Research And Application Of Several Issues On Intelligent Systems

There are many classic intelligent algorithms in pattern recognition. Those algorithms are widely used in intelligent system which makes great impact on the whole system. In the intelligent communication network, network bandwidth control is important for route choice and congestion control, which is depended on effective bandwidth measurement. But current available bandwidth measurement cost too much time, better algorithm is necessary to be put forward. In addition,with the development of optical network, all-optical network becomes more and more popular. Automatic Optical Main Distribution Frame (AODF) as an effective solution for the all-optical switch is valuable for research. At the same time, the popular of network make the quantity of comment data to be huge. How to classify the comments according to its emotion is an important issue. This paper introduces new intelligent algorithms or improves original algorithm to make the three systems mentioned above achieve better performance. The main achievements of this paper can be summarized as follows.1) We proposed a quick available bandwidth measurement algorithm named QuickAbw base on delay increase rate. The foundation of the algorithm is the available bandwidth calculating formula which is only be suitable for single congestion link in the path; thus, we proposed an algorithm for recognizing the count of the congestion links and a iteration algorithm to extend the QuickAbw to the situation which contains more than one congestion link. In the simulation section, the QuickAbw is realized in the NS2platform to confirm the theory and to compare with the other available bandwidth measurement tools. The experiment shows that QuickAbw gets better performance in the Measurement accuracy and measuring speed.2) Investigate an automatic optical wiring system model based on three-level Clos network, which can auto control the optical wiring. Compared with the manual wiring, it costs less money and time. The rearrangement algorithm for Clos network is improved to reduce the time complexity; it can also effectively improve the device response speed and reduce the hardware cost.3) Naive Bayesian classifier is widely used in text classification. However, it requires two basic assumptions as its prerequisite and the performance would have been poor if these two were dissatisfied. A multi-level naive Bayes classifier is proposed to make up the deficiency of the traditional naive Bayes classifier. The research below shows that the multi-level naive Bayes classifier gets better performance than the traditional naive Bayes classifier on the sentiment classification of movie reviews.