If you use this dataset, please cite it in your paper:

Ming Li, Lin Li, Xiaohui Tao, and Jimmy Xiangji Huang. 2024. MealRec+: A Meal Recommendation Dataset with Meal-Course Affiliation for Personal- ization and Healthiness. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR ’24), July 14–18, 2024, Washington, DC, USA. ACM, New York, NY, USA, 11 pages. https://doi.org/10.1145/3626772.3657857

Ming Li, Wuhan University of Technology, China;

Lin Li, Wuhan University of Technology, China;

Xiaohui Tao, University of Southern Queensland, Australia;

Jimmy Huang, York University, Canada

The suit of $MealRec^+$ datasets we released contains two datasets, $MealRec^+_H$ with a user-meal interaction density of 0.77% and $MealRec^+_L$ with a user-meal interaction density of 0.17%.

Each dataset contains files as follows:

• relationship information:
  • user_course.txt : the interaction information between users and courses;
    • format: [user index][\tab][course index]
  • course_category.txt : the correspondence information between courses and categories;
    • format: [course index][\tab][category index]
      • category_index = {0:appetizers, 1:main dishes, 2:desserts}
  • meal_course.txt : the affiliation information between meals and courses;
    • format: [meal index][\tab][course index]
  • user_meal.txt : the interaction information between users and meals. This file is divided into training set, verification set and test set according to the ratio of 8:1:1.
    • format: [user index][\tab][meal index]
• healthiness information
  • course_fsa.txt/course_who.txt: FSA/WHO healthiness scores of courses;
    • format: [healthiness score] (line index == course index)
  • meal_fsa.txt/meal_who.txt: FSA/WHO healthiness scores of meals;
    • format: [healthiness score] (line index == meal index)
  • user_fsa.txt/user_who.txt: Mean FSA/WHO healthiness scores of meals that each user has interacted with historically
    • format: [healthiness score] (line index == user index)
• meta information
  • course.csv: meta data of course;
    • fileds: course_id, course_name, review_nums, category, aver_rate, image_url, ingredients, cooking_directions, nutritions, reviews, tags
  • user_course.csv: meta data of user-course interaction;
    • fileds: user_id, course_id, rating, dateLastModified
  • user2index.txt: a mapping from user id to user index;
    • format: [user id][\tab][user index]
  • course2index.txt: a mapping from course id to course index;
    • format: [course id][\tab][course index]

We provide a data loader (data_loader.py) to load relationship data as matrices and healthiness data as arrays for ease of use.

We provide the evaluation code (healthiness_eval.py) for two healthiness metrics: FSA/WHO healthiness socre and ranking exposure.

Thanks to Nathan Nichols for the suggestion on data loader code.