دانلود رایگان مقاله انگلیسی یک مدل رویداد متحد چهارگانه برای وب کاوی - اشپرینگر 2017

Abstract

In order to improve the quality of web data mining algorithm, this paper summarizes the advantages and disadvantages of several web data source models, including web log, application server log, Client-side log, Packet sniffer, and 5-gram united events model. Based on this analysis, a new 4- gram united events model (UEM4) is proposed in this paper. Simulation experiments were conducted to verify the performance of UEM4, compared with web log and 5-gram united events model. The experiment results show that web log has the worst session identification performance; UEM5 has high accuracy, best online and offline performance, but it needs the application system support the ability to identify the session; UEM4 does not require the application system to support session identification, and also has a good accuracy and performance of session identification. Therefore, this model can be used in e-commerce, which can provide high quality data sources for web mining algorithms and improve the quality of intelligent services.

4 Conclusion

In order to solve the problems of web data source, this paper proposes a web data source model UEM4 based on application layer record. The performance of the model is verified by simulation experiments, and the performance is compared with that of UEM5 and web log. Experimental results show that: (1) web log has the worst performance among the three model. (2) like UEM5, UEM4 has four advantages: firstly, it is more accurate and convenient user session identification than web log, and can solve the problem of a series of web log pre-processing; secondly, it is well integrated with the purchase, browsing and other types of events; thirdly, it is compatible with the existing web mining algorithm; fourthly, it supports multi-dimensional and multi-level web mining analysis. (3) UEM5 has a higher accuracy rate than UEM4, but for UEM5, the application system needs the ability to support session identification, which needs higher requirements on the performance of the application system; UEM4 does not require the application system to support session identification, and also has a good accuracy and performance of session identification. Which model to choose depends on the specific requirements of the user’s UEM system.

In summary, UEM4 model provides a high quality data source for web mining algorithm, and has a good recognition accuracy and performance. The data records of various e-commerce can be easily added in the model. The new Web data source model is proposed, which provides a high quality data source for the intelligent e-commerce site, and thus improves the quality of intelligent service.