BI Application: Data analysis and data mining era coming

After several years of accumulation, most of the large enterprises and institutions have established a relatively perfect CRM, ERP, OA and other basic information systems. The unified characteristics of these systems are: through the operation of business personnel or users, the final database to add, modify, delete and other operations. The above system can be unified known as OLTP (online Transaction process, on-line transaction processing), refers to the system after a period of time, must help enterprises to collect a large number of historical data. However, in the database dispersed, independent existence of a large number of data for the business people, is only a few can not understand the Bible. What business people need is information, the abstract information that they can understand, understand, and benefit from. At this point, how to transform data into information, so that business people (including managers) can fully grasp, use this information, and assist decision-making, is the main problem of business intelligence.

How do you turn the data that exists in the database into the information that business people need? Most of the answers are the reporting system. Simply put, the reporting system is already called BI, which is the low-end implementation of BI.

Now foreign enterprises, most of them have entered the mid-tier bi, called data analysis. Some companies have begun to enter high-end bi, called data mining. But the enterprise of our country, most still stay in the report stage at present.

Data reports cannot be replaced

Traditional report system technology has been quite mature, we are familiar with Excel, Crystal Report, Reporting Service has been widely used. However, with the increase of data and the increase of demand, the traditional report system is facing more and more challenges.

1. Too much data, too little information

The dense tables stack up a lot of data, how many business people look at each data carefully? What information does this data represent, and what trends? The higher the level of leadership, the more concise information is needed. If I were the chairman, I might just need a word: is our situation good, medium or bad?

2. Difficulty in interactive analysis, understanding of various combinations

Customized reports are too rigid. For example, we can list sales of different regions, different products in one table, and sales of customers in different regions and different ages. However, these two tables are unable to answer such questions as "the purchase of digital camera type products by young and middle-aged customers in north China". Business problems often require an interactive analysis of multiple angles.

3. It's difficult to dig out potential rules

The reporting system is often listed on the surface of the data information, but the vast number of data in the depths of the potential to contain what rules? What customer is the most valuable to us, the degree of correlation between products? The deeper the rules, the greater the value of decision support, but the harder it is to dig out.

4. It is difficult to trace history and data to form Islands

There are a lot of business systems and data in different places. Too old data, such as data from a year ago, is often backed up by a business system, making it difficult to make macro-analysis and long-term historical analysis.

Therefore, with the development of the Times, the traditional reporting system can not meet the growing business needs, enterprises look forward to new technology. The age of data analysis and data mining is coming. It is noteworthy that the purpose of the data analysis and data mining system is to bring us more decision support value, not to replace the data report. Report system still has its irreplaceable advantages, and will be long-term and data analysis, mining system together.

Eight-D above data analysis

If OLTP focuses on daily business operations such as adding, modifying, and deleting databases, OLAP (online Analytics Process, on-line analysis System) focuses on macro issues, comprehensive analysis of data, and access to valuable information.

In order to achieve the goal of OLAP, traditional relational database is not enough, need a new technology called multidimensional database.

The concept of multidimensional databases is not complex. For example, we want to describe the April 2003 Coke in the northern region sales of 100,000 yuan, involving a number of angles: time, products, regions. These are called dimensions. As for sales, it's called measures. Of course, there are costs, profits and so on.

As shown in Figure 2, each dimension represents the time, product, and region, and the cell on the cube represents the measure. Further, the dimensions can be divided into different levels.

　　

Figure 1 Analysis of sales in a case and classification of current products