Online TV Program Recommendation System Based On User Behavior Analysis From Internet

Speedy Internet has changed our living habits in a deep way.In the past we can only use television watching some programs,but now we can use mobile phone to find any video or program via a video platform.In the past we can only use computer at specific place to surf the internet,but now we can use plenty of facility to surf the internet at any place.This transform makes it easier to live,and create massive user network behavior.Traditional data process model can’t deal with such massive data,but with the development of cloud computing,it’s possible to storage and process such massive data.All these user network behavior data seem irrelevant,but with the development of business intelligent system and data mining skills,we can get a great influence from the information which we get after the mining of these data.For a internet television platform,we can use these data mining result,combining the user network environment to push the specific videos they want.It will make user stick to this platform and increase the value of the platform.This article based in business intelligence,built a BILOG system.This system can gather user network behavior,store it in distributed database by log.This system use ETL(Extract,Transform Load)to preprocess the data,split these data into table to do query and computation.The environment of the article is Hadoop,using map-reduce to do data washing and generating 12 basic metrics,each metric is save in a csv file as a vector to cluster analysis.We set a 10 classes,use Weka as software platform and k-means algorithm to do clustering.According to our experiment we learned than every class we get has some specific features and we can use these feature to do some classification.We exact the user habits though these features,link these feature with TV programs and do some personal recommendation,in order to gain user flow.This article emphasis on BI system and data mining algorithm to make a personal recommendation,this article will not discuss anything about data gathering.The data source used in this article is Weibo data from an open-source data source called Tianchi.Weibo is a social platform which is the most activated and has plenty of actions such as like,commend,and follow.We can use these different user behavior data to classify the user.Finally,we test our system use the specific Weibo data.According to our experiment,we found that our classification algorithm is accurate and we can use the result of classification to recommend internet TV show.