Softwares and Datasets

Softwares

ORKAD develops software as part of its research activities. The software is a product of its research. They are made available according to the associated licences.

MH-BUILDER

MH-Builder is a C++ framework for designing adaptive metaheuristics for single and multi-objective optimization that is developed by ORKAD team. Its particularity is to be able to modify during the execution the components of your metaheuristic.

Tutorial on: https://mh-builder-orkad.univ-lille.fr/

Code source on: https://gitlab.cristal.univ-lille.fr/orkad-public/mh-builder

MOCAI

MOCA-I is a Pittsburgh classification rule-mining system using partial classification rules particularly well adapted to imbalanced data. It is based on a multi-objective local search using Confidence, Sensitivity and rule length to explore the candidate.

As MOCA-I software is embedded in the Gitlab project "mh-builder.git", the installation process is identical to that of the mh-builder framework.

Documentation is available at: https://mh-builder-orkad.univ-lille.fr/

MARY-MORSTAN

Mary-Morstan is a multi objective modular framework to automatically configure machine learning algorithms (AutoML). This python automated machine learning tool is based on evolutionary algorithms.

Mary-Morstan is modular in such a way that the exploration versus exploitation process can be tuned through the specification of an Evolutionary Algorithm (EA) space. It also allows to deal with big data files and various of classification and regression problems.

Documention is available on: https://mh-builder-orkad.univ-lille.fr/

Source code is available on: https://gitlab.cristal.univ-lille.fr/orkad-public/mary-morstan

Datasets

ORKAD uses and produces datasets as part of its research activities. These datasets allow the results obtained to be reproduced.

E-learning Recommender System Dataset (MARS)

Mandarine Academy Recommender System (MARS) Dataset is captured from real-world open MOOC {https://mooc.office365-training.com/}. The dataset offers both explicit and implicit ratings, for both French and English versions of the MOOC. Compared with classical recommendation datasets like Movielens, this is a rather small dataset due to the nature of available content (educational). However, the dataset offers insights into real-world ratings and provides testing grounds away from common datasets.

The dataset is avaible on https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/BMY3UD

Mandarine Academy Professional Timetabling (MAPT)

Mandarine Academy Professional Timetabling (MAPT) is a real-world dataset suggested to solve Professional Timetabling Problems (PTPs). A rather under-exploited category of the overall Timetabling Problems. However, we believe it can still be applied to traditional problems (education, health, etc.) as a helpful benchmark dataset to assist researchers in comparing different methods. Compared to conventional educational datasets such as (ITC2007), MAPT proposes richer features inspired by real-world data to provide insight into corporate training logistics and timetabling complexities.

MAPT dataset are avaible on https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/A4JU5E

Replication Data for: Bi-level FJSP instances arising from selective deconstruction

Dataset of instances modeling the selective deconstruction field. Each instance contains a series of deconstruction sites, recycling facilities, warehouses, and operative units along with their characteristics.

The dataset is avaible on https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/YRR9G4