Dual Recommendation Analysis

This thesis will investigate the recommendation systems in a radically different perspective. First, instead of focusing and proposing related products/services to customers, we will analyze and suggest recommendations for the producers. Such recommendations taking the form of product suggestions or adjustments will be defined according to a given geographical region and time span. We will also generate maps based on product profiles for a given time period and geographical region. Second, the targeted analytic system will be able to generate a description of both the different product facets and the entire product based on the customers’ reviews and thus fulfill a database describing the products available and their quality in a given time period and a geo-graphical region. Moreover, being able to detect and measure the polarity of written opinions, the system can generate a map for a given time period and a geographical region, showing the most successful products/services. We can complete this map by indicating the degree of similarity between successful (or unsuccessful) products. Once this information is obtained, the system can detect product opportunities and proposes a map of alternatives indicating where and when products that were successful in the past and in other similar regions might have a success in another region. Third, based on the social network between customers, we will determine the strength of the relationships between customers and define their degree of leadership. Based on this information, we can weight more precisely the different customers’ reports, assuming that reviews written by leaders will have a stronger impact than those written by followers. Moreover, after being able to identify the leaders, we could determine how they can improve their status or leadership.