Research Of Data Analysis Process Based On MDA

With the development of computer technology, management information system becomes popular and universal. They help the school to collect a large number of historical data after the system is put into use for a certain period of time. But they can't carry out multi-dimensional data analysis, and they are insufficient in supporting management faculty to make right decision. So people began to think of ways to find the useful information behind such large amounts of data. In this case, many kinds of data analysis techniques have emerged.At present, data analysis has many applications in the commercial area and data analysis also provides a general process. But their ultimate goal is the analysis of financial data. As a nonprofit industry, education is always confronting some difficulties, such as lack of funds and technical personnel. In addition to the distinction between education and business, it's obvious that recent data analysis implements, measures and processing details cannot meet the need, thus grope of brand new data analysis process according to the features of education industry seems badly necessary. Moreover, High cost and obsessive complicated measure is not suitable for education, in order to fill the requirement; it needs to search for new methods. Therefore data analysis technology is facing some new challenges.The points of this paper are mainly as follows:First, establish the data model. Extract raw data from every data source and design for their data structures, data operations and data restrict to establish the data model in middle database according to their feature; after determine analysis theme and analysis facts, abstract the logic relationship between them to form the data model in target data-side. Second, establish data collection model, descript the process of data extraction-data transition-data extraction as a platform-independent model, in which the specific content and details of the adoption are be descript by XML configuration, to overcome the limitations of hard-coded and to enhance the flexibility, scalability and portability. Third, take the combination of Intermediate library and documentation for data extraction, to solve the data acquisition problems in heterogeneous data sources; make different rules to handle different situations of null values; take the packet-based method for the detection of approximately duplicated records, assign different weight for different keywords, according to which to repeat the process of detection for the improvement of Accuracy. Concludes propose a new method for generation of the questionnaire automatically to collect survey results in real time on-line. It overcomes the drawbacks of traditional paper-based questionnaire.In this paper, we do some research about how to conduct data analysis based on model-driven architecture, with the hope that to solve the problem with an effective, common idea and method. The basis of this paper is some famous technology in data analysis area, it not only provides the ideas and methods to enhance the interoperability of data analysis, but also provide some help for the expansion of model-driven architecture applications. So this paper not only has theoretical research value, but also has great application value and practical significance.