SmPFT: Social media based profile fusion technique for data enrichment

Abstract People use different social networking platforms for distinct purposes. The information across each micro-blogging site is often partial. A better profile of an individual can be built, if we amalgamate the complementary information from various sites. This data enrich profile can be useful in a number of online services i.e. marketing of any product across sites, friend recommendation, etc. To integrate profile information, it is essential to identify individuals in distinct social networking platforms. This study aims to identify identical users across different social media platforms. Existing works on user profile matching frameworks are restricted to certain social networks as some of the previously available streaming APIs are not available now. In this work, there are no such dependencies over the streaming APIs as it is based on the uniqueness of usernames, which are identical among various social networking sites. We also efficiently exploit the information redundancies, due to individual similar behavioral patterns which can be used during mapping. We have tested our system over 500 users in the real-time scenario, considering only those profiles which generate their content predominately in the English language. The total dataset comprises of over 1.1 Million tweets and 0.63 Million URLs, in which 35.6% URLs contained the geotagged information. Our model is able to identify 6.3% more identical users than the traditional approaches. There are several application areas such as friends recommendation, future place prediction, leaders identification, and information diffusion across social media sites that can benefit from the outcoming of this work.