CeLaVie: A Personalized Restaurant Recommender System

Abstract

This project introduces a personalized restaurant mobile recommender system called CeLaVie. CeLaVie aims at recommending the best restaurant for users according to their personal preferences at their physical locations. To achieve these goals, we implement an agent in which resides in the users’ smart phone and uses the smart phone Global Positioning System (GPS) to determine the current position of the user. Two kinds of agents were implemented. The Context-Aware Personalized Agent (CAPA) which resides in users’ smartphones to interact with the user and the RDSA (the second kind of agents). CAPA is capable of getting user’s preferences for selecting restaurants. For the mobile user, CAPA can explore the user’s time and spatial context by the built-in sensor capabilities, and the restaurant directory service agent (RDSA). The RDSA is running on a server. Its main goal is to maintain a restaurant directory. A new restaurant must register in the restaurant directory with RDSA before it can be recommended to the user. The entry contains information about the restaurant. Also, the RDSA has the ability of searching restaurants given preference constraints. If the RDSA could not find anything according to these preference constraints, it will propose some modifications to the constraints. And it also has the ability of recommending a restaurant out of many matched ones. To recommend a restaurant for the user, CAPA first acquires the user preferences and establishes user contexts. Then RDSA searches for suitable information, and resolves constraint violations in order to do the recommendation. CeLaVie application works. Amena is a Muslim user. She is visiting Paris Main Following scenario illustrates how our proposed Station. The time is 2:00 P.M, and Amena wants to find a restaurant. She inputs her personal preferences: (1) the meal budget should be below 30 dollars; (2) the restaurant serves Halal meals; (3) non-smoking restaurant. The RDSA finds out that there is nothing matches the preference around Paris Main Station and will relax the constraints and generate some proposal according to previous interaction history. The agent finds out that there are some restaurants serve vegetarian meals and Amena used to choose vegetarian more than other categories in previous interactions. So it will add the proposal of “Change <Category> to vegetarian”. And the agent also finds out that the user used to spend 50 to 100 dollars for her lunch and if it relaxes the budget constraint to 80 and finds some restaurant matched. So it adds the proposal of “Increase <Price> to 80$”. And then, the agent presents the proposals to the user. Finally, Amena decides to choose the first proposal (change <category> to vegetarian) and she gets the final recommendation. The recommending service is not limited to restaurant information only it can be extended to others such as hotels, cinemas, bus stops, etc.