Domestic operators seek the fulcrum of large data mining

The domestic large data development direction, the operator receives the very strong attention. Recently, according to the industry to reporters, the operators in the past year not only the mouth of the "big data" heated discussion, the actual action is very fast. So, at these three levels, what is the location of the operator's goal?

When the IT Times reporter walks into the office of the head of data operations at an operator's provincial company, his desk is in the Big Data Age book. Talking about the goal, he opened the book to the reporter to show a lot of careful annotation: "Only to reach the second, third level, that is, master the idea of large data management mode, to meet the ultimate demand for the development of large data carriers." ”

High data acquisition costs 1 days 10,000 TB

Theoretically, the operator has a huge and complete user data, not a company comparable to the Internet.

And in the age of data explosion, the volume of these data is expanding. Take the information of the mobile Internet log as an example, it is far more than the traditional data volume, reaching more than 1TB per day level, and growing. In fact, in addition to the Internet behavior, traffic information and other people see, operators can also through the base station signaling data to master mobile phone users location information. If all these big data accumulates together, the operator a province company the amount of information one day is highest or can reach 10,000 tb! "Sky Cloud Big Data Company's project director Wang to the IT Times reporter."

In the operator's gross data, there are many non-structural data, which need advanced data mining method to produce value. A large Internet Enterprise data Platform Department to the "It Times" reporter interpretation, a user of the site's click Behavior, may be only about 1/6 of the analysis of user preferences is meaningful, and large data very much emphasis on the combination of data. Traditional relational databases, such as IBM and Microsoft, which were used by operators in the past, are difficult to fulfill, and costly to implement, and must rely on the latest cloud computing technology.

For the specific cost of large data storage, Wang says it stores 1TB of data in a traditional way, costing 390,000-yuan intervals. With the new architecture of cloud computing, the cost can be reduced to 5000 to 15,000 yuan per TB, which is very attractive to operators. Therefore, the current operators generally to be good at data storage, analysis of the company to seek cooperation, and actively grasp the new technology.

The commercialization of its own business services is far from operational

What is the level of use of large data by operators? A large number of domestic internet companies in the data sector said that in the Internet companies like Alibaba has a seller and media personnel "data cube" and "Taobao index" and other data products, in addition to Ali financial use of large data to carry out credit evaluation, has been fired fiery. In contrast, the pace of operators is relatively slow. Big Data What part of the mine is dug first and which parts are not suitable for digging? How to use after digging, how to make money? These problems have left domestic operators with headaches.

A data company confirmed to reporters that its cooperation with domestic operators in the current large data applications are generally still for their own business development services, focusing on traffic management and precision marketing direction. Traffic management can analyze the flow of mass users, in order to design a more targeted flow package. Precision marketing by capturing the user's online browsing information, and so on, in the future can put accurate ads to users. "But in these two directions, there are no successful cases known to the outside world," he said. Perhaps some of the carrier's provincial companies have made some attempts based on existing data. said the person. In this respect, the operator's supervisor confirmed to reporters that at present it is still in the early stage of accumulation: "As for the data to the third party to share the business model, in fact, all parties would like to do, there have been some experimental nature of discussion, but the privacy data companies are actually very much hope that operators to share as many valuable data, and has produced a fine business development plan and its discussion. But in the end, the two sides stopped in the same doubt: "If the cost of tens of millions of or even billions of dollars to build such a platform, once used by the user complaints, by policy restrictions, how the money back?" ”

There's still a shortage of technology.

The mature application case is not many, another factor is to capture the user behavior information more accurately, still exist the technical bottleneck. It is understood that operators can now do is to understand when and where users browse which sites, which channel, but can not grasp the exact content of the page. "For example, you can see a user browsing the car channel of a portal, but don't know which car to look at specifically." Many operators have asked us whether it is possible to realize what brand of cars the user is looking at, and even what color appearance and displacement they prefer. This technology is quite complex, radical internet companies are willing to try, but the operators have a high demand for the stability of the system platform, and some costly and risky programs are eventually shelved. ”

Based on this, the personage thinks individual operator discloses big data practice case, have been exaggerated by the outside world. For example, there is news that a carrier province companies borrow large data mining, accurate monitoring of user use of micro-letter to the impact of SMS, the current technology is not likely to do: "Only a cursory view of the user in the micro-letter on how much traffic, but whether it is to send communications information, or just in the circle of friends to update content, etc., can not accurately grasp. ”

Talent accumulation can not only heavy technology

However, the relevant partners believe that operators ' movements in big data are swift: "The Big Data concept is very hot in the last year, but compared to industries such as energy and power, operators have been doing a lot in this short time, and they have a very high demand for big data technologies." I can say that the three major operators of various provincial companies, and even some subordinate agencies have been with the domestic data companies have contact. ”

In the interview, the operator's supervisor showed the reporter an analysis of the large data tools, including a detailed analysis of the characteristics of Hadoop (the software framework for distributed processing of large data).

Partners related to confirm that the operator's technology accumulation is very rapid: "If through self-study technology, and so on, training a large data technology personnel roughly need six months or longer, but operators through the external exchange, training, two months can well grasp the relevant technology." It should be said that in the past year, operators on the large data cutting-edge technology has been very thorough grasp. ”

But the operator's boss admits currently only in the "own data" and "Storage data mining technology" phase, as for the next level "based on large data application of thinking" still need to open up: "The domestic large data industry mature, is expected to be 3-5 years, during which operators seek more reasonable business model, Is the main appeal of the next stage. ”

In this respect, industry insiders pointed out that operators currently have, more data warehousing engineers and analysts, such as technical personnel, the rest of the chain of talent is still in the Internet enterprises: "In fact, in addition to technical personnel, the most critical or thinking and mechanism of the need for innovation, training of large data business strategy to understand the staff." ”

Operators mining Large data advantages

1. Master a more comprehensive data source, covering the overall day-to-day behavior of users

2. More attention to large data than other industries, rapid response

3. Has built the data center, the basic resources, the network quality and so on has the superiority

4. High frequency of foreign cooperation and rapid progress in mastering new technologies

Operators mining Large data bottleneck

1. Many data sources, complex and diverse forms, high processing costs

2. By "privacy" and other constraints, the commercialization of the model to promote the speed of not fast

3. Under the system constraints, in the management of norms, operational models are still exploring

4. At the same time, the training of technical personnel, lack of large data strategic decision-making talents