Article Preview
Top1. Introduction
Multimedia has become the resounding concept in the current social networks and it is providing a large number of videos so the users are finding tough time to capture their related interests on the quick basis. Despite that present recommendations are always not explicit and are not well aimed with the interests of the end users. Recommendation is anticipated to be one of the vital services that can give such customized multimedia contents to users. (Wang et al., 2013) The privacy of users’ factors and video service marketers’ stores, which are of remarkable value and are very fragile maintaining these are a huge problem with the existing proposals.
Existing Collaborative filtering(CF) recommendation system calculate on content acceptance and ample user transactions histories, this system depends upon adequate history consumption report and assessment which won’t be befitting real-time recommendation. Content based filtering (CB) recommendation systems mainly target the affinities of tags, titles of contents, descriptions and rely on the user-interest things based on the individual reading history of end-users. Even though the deployment of such a recommender system is easy. Nevertheless, just using a bag of words depicting the profile information of a user is not pleasing users to acquire their exact interests. Next building a graph to take account of similarities between the recommendation items comes under graph-based recommendation systems (GB). Here the node selection problem turns into a recommendation problem, apart from that based-on user’s sentiment, behaviors and friendship forming a graph in social networks. Even merging graph theory with the recommendation is a fabulous idea but it won’t be acceptable or convincing because graphs should be changed continuously.
At present, there is no context based video recommendation system in social networks. Most of the current work allocations systems are based on equal job allocation which will not be efficient when the job durations are very high. The majority of online social network video recommendations is based on the number of times the video is watched and it is common to all users Immaterial of their age, profession, location. This type of recommendation is not convincing users all the time, people in one region might like to watch videos related to their region even though numbers of clicks are less than very popular videos. So, this paper introduces context extractor based video recommendation on cloud.
To make the task easier, we suggest a cloud based video recommendation system. The system which studies cluster behavior of the end user rather than individuality will help reduce network overload and increase recommendation speed and accuracy. To avoid network explosion, the users will be clustered into various clusters using certain clustering rules based on user context details that are collected on video-sharing online platforms. The proposed approach will have high precision, high recall and a low response delay.
Advantages:
- •
Efficiency of cloud storage
- •
No same videos are recommended
- •
Video selection strategy based on cluster ids
- •
Proper performance guarantee
- •
Reduces network overhead