A Study For Discretization Of Real Value Attributes And LMS Algorithm

The discretization of continuous feature values is an effective technique to deal with continuous attributes for machine learning and data mining. Some algorithms in Rule extraction and feature classification can only handle categorical attribute. Discretization is a technique to partition continuous attributes into a finite set of adjacent intervals in order to generate attributes with a small number of distinct values. Reasonability of a discretization process is determined by the accuracy of expression and extraction for information. Discrete algorithm is the key to how to obtain the optimal partition, to maximize the significance of maintaining the information that reduce the loss of information. Rough Set theory by Poland scientist Z.Pawlak puts forward in 1982, it can be used for processing the decision information table not sure knowledge, in data analysis, data mining etc had been used widely, traditional Rough Set theory to the database discrete only attribute processing.The paper from the viewpoint of rough sets, and ensuring the condition attributes and the decision attribute relative relation is changeless, puts forward a new method of interval segmentation discretization algorithm experimental results and theoretical proves the effectiveness of the algorithm.Artificial neural network (ANN) unique information processing and computing power, people on it gives more attention. ANN used neuronal group of forming a nets, can simulate any complicated nonlinear system that is a kind of to solve practical engineering problems, because the effective tools hidden units have been introduced ANN that has stronger classification and memory. In order to expand the application range of the neural network by BP model building, the nonlinear relationship between the input vector, deviation and learning steps, this paper proposes a new based on neural network control variable step length LMS algorithm. It through adaptive learning process to determine the steps that also puts forward how to determine whether a similar method input signal.In recent years the development from adaptive filter algorithm is very quickly, adaptive filter algorithm is the important foundation, the signal processing in various fields made extensive application.The simulation experiment also showed that performance of the algorithm is improved.The main work of four aspects of this paper as follows:1> Discussing our research background of discretization, the rough set theory and the neural network theory , and structure of this article.2> Introducing the process of discretization and the status of discrete algorithm conducted a comprehensive study.3> Proposing based on rough set theory a new algorithm for discretization of continuous attributes, The experiments are performed respectively with the results of discreted data by using C4.5 and SVM. The results show that the presented algorithm is effective.4> Describing the neural network is a very good data mining tools, has a good non-linear processing. But the neural network to learn the information is implicit in a large number of connection weights, it is difficult to understand. This paper presents a new neural network control based on variable step size LMS algorithm. And the correlations of the simulation experiments...