Current location - Plastic Surgery and Aesthetics Network - Clothing company - Seven key steps for enterprises to successfully deploy big data.
Seven key steps for enterprises to successfully deploy big data.
Seven key steps for enterprises to successfully deploy big data.

Today, big data has become a delicious "big cake" in the market, and most companies want a piece of it. Most enterprises are ready to lay out big data. So, what can be done to successfully deploy big data?

Recently, Zhou Tao, a professor at the University of Electronic Science and Technology of China and a partner in the Big Data Lab of Cloud Base, said in an interview that there are seven key steps for ordinary enterprises to become big data enterprises through cultivation:

1. What kind of data do enterprises need to save in order to make a good plan for this, people-centered data or product-centered data, or pay more attention to enterprise operation? It needs to make such a plan and then save the data in the production and operation of the enterprise. Even data that seems useless now may be of great value in the future. For example, sales offices, experience stores, etc., need to completely record the customer's visit data. Including how to come over, whether one person or several people came, whether there were old people or children, what kind of clothes to wear, and so on. , as well as the customer's mood, what they saw, what questions they asked and what they finally bought are all very important data.

In addition, all aspects of human resources within the enterprise can also be recorded, and can be excavated and analyzed. For example, he said that Changhong Company has set up many sensors on its own production line to monitor factors such as temperature, humidity, vibration, noise and particles, hoping to know which factors will have a significant impact on employees in the production process. They all think that temperature and particles may have the greatest influence on employees' operation and product quality, but in fact, as a result of final data analysis, temperature has no influence. The contribution of constant temperature control to production efficiency and qualified rate is not as great as expected, but the influence of noise on employees' mood and production is very important. To become a big data enterprise, the first step must be to realize data.

2. Enterprises should cultivate some big data concepts or small data mining teams. Big data, enterprises have different scales and different requirements. If the enterprise is large enough, such as telecom operators or industries such as power and banking, a big data team may be formed. If not, such as a simple service enterprise, then it is enough to form an idea. Now we think that a better data scientist does not mean that he is particularly good at or adapted to the network. Such people are not important. It is important to have weapons and know how to solve any problems.

The key to our understanding is to cultivate four concepts:

(1) In addition to structured data, there are text, audio, images, remote sensing, networks, behavior tracks and time data. How to deal with these data, and what is its big challenge.

(2) Be sure to understand forecasting, because most big data applications are regression forecasting, and many methods in forecasting are benchmark learning, and the hottest direction of benchmark learning is clustering learning.

(3) As far as distributed storage computing is concerned, this definitely does not mean that I know that Hadoop, Mapreduce and Hbase are enough. The key problem is to know how to make a hybrid first. When your data comes, should I sacrifice my consistency or operability? What is the approximate cost? What important data mining algorithms should I implement in Hadoop and Mapreduce? Which algorithms should pass SPTA? Variable logic governance is in hardware.

(4) Need the outward development of the whole data, and know which data may produce what kind of important value to the outside world, or what kind of important value external data can produce in your enterprise. Enterprises should cultivate these four abilities and establish a talent team for enterprise data mining.

3. Enterprises must make their own external data reserves. We all say that "books are less annoying when used". For many enterprises, such as traditional industries like clothing sales, how do I sell the goods I want to buy on Taobao and Tmall? What about Taobao or Tmall? What are the prices and sales of its competitive brands? For such data, it is often too late to find it when needed. In the same way. For example, banks lend loans to small and medium-sized enterprises and want to know its water consumption, electricity consumption, production and transportation data. For example, it can know how many cars are running in this enterprise through a camera. These data may be very important for SMEs to make loan decisions. But when you want to borrow money, there is no chance to ask again, or the cost is too high. We suggest that enterprises should learn to customize their external data and strategic data according to their business needs through public channels or data exchange.

4. Enterprises should establish their own big data management and application platform. For many enterprises, making big data does not mean building your own data center. With the emergence of cloud computing and cloud data centers, the cost of using external data centers has been very low, and the cost of data storage has also dropped exponentially. But to do big data, enterprises must have a good data office architecture on IT infrastructure and use larger tools such as data distributed storage and Hadoop. Key enterprises should not only have the hardware of a data center, but also consider the combination with the business direction of the enterprise, including not only data acquisition, database architecture, upward analysis module, upward API data export, horizontal business module and export. In order to be an enterprise's big data management application platform, we emphasize that we must start from the business of the enterprise and tailor it to suit our needs. Enterprises must first find out what their business forms are.

5. Large enterprises should have the ability of data detection, and people with innovative thinking need to think about these issues at any time, such as what great role the data owned by enterprises can play in the outside world. Just like we often take the Yachang Art Center as an example, it stores a lot of data of artworks, so it can finally publish the art index. Similarly, the State Grid has also released two indices, one is called the heavy industry electricity consumption index, and the other is called the light industry electricity consumption index. Taobao has its CPI index and some data of many enterprises, which can actually play an unimaginable value.

6. A big data enterprise, including a modern enterprise in the future, must have an open attitude. On the one hand, enterprises need to socialize many of their own problems. On the other hand, enterprises should try their best to share data with each other through some equal methods and data exchange.

7. Enterprises should also make strategic investments in data. I think there are three more advanced models.

One mode is called industrial chain layout. For example, Haier and Changhong can invest in the Internet of Things and the innovation of Internet of Things enterprises. For example, CITIC Group can pay attention to medical care and look for relevant data applications in this area.

The second aspect is technology. You need to know what hard technology innovation is, especially at the infrastructure level, such as accelerated storage, some technologies of cloud computing, such as data mining and vertical application analysis. Many innovations in this area can also be formed on a large scale.

The third mode is data set investment. We know that Alibaba invested in Gaode for data, and invested in Sina Weibo not only for investment, but also for purchasing data. The essence of all this is to move the data stream to do bigger things. This investment is to integrate data and emphasize the mobility of data. There are several points to pay attention to in these investments. First, we should pay attention to the data value of enterprises. Second, we should pay attention to early investment and long-term guidance rather than short-term pursuit of returns. Finally, we should pay more attention to traditional industries.

Professor Zhou Tao pointed out that the essence of big data is not how much data there is, nor whether it is heterogeneous data, but that the data are interrelated and the whole data can flow. He believes that cross-domain correlation and generating value far greater than two through one plus one is the essence of big data.

Of course, data itself does not generate value. Only solving problems through big data analysis is value, and the role of big data in corporate marketing can be big or small. However, in this era with the concept of big data, enterprises still have to do a good job. Prepare for the layout of big data and practice with big data companies.