数据挖掘概要.ppt

Any Questions? * Why we want to make this review? There are some motivations Today’s internet evolves more and more complicated. many applications use techniques for avoiding detections, such as (random ports, encrypted data transmission, proprietary communications, etc) On the other side, there are many classification works in research area. Usually, researchers collect data and do an analysis. But we do not find any systematic results Therefore we try to fill the gap * Flows * Here is how k-Nearest Neighbors method works. The k-NN is a method for classifying objects based on the closest k training instances in the n-dimensional feature space. Suppose we have two application classes on two dimensional feature space, Say, Feature X and Y. Feature X could be the size of the first packet in an application flow, And the Feature Y could be, the size of the 2nd packet. At first, K-NN just memorizes the location of all the training instances in the two dimensional feature space, based on their feature values, like this. And then, when a new training instance arrives, it just assigns the testing instance To the majority class of its k-nearest training instances. In this way, this algorithm just considers the inherent euclidean closeness or farness between feature values of given training and testing instances, nothing else, Thus often used to measure discriminative power of features. * * * Data Mining: Concepts and Techniques 智能信息处理32学时 费高雷 fgl@ 电子科技大学 通信与信息工程学院 教师信息 费高雷 电话邮箱：fgl@ 地址：科研楼B325 研究方向： 网络层析成像 反演理论与方法 复杂多维信息处理 提纲 引言 数据挖掘概念及必要性 数据挖掘的主要任务 案列分析 Teaching Material 数据挖掘：概念与技术（原书第3版） [美]Jiawei Han?，等?著 范明?，孟小峰?译 机械工业出版社 特点：大量的图解、实例和练习 参考：第2版 Reference Books 考核方式 成绩构成： 期末70%、期中5%、平时25% （平时 ＝ 实验15% + 考勤10%） 考试方式： 期中随堂开卷、期末开卷 实验成绩： 结果分析50%、报告50% 课堂讲述、课后完成 Content 1、引言 2、认识数据 3、数据预处理 4、数据仓库与联机分析处理（自学） 5、数据立方体技术 （自学） 6、挖掘关联规则（重点） 7、高级模式挖掘（自学） 8、分类：基本概念（重点） 9、分类：高级方法（自学） 10、聚类分析：基本概念和方法（重点） 11、高级聚类分析（自学） 12、离群点检测 13、智能信息处理技术前沿 数据挖掘技术不断进步