Abstract:
Social networking website creates new ways for engaging people belonging to different communities, moral and social values to communicate and share valuable knowledge, therefore creating a large amount of data. The importance of mining social media cannot be over emphasized, due to significant information that are revealed which can be applied in different areas. In this thesis, a systematic approach for traversing the content of weblog, considering location and time (spatiotemporal) is proposed. The proposed model is capable of searching for subjects in social media using Boyer Moore Horspool (BMH) algorithm, analyzing the content in terms of location and time. BMH is an efficient string searching algorithm, where the search is done in such a way that every character in the text needs not to be checked and some characters can be skipped without missing the subject occurrence. Semantic analysis is carried out on the subject by computing the mean occurrence of the subject with the corresponding predicate and object from the total occurrence of the subject. Experiment were carried out on two datasets: the first category was crawled from twitter website from September to October 2014 and the second category was obtained from spinn3r dataset made available through the International AAAI (Association for the Advancement of Artificial Intelligence) Conference on Web and Social Media (ICWSM). The results obtained from tracking some subjects such as Islam and Obama shows that the mean occurrence of the analysis of the subject successfully reveals the pattern of the subject over a period of time for a specific location. Evaluation of the system which is based on performance and functionality reveals that the model performs better than some baseline models. The proposed model is capable of revealing spatiotemporal pattern for a subject, and can be applied in any area where spatiotemporal factor is to be considered.
