in the industrial Internet era, the industrial Internet era of enterprise material assets are gradually replaced by big data assets, big data is the production of industrial Internet era. It depends on the data of the Internet industry is not only the engine, data will be whether the Internet standard, whether the formation of closed loop data is the key, it is not a simple software and information, IT, but the enterprise’s overall ecosystem data. Today, the electricity supplier 3 era has quietly come (2012 to date), is the era of brand, market segmentation era, but also data operation era, but also a lot of enterprise data began. For electricity providers, the data is life. Then, in the electricity supplier 3 era, how to do data management? How can we deal with the data properly?
first method: control
contrast, commonly known as "contrast", "look at a data alone" will not feel, and must be compared with another data will feel.
this is the most basic thinking, but also the most important thinking. In the reality application is very wide, such as the selection and measurement models monitor store data, the process is in control, analysis personnel to get the data, if the data is independent, can not compare, it is unable to judge, not equal to the data read from the useful information.
second method: split
The word "
analysis" literally means splitting and parsing. Thus, the importance of splitting in data analysis is visible. In the "generation" can also be seen everywhere "split" the word, many writers will use this tone: after the split, we will clear…….
we go back to the first mental contrast, and when a dimension can be compared, we choose to contrast. After a comparison, it is time to find out why the problem needs to be found out Or no comparison at all. At this point, the split is on the stage.
split results, relative to the resolution before a lot of clarity, easy to analyze, to find details. Visible, split is one of the essential thinking of analysts.
third method: dimension reduction
have a lot of data in the face of the situation, but have no experience? When there are too many data dimensions, it is impossible for us to analyze each dimension. There are some related indicators that can be used to filter out the representative dimensions.
so many dimensions, in fact, does not have to be analyzed by each. We know that the number of transactions users / visitors = conversion rate, when the existence of such dimensions can be calculated through the other two dimensions, we can reduce the dimension. The number of users, the number of visitors and conversion rate, as long as three election two can. In addition, the number of transactions * customer unit = sales, the three can also choose three two.
fourth method: increasing dimension
increases and decreases dimensionality is correspondence, there is decrease, there will be increase. When our current dimension cannot be >