Résumé:
Over the last decade, intelligent systems that provide public services have been successfully
applied in all areas of the city, leading to a significant shift in the city’s living environment
towards the concept of the smart city. For this purpose, we focused our research on the tourism
and mobility domain to create an intelligent strategy to propose a recommendation system with
hybrid filtering sensitive to context, location and user preferences.
To build our recommender system, we must first establish a user model that allows us to
specify the features that will be included in the system. These features are offered by three
profiles : the demographic profile, the preference profile and the location profile. Following this
modeling, we developed our solution based on three factors : a key factor that allows us to
perform a sentiment analysis of the user’s opinion in the form of a textual comment and make
a prediction of this comment in the form of a score, a user preference factor that allows us to
know the most similar and appropriate restaurants with respect to these preferences, and finally
a location factor that facilitates intelligent mobility in the city in order to find the restaurants
closest in distance. The results are then merged in our system to improve the suggestions based
on these three factors.
In the experimentation phase, the sentiment analysis subsystem was developed in a Colab
environment with two datasets, allowing the creation of a sophisticated prediction model with
an accuracy rate of 60%, and the recommendation system, which is developed in the Jupyter
environment based on the previous three factors we were able to discover a usable model to
create better suggestions for the user.
