionary Challenge Singles’ Lunch Goofy The Comedy and Hypnosis of Ricky Kalmon} {Fun for All Ages Fun for All Ages Adults Characters Fun for All Ages}
59
Vansteenwegen, P., Souffriau, W., Van Oudheusden, D. (2011). The orienteering problem: A survey with time windows. European Journal of Operational Research, 209(1):1-10, 2011.
Objective function Start and Destination constraint
∀𝑙 = 2, … , 𝑂 − 1
Unique visit constraint Time budget constraint
𝑡𝑢𝑏𝑠𝑢𝑗 + 𝑢𝑗𝑘 − 𝑡𝑢𝑏𝑠𝑢𝑘 ≤ 𝑁 1 − 𝑦𝑗𝑘 , ∀𝑗, 𝑘 = 1, … , 𝑂; 𝑁 = 𝑑𝑝𝑜𝑡𝑢
Timeline of the path constraint
𝑢𝑡 𝑗 ≤ 𝑡𝑢𝑏𝑠𝑢𝑗 ≤ 𝑢𝑓 𝑗 , ∀𝑗 = 1, … , 𝑂
Activity availability constraint
𝑡𝑢𝑏𝑠𝑢𝑗 + 𝜀𝑗 ≤ 𝑢𝑓 𝑗 , ∀𝑗 = 1, … , 𝑂
Activity completion constraint
𝑦𝑗𝑘 = ቊ1, if 𝑘 is visited after 𝑗 0, otherwise
Decision variable
60
i j i j ? ? ? 𝑁𝑏𝑦
𝑏∈𝜊
𝜍(𝑣, 𝑏) ⇒ ∀𝑏𝑙𝜗𝜊: 𝑁𝑏𝑦𝑆𝑏𝑢𝑗𝑝𝑙
i j k
t
i j k
Best Insertion Search Feasibility Check
✓
Insertion Current Solution ILS_TP 𝑆𝑏𝑢𝑗𝑝𝑙 = 𝜍𝑙 ∙ 𝑄𝑈 𝑏𝑙−1 → 𝑏𝑙 𝑇ℎ𝑗𝑔𝑢𝑙 ILS 𝑆𝑏𝑢𝑗𝑝𝑙 = 𝜍𝑙
2
𝑇ℎ𝑗𝑔𝑢𝑙
𝑢
𝑢𝑡 𝑏, 𝑡 𝑢𝑓 𝑗 𝜀
×
Time window Time spent at node × Arrival time
ILS ILS_TP
𝑢
Time window Time spent at node
𝑗 𝑘
× ×
Arrival time Travel time between nodes
𝑏𝑘 𝑡𝑘 𝑋𝑏𝑗𝑢𝑘
×
Waiting time
𝑢(𝑗, 𝑘)
Time spent at node with the latest possible start
𝑁𝑏𝑦𝑇ℎ𝑗𝑔𝑢𝑘 𝑚𝑏𝑢𝑓_𝑡𝑘
𝑢
Time window Time spent at node
𝑗 𝑘
× ×
Arrival time Travel time between nodes
𝑏𝑘 𝑡𝑘 𝑋𝑏𝑗𝑢𝑘
×
Waiting time
𝑢(𝑗, 𝑘)
Time interval during which the start
𝑁𝑏𝑦𝑇ℎ𝑗𝑔𝑢𝑘
Vansteenwegen, P., Souffriau, W., Vanden Berghe, G., & Van Oudheusden, D. (2009). Iterated local search for the team orienteering problem with time windows. Comput. Oper. Res., 36(12), pp. 3281-3290.
61
i j i j ? ? ? 𝑁𝑏𝑦
𝑏∈𝜊
𝜍(𝑣, 𝑏) ⇒ ∀𝑏𝑙𝜗𝜊: 𝑁𝑏𝑦𝑆𝑏𝑢𝑗𝑝𝑙
i j k
t
i j k
Best Insertion Search Feasibility Check
✓
Insertion Current Solution ILS_TP 𝑆𝑏𝑢𝑗𝑝𝑙 = 𝜍𝑙 ∙ 𝑄𝑈 𝑏𝑙−1 → 𝑏𝑙 𝑇ℎ𝑗𝑔𝑢𝑙 ILS 𝑆𝑏𝑢𝑗𝑝𝑙 = 𝜍𝑙
2
𝑇ℎ𝑗𝑔𝑢𝑙
𝑢
𝑢𝑡 𝑏, 𝑡 𝑢𝑓 𝑗 𝜀
×
Time window Time spent at node × Arrival time
ILS ILS_TP
𝑢
Time window Time spent at node
𝑗 𝑘
× ×
Arrival time Travel time between nodes
𝑏𝑘 𝑡𝑘 𝑋𝑏𝑗𝑢𝑘
×
Waiting time
𝑢(𝑗, 𝑘)
Time spent at node with the latest possible start
𝑁𝑏𝑦𝑇ℎ𝑗𝑔𝑢𝑘 𝑚𝑏𝑢𝑓_𝑡𝑘
𝑢
Time window Time spent at node
𝑗 𝑘
× ×
Arrival time Travel time between nodes
𝑏𝑘 𝑡𝑘 𝑋𝑏𝑗𝑢𝑘
×
Waiting time
𝑢(𝑗, 𝑘)
Time interval during which the start
𝑁𝑏𝑦𝑇ℎ𝑗𝑔𝑢𝑘
Vansteenwegen, P., Souffriau, W., Vanden Berghe, G., & Van Oudheusden, D. (2009). Iterated local search for the team orienteering problem with time windows. Comput. Oper. Res., 36(12), pp. 3281-3290.
Conclusions & Perspectives Objectives & Research Questions RSSI : Recommendation
Items ANASTASIA: approach for solving RSSI Fantasy_db & DEvIR: datasets for RSSI and evaluation of ANASTASIA
12 min 13 min 10 min 3 min 5 min
62
You are here! Time Windows Coordinates Service Time Categories
Attributes
ITEM
SEQUENCE
USER
Historical data
USER-ITEM
USER-USER
63
64
[1]
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [12] [13] [14] [11,12] [12] [12]
65
[1]
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [12] [13] [14] [11,12] [12] [12]
66
[1]
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [12] [13] [14] [11,12] [12] [12]
Dataset Statistics Number of users 23 Number of activities in overall program 595 Number of days 7 Number of DCL categories 10 Number of No DCL categories 42 Number of locations 47 Average time of completion 50min-1h
Link: https://goo.gl/forms/ZEX4LPhcg0qDAzlr1 67
D.Nurbakova, L.Laporte, S.Calabretto, J.Gensel. Itinerary Recommendation for Cruises: User Study. In Proceedings of the 2nd Workshop on Recommenders in Tourism co-located with 11th ACM Conference on Recommender Systems (RecSys 2017), 2017. 31–34
68
# events – 9,771 # location – 94 # categories – 465 # tags – 453 # users – 36,101 # user-user – 28,255 # user-event – 1,228,923 # years - 5
DEvIR
!
very high density
D.Nurbakova, L.Laporte, S.Calabretto, J.Gensel. DEvIR : Data Collection and Analysis for the Recommendation of Events and Itineraries. Proceedings of the 52nd Hawaii International Conference on System Sciences, HICSS-52. 2019 (to appear)
Users Activities (Items) 𝑢 TEMPORAL DATA PARTITIONING ? ? ? ? ? ? ? ? Users 𝑢
Future activities Past activities
? Train data Test data 𝜐
69
70
71
History days
Categorical score
Strategy 1 Strategy 2
Textual score
Strategy 1 Strategy 2
Temporal score
Strategy 1 Strategy 2
𝑠
ℎ𝑧𝑐 𝑏𝑗,𝑣= ൫ ൯ 𝛿𝑣 ⋅ Ƹ 𝑠𝑑𝑏𝑢 𝑏𝑗,𝑣 + 𝜀𝑣 ⋅ Ƹ 𝑠
𝑢𝑓𝑦𝑢 𝑏𝑗,𝑣⋅ Ƹ 𝑠
𝑢𝑗𝑛𝑓 𝑏𝑗,𝑣Hybrid score - Combination
Strategy 1 Strategy 2
𝑠
𝑚𝑝 𝑏𝑗,𝑣 =1 1+ 𝑓−(𝜃0+𝜃1𝐲)
Hybrid score – Logistic regression
Strategy 1 Strategy 2
History days 1 2 3 4 5 6 Average Strategy 1 6.1 3.4 6.5 6.2 9.9 11.3 7.4 Strategy 2 25.0 10.3 4.4 13.5 14.3 11.3 13.5 Improvement of ANASTASIA over ILS in term of precision, % Hybrid scores of activities are considered
72
73
74
75 Conclusions & Perspectives Objectives & Research Questions RSSI : Recommendation
Items ANASTASIA: approach for solving RSSI Fantasy_db & DEvIR: datasets for RSSI and evaluation of ANASTASIA
12 min 13 min 10 min 3 min 5 min
You are here! DEvIR Fantasy_db RSSI ANASTASIA
Dataset Background Approach
76
RSSI
Background
77
RSSI
Background Approach
How to define the problem of recommendation of sequences of spatial items (RSSI)? formalising and defining the RSSI positioning of recommendation of activity sequences during distributed events with respect to POI, Event and Trip recommendation
RQ1
78
RSSI ANASTASIA
Background Approach
79
RSSI ANASTASIA
Background Approach
What constitutes an integrated model for RSSI during distributed events? recommending personalised itineraries during distributed events bridging the users sequential behavior and a set of spatio-temporal constraints
RQ2
80
DEvIR Fantasy_db RSSI ANASTASIA
Dataset Background Approach
81
DEvIR Fantasy_db RSSI ANASTASIA
Dataset Background Method
What datasets can be used for evaluation of solutions for RSSI during distributed events? Definition of the requirements for a dataset Creation of test collections
RQ3
82
83
SHORT-TERM
ANASTASIA in application to this dataset
profiles and their selection of events at Comic-Con
Online Survey
Psychology-driven Model
Workshop on Recommender Systems for Citizens co-located with 11th ACM Conference on Recommender Systems (CitRec@RecSys). ACM, 3:1–3:4. 2017
84
SHORT-TERM
ANASTASIA in application to this dataset
profiles and their selection of events at Comic-Con
Online Survey
Psychology-driven Model
LONG-TERM
recommendation – new PhD Thesis in the team
Interactions
Workshop on Recommender Systems for Citizens co-located with 11th ACM Conference on Recommender Systems (CitRec@RecSys). ACM, 3:1–3:4. 2017
85
Diana Nurbakova, Léa Laporte, Sylvie Calabretto and Jérôme Gensel. (2018). DEvIR : Data Collection and Analysis for the Recommendation
Diana Nurbakova. (2018) Recommendation of Activity Sequences during Distributed Events. ACM UMAP 2018 – Doctoral Consortium Diana Nurbakova, Léa Laporte, Sylvie Calabretto and Jérôme Gensel. (2017) Recommendation of Short-Term Activity Sequences During
Distributed Events. In Procedia Computer Science. Proceedings of the 17th International Conference on Computational Science, ICCS '17, Zurich, Switzerland, 12-14 June 2017
Diana Nurbakova, Léa Laporte, Sylvie Calabretto and Jérôme Gensel. (2018) Recommandation de séquences d'activités pendant des
événements distribués. CORIA-TALN 2018, Rennes, 14-18 May 2018
Diana Nurbakova, Léa Laporte, Sylvie Calabretto and Jérôme Gensel. (2016) ANASTASIA : recommandation de séquences d'activités spatio-
Diana Nurbakova, Léa Laporte, Sylvie Calabretto, and Jérôme Gensel. 2017. Users Psychological Profiles for Leisure Activity
Recommendation: User Study. In Proceedings of CitRec, Como, Italy, August 27, 2017, 4 pages.
Jie Yang, Iván Cantador, Diana Nurbakova, María E. Cortés-Cediel, Alessandro Bozzon. 2017. Recommender systems for citizens: the
CitRec'17 workshop manifesto. In Proceedings of CitRec, Como, Italy, August 27, 2017.
Diana Nurbakova, Léa Laporte, Sylvie Calabretto, and Jérôme Gensel. 2017. Itinerary Recommendation for Cruises: User Study. In
Proceedings of RecTour, Como, Italy, August 27, 2017, 4 pages.
86
G. Adomavicius and A. Tuzhilin. Toward the next generation of recommender systems: A survey of the state-of-the-art and possible extensions. IEEE Trans. on Knowl. And Data Eng., 17(6):734–749, June 2005 Y. Zheng. Location-Based Social Networks: Users, pages 243–276. Springer New York, NY, 2011. H. Yin, X. Zhou, Y. Shao, H. Wang, and S. Sadiq. Joint modeling of user check-in behaviors for point-of-interest recommendation. In Proc. of the 24th ACM CIKM ’15, pages 1631–1640, 2015. E. Minkov, B. Charrow, J. Ledlie, S. Teller, and T. Jaakkola. Collaborative future eventrecommendation. In Proc. of the 19th ACM CIKM ’10, pages 819–828, 2010. A.Q. Macedo, L.B. Marinho, and R.L.T. Santos. Context-aware event recommendation in event-based social networks. In Proc. of the 9th ACM RecSys ’15, pages 123–130. 2015. H. Yin, X. Zhou, Y. Shao, H. Wang, and S. Sadiq. Joint modeling of user check-in behaviors for point-of-interest recommendation. In Proc. of the 24th ACM CIKM ’15, pages 1631–1640, 2015. Z. Yu, H. Xu, Z. Yang, and B. Guo. Personalized travel package with multi-point-of-interest recommendation based on crowdsourced user footprints. IEEE Trans Hum Mach Syst., 46(1):151– 158, 2016. I.R. Brilhante, J.A.F. de Macêdo, F.M. Nardini, R. Perego, and C. Renso. On planning sightseeing tours with tripbuilder. Inf. Process. Manage., 51(2):1–15, 2015. X. Ji, Z Qiao, M. Xu, P. Zhang, C. Zhou, and L. Guo. Online event recommendation for event-based social networks. In Proceedings of the 24th WWW ’15 Companion, pages 45–46, 2015. K. Kaneiwa, M. Iwazume, and K. Fukuda. An upper ontology for event classifications and relations. In Proc. of the 20th Australian Joint Conference on Advances in Artificial Intelligence, AI’07, pages 394–403, 2007. Y. Zheng. Trajectory data mining: An overview. ACM Trans. Intell. Syst. Technol., 6(3):29:1–29:41, 2015. M. Quadrana, P. Cremonesi, and D. Jannach. Sequence-aware recommender systems. ACM Comput. Surv., 51(4):66:1–66:36, July 2018. Z. Friggstad, S. Gollapudi, K. Kollias, T. Sarlos, C. Swamy, and A. Tomkins. Orienteering algorithms for generating travel itineraries. In Proceedings of the 11th ACM WSDM ’18, pages 180– 188, 2018. S. Rani, K.N. Kholidah, and S.N. Huda. A development of travel itinerary planning application using traveling salesman problem and k-means clustering approach. In Proceedings of the 2018 7th ICSCA 2018, pages 327–331, 2018. J. Chen, X. Li, W.K. Cheung, and K. Li. Effective successive POI recommendation inferred with individual behavior and group preference. Neurocomputing, 210: 174–184, 2016. F. Figueiredo, B. Ribeiro, J.M. Almeida, and C. Faloutsos. Tribeflow: Mining & predicting user trajectories. In Proceedings of the 25th WWW ’16, 2016 J. He, X. Li, and L. Liao. Category-aware next point-of-interestrecommendation via listwise bayesianpersonalized ranking. In Proceedings of the 26th IJCAI 2017 S.H. Hashemi and J. Kamps. Where to go next?: Exploitingbehavioral user modelsin smart environments. In Proceedings of the 25th UMAP ’17 Sang, J., Mei, T., & Xu, C. (2015). Activity Sensor: Check-In Usage Mining for Local Recommendation. ACM Trans. Intel. Syst.. 6, pp. 41:1--41:24. ACM. Zhang, J.-D., & Chow, C.-Y. (2015). Spatiotemporal Sequential Influence Modeling for Location Recommendations: A Gravity-based Approach . ACM Trans. Intel. Syst. 7 (1), pp. 11:1--11:25. ACM. Vansteenwegen, P., Souffriau, W., Vanden Berghe, G., & Van Oudheusden, D. (2009). Iterated local search for the team orienteering problem with time windows. Comput. Oper. Res., 36(12), pp. 3281-3290. Vansteenwegen, P., Souffriau, W., Van Oudheusden, D. (2011). The orienteering problem: A survey with time windows. European Journal of Operational Research, 209(1):1-10, 2011.
87
[1] TREC CS’13: [2] TREC CS’14: [3] TREC CS’15: [4] Yelp Challenge: [5] Foursquare_1: [6] Foursquare_2: [7] Flickr: [8] Twitter: [9] Meetup_1: [10] Meetup_2: [11] MCTOPMTW: [12] Other OPTW, Other OP-based: [13] TripBuilder: [14] Geolife: Adriel Dean-Hall, Charles L. A. Clarke, Jaap Kamps, Paul Thomas, Nicole Simone, and Ellen Voorhees. Overview of the trec 2013 contextual suggestion track. In Proc. of the 22nd Text REtrieval Conference (TREC 2013), 2013. Adriel Dean-Hall, Charles L. A. Clarke, Jaap Kamps, Paul Thomas, and Ellen Voorhees. Overview of the trec 2014 contextual suggestion track. In Ellen M. Voorhees and Angela Ellis, editors, Proc. of the 23rd Text REtrieval Conference (TREC 2014), 2014. Adriel Dean-Hall, Charles L. A. Clarke, Jaap Kamps, Julia Kiseleva, and Ellen Voorhees. Overview of the trec 2015 contextual suggestion track. In Proc. of the 24th Text REtrieval Conference (TREC 2015), 2015. Jacob Eisenstein, Brendan O’Connor, Noah A Smith, and Eric P Xing. A latent variable model for geographic lexical variation. In Proc. of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 1277–1287, 2010. Xingjie Liu, Qi He, Yuanyuan Tian, Wang-Chien Lee, John McPherson, and Jiawei Han. Event-based social networks: Linking the online and offline social worlds. In Proc. of the 18th ACM SIGKDD conference on Knowledge Discovery and Data Mining, KDD’12, 2012. Augusto Q. Macedo, Leandro B. Marinho, and Rodrygo L.T. Santos. Context-aware event recommendation in event-based social networks. In Proc. of the 9th ACM Conference on Recommender Systems, RecSys ’15, pages 123–130. ACM, 2015. Igo Brilhante, Jose Antonio Macedo, Franco Maria Nardini, Raffaele Perego, and Chiara Renso. Where shall we go today?: Planning touristic tours with tripbuilder. In Proc.
Wouter Souffriau, Pieter Vansteenwegen, Greet Vanden Berghe, and Dirk Van Oudheusden. The multiconstraint team orienteering problem with multiple time windows. Transportation Science, 47(1):53–63, 2013. Pieter Vansteenwegen, Wouter Souffriau, and Dirk Van Oudheusden. The orienteering problem: A survey. European Journal of Operational Research, 209(1):1 – 10, 2011. Dingqi Yang, Daqing Zhang, and Bingqing Qu. Participatory cultural mapping based on collective behavior data in locationbased social networks. ACM TIST, 7(3):30, 2016. Yu Zheng, Lizhu Zhang, Xing Xie, and Wei-Ying Ma. Mining interesting locations and travel sequences from gps trajectories. In Proc. of the 18th International Conference on World Wide Web, WWW ’09, pages 791–800, 2009. Bart Thomee, David A. Shamma, Gerald Friedland, Benjamin Elizalde, Karl Ni, Douglas Poland, Damian Borth, and Li-Jia Li. Yfcc100m: The new data in multimedia
http://www.yelp.com/dataset_challenge https://github.com/jalbertbowden/foursquareuser-dataset
88