Research Of A Rule-based Personalized Location-based Recommendation System In Mobile Application

Mobile phones have taken the place of desktop computers and become themost preferred Internet terminal by the year of2012. As a result, the mobiledevices, including mobile phones and tablet PCs, turn into a main platform,where users can obtain information. The location-based service (LBS), which isa typical convenient-service for mobile devices, faces with the problem ofinformation overload as a consequence of the Internet information explosion.Mobile platform makes it easier for users to access a variety of information abouttheir surrounding businesses, not only from the traditional ways, such as blogsand websites, but also from the mobile applications which are more convenientfor users. Due to the convenience of location-based services, users may need toobtain the nearby restaurant information at any time, so the system must be ableto provide users with personalized recommendations based on the user’spreference, combined with the current location and contextual information.This paper summarized the main problems of the current LBSrecommendation systems by analyzing the particularity of the LBSrecommendation systems and mobile applications and comparing thecharacteristics and functions of different LBS recommendation systems. Then,the study put forward a new framework of a rule-based LBS recommendedsystem based on the results. The framework uses the demographic information ofusers which can be obtained during the cold start phase, integrates the richcontextual information such as users’ location, the current weather and otherenvironmental factors, which are identified by the GPS technology, and uses therule-based recommendation algorithm to reduce the effort for new users and isable to recommend under the condition of no explicit input of the userspreference. By combining contextual information, users’ preference is no longerstatic, different situations can cause users’ preference changing dramatically, as aresult it is necessary to divide users’ preference into short-term and long-term.Short-term preference refers to users’ temporary preference under the currentspecific context information; while long-term preference is the users’ stablepreference habits derived from a large number of historical data. Users’long-term preference and short-term preference are identified by different rulebase, and is unified in the final recommendation process.At last, a demonstration system was developed. Thirty test users were invited to use and test demonstration system. The specifications of the recommendedsystem, including effectiveness, customer satisfaction, accuracy, coverage,context sensitivity, were evaluated. The recommendation system put forward inthis paper was proved to be able to provide accurate and effective r ecommendedresults. Also, this recommendation system can solve the cold start problem andreduce the user cost of learning.