User tweets based genre prediction and movie recommendation using LSI and SVD

The emerging popularity and raise in users' posts on social media gave birth to numerous research challenges. Out of all challenges users' centric context information based recommendation is one prime research area to recommend jobs, events and movies. Here in this research work we focus on movie context aware recommendation and for this purpose, we analyze users' posted movie tweets to understand their intentions for the same. Therefore, the objective of this research work is to predict genre of movies based on user's posted movie tweets and recommending movies to users' according to predicted genre. For this purpose, we pre-processed twitter extracted movie tweets using tokenization, porter stemming, stop word removal and use Word-Net dictionary for synonym matching. Further, we apply Latent Semantic Indexing technique which in turn involves Singular Value Decomposition on this pre-processed data and predicts genre on the basis of IMDb movie genre categorization. The predicted genre conveys the movie interest of the user and to recommend movie on the basis of predicted genre which is measured through euclidean distance. We have extracted IMdb given movie data and further predicted genre using our proposed technique. To validate this we divided our dataset using pareto principle and matched with IMDb given genre data set and achieved approximate 70% accuracy using our approach.