 |
 |
Volume 36 - Numéro spécial CRI 2021 - 2022
Numéro spécial CRI 2021 - 2022
1. Comparaison d'ensembles d'arbres à l'aide de descripteurs vectoriels basés sur les modèles de Markov cachés
Trees are among the most studied data structures and several techniques have consequently been developed for comparing two trees belonging to the same category. Until the end of year 2020, there was a serious lack of suitable metrics for comparing two weighted trees or two trees from different categories. The problem of comparing two tree sets was not also specifically addressed. These limitations have been overcome in a paper published in 2021 where a customizable metric based on hidden Markov models has been proposed for comparing two tree sets, each containing a mixture of trees belonging to various categories. Unfortunately, that metric does not allow the use of non metric-dependent classifiers which take descriptor vectors as inputs. This paper addresses this drawback by deriving a descriptor vector for each tree set using meta-information related to its corresponding models. The comparison between two tree sets is then realized by comparing their associated descriptor vectors. Classification experiments carried out on the databases FirstLast-L (FL), FirstLast-LW (FLW) and Stanford Sentiment Treebank (SSTB) respectively showed best accuracies of 99.75%, 99.75% and 87.22%. These performances are respectively 40.75% and 20.52% better than the tree Edit distance respectively for FLW and SSTB. Additional clustering experiments exhibited 54.25%, 98.75% and 75.53% of correctly clustered instances for FL, FLW and SSTB. No clustering was performed in existing work.
2. Construction d'une plateforme de diffusion d'informations par publication/abonnement pour les réseaux sociaux ad hoc mobiles hybrides sur les appareils androïdes
Mobile ad-hoc social networks (MASNs) have been the subject of several research studies over the past two decades. They allow stations located in a small geographical area to be connected without the need for a network infrastructure and offer them the possibility to communicate any time anywhere. To communicate, stations regularly broadcast their interests in the form of keywords. Stations with a high degree of similarity among their keywords can communicate with each other. However, the coverage of MASNs is limited to a small geographical area, due to the limited communication range of mobile ad-hoc networks (MANET) stations. In this paper, we present an architecture and implementation of hybrid mobile ad-hoc social networks (MASNs coupled to infrastructure networks) of Android mobile devices for information dissemination. Stations can use the infrastructure network to communicate and rely on the mobile ad-hoc network when the infrastructure is not available.Rather than communicating synchronously as this is the case in the similar works found in the literature, in our approach, the stations communicate using a publish/subscribe communication protocol, which is perfectly suited to this type of network thanks to the decoupling in time and space it provides.
3. Système de recommandation avec prise en compte de la prévision de disponibilité des catégories de produits
Recommending suitable products to users is crucial in e-commerce and streaming platforms. In some situations, a customer has a preference for a product based on the product features and the current temporal context. It is therefore wise to take these aspects into account in order to improve the quality of the recommendations. In this paper, we propose recommender systems based on the availability prediction of product categories according to the temporal context. Indeed, the classification of the Top-N recommendations proposed by the initial recommender system is updated in such a way as to favor products with categories predicted available. Furthermore, we propose an algorithm for the choice of the appropriate temporal context to consider for the availability prediction of categories. Experiments are carried out on four datasets and comparisons are made on the results of three basic recommender systems with and without integration of availability forecasts, according to the Hit-ratio, MAP and F1-score evaluation metrics. We note that in 75% of cases, to have the best performance, it is necessary to integrate the availabilities prediction of the categories. This gain can even go to more than 12% regardless of the dataset. All this confirms the relevance of our contribution.