Reading big data from video sites

The popularity of the card house, so that the world's American fans know the Netflix, so that the world's plays can also be realized, so that the world's big data practitioners see the benchmark, "card House" success is a big data victory.

Netflix has 27 million subscribers in the United States and 33 million users around the world, and these users ' data tell Netflix:fincher (social network, seven crime director) is a reliable director, they all like Spacey this actor, the British play version of the "card House" is very popular, When this information comes together, the ratings of the card house are the future that Netflix has already foreseen.

So, this issue of CNET number reading, is to tell the Chinese video site in the large data application of exploration.

The realm of large data: Using Analysis to guide the original

Chen (Director of Chen Jie New media business)

China's dual technology is not a video site, but its clients include radio and television new media industry video site to help them upgrade in the PC terminals, mobile terminals, set-top boxes, smart TV user experience.

National dual science and Technology new media business Director Chen introduced, in the internet age, the TV station can use its own new media platform, by sending the invitation code way, so that some viewers in the PC, tablet computers, mobile phones and other terminals to watch new programs, and the user watch video behavior collection. In addition, the program group in the issuance of invitation code, can be registered to understand the audience's gender, age, occupation and other information, collected users to watch the data can reflect the different audiences are how to watch the new program.

And the user feedback is also varied.

"For example, when you see some video clips, the audience began to be impatient and began to drag on and on, and when he saw the footage, he thought it was particularly interesting, and he would look back one or two times; there were some modules for comments, such as the lighting of the scene or the costume of the host, etc. His feedback or comments can also be raised in the middle of the process. "Chen said.

The purpose of this scheme is to do some backtracking on the behavior of the users, so that the program group can see how the users look. The program team found that in a user lost link, you can consider whether to add a good link to the notice, how the notice can make the audience to leave the number reduced, or let the audience as little as possible drag, these are for the program clips provide data, and ultimately respond to customer demand.

In fact, there are already some TV customers to the country double put forward such a demand, intends to try to take the new program to the Internet to pilot, look at user feedback to adjust, they want to maximize the value of the new media platform, the new media departments have data can be feedback to the television program production department.

Of course, there are still some TV programs after the broadcast, see the Internet users to watch the feedback, thereby improving the clips, which is another reflection of user feedback. Postal

(Yao) The CTO of Youku Potato Group

This year, the Youku index has evolved into a "China Network video index", adding video data to Tudou and mobile clients.

As a platform of products, "China Network Video Index" in the Youku Potato Group reference value everywhere, from advertising sales, to copyright purchase, and then to the player product optimization, and so on, everywhere can be used as a guide.

Youku Potato Group reports its data to program makers, film and television companies, third-party analysis agencies to understand the video broadcast information, and provide the basis for the analysis of the audience; in advertising sales, can show the user behavior characteristics of advertisers, provide the value of advertising analysis; when making a copyright purchase, can help to make decisions based on the direction of the index; within the company, even the user experience of the player product optimization, you can view the data analysis results, see the button display and use frequency.

The value of these practices is obvious, or the results of data analysis can be used to guide the homemade content of the Youku potato Group.

For example, Youku has a lot of homemade content, there are a lot of micro-movies, variety shows, and so on, these playback data can show which themes are users like, where users see where to see, where is the drag and drop watch, a series of user behavior can clearly tell the content of the production staff, how to clip video, How to choose Content theme.

In fact, this process is also the process of video quality analysis, in the Youku potato search, recommended in accordance with the quality of the video ordering, in turn, improve the recommended success rate. Postal

Getting Started with big data: starting with data accumulation

Chen

Chen Jie, director of new media business

Chen stressed that the grasp of user preferences is accurate or not, the key point is the accumulation of data, specifically, the user content of the association. It is important to help customers make the label of each video complete and fine.

For example, a video is a comedy, then, this video is which country's comedy, who is the actor, who is the director, and so on this series of attribute tags to play, the more labels, in the larger data mining, the more information to dig out.

The meaning of these labels, is the data transmission, you can describe the video richer information, the more detailed the label, dug out the more accurate data, the user's preferences to grasp the more accurately.

If compared with the popular American TV series "Card House", the "big Data Mining" made by the two countries also takes the final application as the starting point, and the large data will be finally settled into the "small" data. "Small Data", not refers to the small amount of data, but the massive data mining and analysis, so that the results at a glance.

Chen explained that the data analysis, mining finally is to fall on a few points, that is, to help complete what kind of work, what kind of results, this more specific, focused work; this "small" data on the ground is the value of the data. Postal

Yao

Yao Potato Group CTO

Youku's new online home page, through a technology called "collaborative Filtering Recommendation", for different users to recommend their favorite video.

It is reported that the collaborative filtering recommendation (collaborative filtering recommendation) is rapidly becoming a popular technology in information filtering and informational system, which is different from the traditional content filtering analysis content directly, and collaborative filtering also analyzes the user's interest, To find the similar (interested) users of the specified user in the user group, to synthesize the evaluation of the information by these similar users, and to form a system to predict the degree of preference of the specified user.

Thus, seemingly simple related recommendations, in fact, in the Youku video recommendations involving hundreds of parameters, each time to adjust the parameters, you have to manually adjust more than 10 or even dozens of parameters, the daily recommended video data model to cover up to billions of.

Of course, the numbers confirm that the video recommended to users through this collaborative filtering is reliable, because the cost of finding their favorite video in a huge amount of video is high, and the recommended video opening rate is satisfying. Postal