原创力文档- 1、原创力文档（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。。
- 2、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击 这里二次下载。
- 3、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。
- 4、该文档为VIP文档，如果想要下载，成为VIP会员后，下载免费。
- 5、成为VIP后，下载本文档将扣除1次下载权益。下载后，不支持退款、换文档。如有疑问请联系我们。
- 6、成为VIP后，您将拥有八大权益，权益包括：VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
- 7、VIP文档为合作方或网友上传，每下载1次， 网站将根据用户上传文档的质量评分、类型等，对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档
查看更多
DB Seminar Series_ HARP_ A Hierarchical Algorithm with Automatic ...
DB Seminar Series: HARP: A Hierarchical Algorithm with Automatic Relevant Attribute Selection for Projected Clustering Presented by: Kevin Yip 20 September 2002 Short Summary Our own work (unpublished), supervised by Dr. Cheung and Dr. Ng Problem: to cluster datasets of very high dimensionality Assumption: clusters are formed in subspaces Short Summary Previous approaches: either have special restrictions on the dataset or target clusters, or cannot determine the dimensionality of the clusters automatically Our approach: not restricted by these limitations Presentation Outline Clustering Projected clustering Previous approaches to projected clustering Our approach: HARP Concepts Implementation: HARP.1 Experiments Future work and conclusions Clustering Goal: given a dataset D with N records and d attributes (dimensions), partition the records into k disjoint clusters such that Intra-cluster similarity is maximized Inter-cluster similarity is minimized Clustering How to measure similarity? Distance-based: Manhattan distance, Euclidean distance, etc. Correlation-based: cosine correlation, Pearson correlation, etc. Link-based (common neighbors) Pattern-based Clustering 2 common types of clustering algorithms: Partitional: selects some representative points for each cluster, assigns all other points to their closest clusters, and then re-determines the new representative points Hierarchical (agglomerative): repeatedly determines the two most similar clusters, and merges them Clustering Partitional clustering: Clustering Hierarchical clustering: Projected Clustering Assumption (general case): each cluster is formed in a subspace Projected Clustering A 3-D view: Projected Clustering An example dataset: Projected Clustering Projected clustering v.s. feature selection: Feature selection selects a feature set for all records, but projected clustering selects attribute sets individually for each cluster Feature selection is a preprocessing task, but projected clustering select
文档评论（0）