Place Recommendation System Using Demographic Data Analysis Principle from Social Network

Abstract

The objective of this master project is to present rule and tool for place recommendation system using population demographic information from Facebook using Naïve Bayes classification. The demographic information is composed of 7 items; gender, age,
relationship, education, occupation status, hometown and current address. The recommended places were classified into 4 types: restaurant, hotel, outdoor visiting place, and department store and retail store. Moreover, the subcategories of each category were defined. The frequency of place visiting based on user check-in information was used to identify each user group for training data. The total frequency of check-in transaction was 35,084 items from 600 Facebook’s users. The effectiveness of the proposed rule was used accuracy metric. The accuracy values of rule for recommended places; the restaurant, hotel, outdoor visiting place, and department and retail store; were 78.7, 72.5, 67.4 and 76.7% respectively. The developed tool based on the proposed rule could be used to search the recommended places using user profiles. In addition, this tool provided the updating function to adjust the check-in information of the target sample users in order to improving the training data set. This would enhance the recommendation system in term of the up-to-date information and user interest.