Smart Web Recommender systems using parallel processing SPARK Hadoop platform

Due to the increasing volume of data there found a necessity to a fast and pertinent explore and extract of information more than before, hence there is a need for designing systems that are capable of fast attainment of interested information by users on the one hand, and on the other hand inclination toward proper analytical method for large volumes of data, felt as well. At the present time, considering the continued rapid expansion of the Internet use, the need for a recommender system effective for refining expanding volume of information has increased greatly. Recommender systems aim to provide a list of the user s favorite items to him, due to the increased volume of available data, tools used previously to process this amount of data is not appropriate. In this research to solve the proclaimed problems a recommender system applied that for detailed recommendations to the user, utilize the user comments and apply Spark processing engine in the context of HADOOP. In this thesis a combination of two-step procedure approached to recommend the user. In the first stage, for each item based on its ID, all users comments with regard to the use of dictionaries and dictionary-based approaches to traditional WordNet , classified into classes of positive and negative categories. In the second stage, using algorithms based on cooperation (Collaborative Filtering) and calculating the similarity between users, and active users items, the most similar item to the active user items situated in a list for suggestion.in this stage the previous suggested list with considering attained results from users in the second stage combined and items that earn negative views are omitted from final recommended list for users. The attained results indicate that the present method is more efficient in comparison with usual recommending method.

کلیدواژه ها:

Big data ، recommended systems ، Spark ، Hadoop ، exploring ideas ، Comments analysis