| The development of Internet technology and the drive by human needs, together with the development of database technology, provide the human with the ability to store and manage efficiently huge amounts of data. The human accumulate more and more data, and it’s so-called big data era. Facing huge amounts of data, can human find useful information and guide human behavior? The answer is yes, in fact, the data mining technology can solve the problem. Data mining has been widely applied in all walks of life. Research work in this paper is under the big background and demand.Because of C++ language owing the compatibility with the C language, and syntax and structure are different in different development language, which means the API sequence extraction method is different. This article choose more mainstream language- C/C++ language, which means the input of data mining research in this paper is C/C++ code library. When C/C + + developers use of API,they used to take the way the manual search and SE(search engine) search, finding what they need is very difficult.Although related code can be found by search technology now, code snippet found is very much, and one by one analysis is also very time-consuming. The work in this paper basing on C/C + + code library is to research API invocation pattern mining and realize the system by software. The main work of this paper is as follows:1. In this paper, after researching the extraction technology of the sequence about C/C++ language, put forward an effective method to extract sequence.2. By designing the database under this application scenario, the system can access to database remotely, which lay the foundation of parallel processing for system.3. After researching the clustering algorithm, this paper designs and realizes the application of clustering algorithm.4. After researching effect evaluation standards about the clustering, boldly put forward the reasonable hypothesis, puts forward the judgment standard of clustering results under this application scenario.5. Designed and implemented the search module, the module has the abilty of clustering quickly and query, return the query results to the user.6. Designed and implemented mining system tool of API invocation pattern basing on C/C++ code library.This tool has four characteristics, as follows1. High performance, the tool can complete extracting API sequence of 100 k C/C++ code in 5 minutes, and the time of usage pattern search response on every ten thousand API is less than 2 second.2. Dealing with large data, the mining tools can finish mining API usage pattern about millions of lines of C/C++ code.3. Special functions, complete the cluster analysis and support special syntax analysis, like overloaded functions and scope in C++.4. Modular, pattern clustering function is not coupled with other function, can be used alone, support pattern mining and clustering analysis of multiple code snippet.In this paper, I research and implement mining system, API sequence extracted is complete and effective by the system. The system can complete the clustering analysis,and has a high degree of clustering accuracy, can meet the demand of practical application. |
