Improvement of a location-aware recommender system using volunteered geographic information

• Published in: Geocarto international 2019-11, Vol.34 (13), p.1496-1513
• Main Authors: Honarparvar, Sepehr, Forouzandeh Jonaghani, Rouzbeh, Alesheikh, Ali Asghar, Atazadeh, Behnam
• Format: Article
• Language: English
• Subjects: context-aware recommender system; Location-aware recommender system; spatial data quality; volunteered geographic information
• ISSN: 1010-6049; 1752-0762
• DOI: 10.1080/10106049.2018.1493155

Recommender systems (RS), as supportive tools, filter information from a massive amount of data based on the determined preferences. Most of the RS require information about the context of users such as their locations. In such cases, location-aware recommender systems (LARS) can be employed to suggest more personalized items to the users. The most current research projects on LARS focus on the development of algorithms, evaluation methods and applications. However, the role of up-to-date spatial databases in LARS is not a well-researched area. The up-to-date spatial information would potentially improve the accuracy of items which are recommended by LARS. Volunteered geographic information (VGI) could be a low-cost source of up-to-date spatial information for LARS. This article proposes an approach to enrich spatial databases of LARS by VGI. Since not all records of VGI are fitted for use in LARS, a mechanism is developed to identify useful information. Some VGI data sets refer to existing spatial data in the database while other VGI data sets are shared for the first time. Therefore, the proposed method assessed the quality of VGI with reference source (for VGI which is existed in the database) and VGI without reference source (for VGI which is shared for the first time). To demonstrate the feasibility of the proposed approach, a mobile application has been developed to recommend suitable restaurants to the users based on their geospatial locations. The evaluation of the method indicates that VGI can potentially enhance the functionality of the LARS in predicting the users' interests.