+Tour was tested in Debian “bullseye” 11. Although it may work on some other Debian-based Linux distributions, we do not guarantee that all features will work well.
These instructions will guide you to get the environment up and running.
sudo apt update
sudo apt install python3 python3-pip git libboost-all-dev
pip3 install --upgrade docplex
It is also necessary to install the IBM ILOG CPLEX Optimization Studio:
-
Download the binary on the project's website;
-
Run the following command in the directory where the file is located, replacing
<CPLEX_BINARY_FILE.bin>with the correct filename:
bash <CPLEX_BINARY_FILE.bin>
- Install CPLEX at the following directory, replacing
<USER>with your username:
/home/<USER>/CPLEX_Studio221
- After the installation, configure the python API with:
sudo python3 /home/$USER/CPLEX_Studio221/python/setup.py install
- Insert the path where CPLEX is installed in the
~/.bashrcfile with:
export PATH=$PATH:/home/$USER/CPLEX_Studio221/cplex/bin/x86-64_linux
- Finally, execute:
source .bashrc
git clone https://github.com/LABORA-INF-UFG/plusTour.git
The complete data analysis is available at Analysis/, where eahc directory is named after each corresponding figure in the paper.
The data sets used in our work can be found at Datasets/, where:
- Datasets/Raw/: contains the raw data extracted from Flickr;
- Datasets/Model/: contains the data processed for the model (+Tour);
- Datasets/Analysis/: contains the data processed for the data analysis.
The datasets used for the data analysis are organized as follows:
- seqID: an identifier for the sequence;
- shortSeq: a boolean that identifies if the sequence is short (True) or not (False). A short sequence is a sequence in which the user visited less than 3 different POIs (Points of Interest);
- startTime: the date when the sequence started, i.e., when the user took the first photo;
- endTime: the date when the sequence ended, i.e., when the user took the last photo;
- seqDuration: the sequence duration in seconds;
- numberOfPhotos: the number of photos taken by the user in the sequence;
- numberOfDifferentPOIs: the number of different POIs the user visited in the sequence;
- photoID: an identifier for each photo;
- poiID: an identifier for each POI;
- poiName: the POI name;
- poiTheme: the theme of the POI;
- poiLat: the POI latitude;
- poiLon: the POI longitude;
- dateTaken: the date when the photo was taken;
- unixTimestamp: the date when the photo was taken in Unix timestamp;
- userID: an identifier for the user.
Here we detail how to run +Tour:
This phase is responsible for generating itineraries. It is written in C++, and is located at plusTour/espprc. You can compile the model with:
make -C plusTour/espprc/
Then, to run Phase 1 of +Tour, enter the directory:
cd plusTour
And execute:
./espprc/tourist -solve α type city
Where:
- α – Represents the weight given to the user’s interest or the POI popularity (α).
- type – Represents the type of user interest. The available values are:
- 1: frequency-based interest;
- 2: time-based interest;
- 3: time-based interest (weighted).
- city – Represents the city considered in the model. The available cities are:
Athens,Barcelona,Budapest,Edinburgh,Glasgow,London,Madrid,Melbourne,NewDelhi,Osaka,Perth,Toronto, andVienna. To solve for all cities, useall.
Here is an example:
./espprc/tourist -solve 0.5 2 Athens
After the execution, three files will be generated in a directory called results: plot_time_fase_1_file.in, plot_num_of_tours_file.in and tour_file.in. Where:
- plot_time_fase_1_file.in: Contains the execution time of Phase 1 in seconds;
- plot_num_of_tours_file.in: Contains the number of tours generated for each user;
- tour_file.in: Contains the specifications of the tours generated, where the first line represents the number of users, and the rest containing the tours per user in the following format:
- User ID: identity of a user;
- Number of tours: number of recommended tours generated for that user;
- Tours: A list of tours with
Number of toursitems. For each entry inTours, we have the following structure representing the tour characterization:- Sequence ID and original sequence: Identification of the original sequence and the POIs visitation in that sequence respectivelly;
- Profit: The profit perceived in that tour;
- Tour path: The sequence of POI visitation on the tour;
- Time allocation: A 480 item list representing each minute in 8 hours, where for each item, we have 0 if in that minute the user is not consuming resources when using the application and 1 otherwise;
- Number of POIs: Number of POIs available in that city;
- Time allocation per POI: The same as
Time allocationbut in this case we have a line for each POI, hence mapping the resource usage in each one of the POIs;
This phase is responsible for allocating resources, it is written in Python and it is represented by the file plusTour.py, located at plusTour/ directory.
To run Phase 2 of +Tour (assuming you are already inside the plusTour/ directory):
python3 plusTour.py --city City --instance instance
Where:
- city – Represents the city considered in the model. The available cities are:
Athens,Barcelona,Budapest,Edinburgh,Glasgow,London,Madrid,Melbourne,NewDelhi,Osaka,Perth,Toronto, andVienna. To solve for all cities, useall. - instance – Represents the network overload. The available values are:
- high overload;
- medium overload;
- low overload.
Here is an example:
python3 plusTour.py --city Athens --instance high
After the execution, a results.json file will be generated in results. In the following format:
- User – Number of sequence of the user;
- User code – Same as
User ID; - ID – Second number of sequence of the user;
- Allocation – Same as
Time allocation; - Allocation at POI – Same as
Time allocation per POI; - Profit – Same as
Profit(intour_file.in); - Path – Same as
Tour path; - Original sequence – Same as
original sequence; - APP profile – The profile of the aplication used by the user, with the following structure:
- Name – Name of the application;
- APP min bandwidth – Minimum value of bandwidth required by the application;
- APP max bandwidth – Maximum value of bandwidth required by the application;
- APP min computational – Minimum value of computational resources required by the application;
- APP max computational – Maximum value of computational resources required by the application.
- Min bandwidth – Minimum value of bandwidth required between all user applications;
- Max bandwidth – Maximum value of bandwidth required between all user applications;
- Min computational – Minimum value of computational resources required between all user applications;
- Max computational – Maximum value of computational resources required between all user applications;
- Total bandwidth alloc – Total value of bandwidth required between all user applications during all the time (8 hours);
- Total computational alloc – Total value of computational resources required between all user applications during all the time (8 hours);
- Visited – POI visitation of user for each minute;
- Visits – The POI visitation based on the tour;
- Num visits – Length of the tour;
- Bandwidth allocation – Bandwidth allocated for each POI visit;
- Computational allocation – Computational allocation for each POI visit;
- Sequence ID – Same as
Sequence ID(intour_file.in).
For your convenience, we created a script to process the results, which you can run with the following command (assuming you are already inside the plusTour/ directory):
python3 generate_json_for_plot.py --cities City --instance instance
Where:
- city – Represents the city considered in the model. The available cities are:
Athens,Barcelona,Budapest,Edinburgh,Glasgow,London,Madrid,Melbourne,NewDelhi,Osaka,Perth,Toronto, andVienna. To solve for all cities, useall. - instance – Represents the network overload. The available values are:
- high overload;
- medium overload;
- low overload.
Here is an example:
python3 generate_json_for_plot.py --cities Athens --instance high
After the execution, a first_phase_num_tours.csv file will be generated in results, containing the number of tours generated in Phase 1 per user and total (last line). A second file plusTour_<instance>.json contains the results of both Phases for each city in the following structure:
- Time – Represents the execution time for all Phases and total, with the following structure:
- Phase 1 time: Execution time of Phase 1;
- Phase 2 time: Execution time of Phase 2;
- Total time: Execution time total.
- Sequence results – Tour recommended to a user, with the following structure:
- User code: Same as
User ID; - User: Same as
User(inresults.json); - App profile: The profile of APP utilization of the user;
- Seq ID: Same as
Sequence ID; - Profit n: Normalized profit;
- Real profit: Same as
Profit; - Original: Same as
original sequence; - Generated: Same as
Tour path; - Recall: Value of the Recall metric;
- Precision: Value of the Precision metric;
- F-Score: Value of the F-Score metric;
- AE: Value of the Allocation Efficience metric;
- UE: Value of the User Experience metric.
- User code: Same as
For more information, read the +Tour Elsevier Computer Networks Paper.
Feel free to use the data sets and +Tour! Please do not forget to cite our paper! :)
@article{ESPER2025111118,
title = {{+Tour: Recommending personalized itineraries for smart tourism}},
journal = {Computer Networks},
volume = {260},
pages = {111118},
year = {2025},
issn = {1389-1286},
doi = {https://doi.org/10.1016/j.comnet.2025.111118},
url = {https://www.sciencedirect.com/science/article/pii/S1389128625000866},
author = {João Paulo Esper and Luciano de S. Fraga and Aline C. Viana and Kleber Vieira Cardoso and Sand Luz Correa},
keywords = {Travel itinerary recommendation, Next-generation touristic services, Multi-access edge computing, Advanced mobile networks}
}
If you would like to contact us to contribute to this project, ask questions or suggest improvements, feel free to e-mail us at: joaopauloesper@gmail.com and lucianosouza@discente.ufg.br.