| Titre : | Strong Tail Recommendation System | | Type de document : | projet fin études | | Auteurs : | Ilias Azizi, Auteur | | Langues : | Français (fre) | | Catégories : | BIG DATA
| | Index. décimale : | mast 238/19 | | Résumé : | With the advent of the web and technological developments, among others, the mass
of data to be exploited or analyzed has become very large. As a result, it has become
difficult to know what data to look for and where to find them. Computer techniques
have been developed to ease this research as well as the extraction of relevant information.
The one we focus on in this work is the recommendation system. This is to
guide the user in their exploration of the data so that it finds relevant information.
We propose a novel recommendation system, based on the proposed Strong tail similarity,
Graph embedding, community detection. In our work, using user-item’s bipartite
graph and strong tail similarity, we first construct the strong tail item graph, we apply
Node2vec algorithm for graph embedding to represent items in a continuous space Rd,
and based on the strong tail item graph, we split items into communities using Louvain
community detection algorithm .We construct the embedding space for users and
communities from items embedding space, and we aggregate the latent vector of each
entity with the corresponding side information to enhance the representation of users
and items in the embedded space. we train our Neural Network model to predict rating
using users and items latent vectors with respect to the rating user-item’s matrix,
the model is used in the framework online process to help in generating relevant rank
recommendation list of items. Focusing on the case of movies recommendation, extensive
experiments on a real-world dataset demonstrate the effectiveness of the proposed
framework. |
Strong Tail Recommendation System [projet fin études] / Ilias Azizi, Auteur . - [s.d.]. Langues : Français ( fre) | Catégories : | BIG DATA
| | Index. décimale : | mast 238/19 | | Résumé : | With the advent of the web and technological developments, among others, the mass
of data to be exploited or analyzed has become very large. As a result, it has become
difficult to know what data to look for and where to find them. Computer techniques
have been developed to ease this research as well as the extraction of relevant information.
The one we focus on in this work is the recommendation system. This is to
guide the user in their exploration of the data so that it finds relevant information.
We propose a novel recommendation system, based on the proposed Strong tail similarity,
Graph embedding, community detection. In our work, using user-item’s bipartite
graph and strong tail similarity, we first construct the strong tail item graph, we apply
Node2vec algorithm for graph embedding to represent items in a continuous space Rd,
and based on the strong tail item graph, we split items into communities using Louvain
community detection algorithm .We construct the embedding space for users and
communities from items embedding space, and we aggregate the latent vector of each
entity with the corresponding side information to enhance the representation of users
and items in the embedded space. we train our Neural Network model to predict rating
using users and items latent vectors with respect to the rating user-item’s matrix,
the model is used in the framework online process to help in generating relevant rank
recommendation list of items. Focusing on the case of movies recommendation, extensive
experiments on a real-world dataset demonstrate the effectiveness of the proposed
framework. |
|
Faire une suggestion Affiner la recherche
Strong Tail Recommendation System / Ilias Azizi
