《【精品】weka教程-数据挖掘英文PPT课件.ppt》由会员分享，可在线阅读，更多相关《【精品】weka教程-数据挖掘英文PPT课件.ppt（173页珍藏版）》请在三一文库上搜索。
1、Machine Learning with WEKA,6/28/2019,University of Waikato,2,WEKA: the bird,Copyright: Martin Kramer (mkramerwxs.nl),6/28/2019,University of Waikato,3,WEKA: the software,Machine learning/data mining software written in Java (distributed under the GNU Public License) Used for research, education, and
2、 applications Complements “Data Mining” by Witten & Frank Main features: Comprehensive set of data pre-processing tools, learning algorithms and evaluation methods Graphical user interfaces (incl. data visualization) Environment for comparing learning algorithms,6/28/2019,University of Waikato,4,WEK
3、A: versions,There are several versions of WEKA: WEKA 3.0: “book version” compatible with description in data mining book WEKA 3.2: “GUI version” adds graphical user interfaces (book version is command-line only) WEKA 3.3: “development version” with lots of improvements This talk is based on the late
4、st snapshot of WEKA 3.3 (soon to be WEKA 3.4),6/28/2019,University of Waikato,5,relation heart-disease-simplified attribute age numeric attribute sex female, male attribute chest_pain_type typ_angina, asympt, non_anginal, atyp_angina attribute cholesterol numeric attribute exercise_induced_angina no
5、, yes attribute class present, not_present data 63,male,typ_angina,233,no,not_present 67,male,asympt,286,yes,present 67,male,asympt,229,yes,present 38,female,non_anginal,?,no,not_present .,WEKA only deals with “flat” files,Flat file in ARFF format,6/28/2019,University of Waikato,6,relation heart-dis
6、ease-simplified attribute age numeric attribute sex female, male attribute chest_pain_type typ_angina, asympt, non_anginal, atyp_angina attribute cholesterol numeric attribute exercise_induced_angina no, yes attribute class present, not_present data 63,male,typ_angina,233,no,not_present 67,male,asym
7、pt,286,yes,present 67,male,asympt,229,yes,present 38,female,non_anginal,?,no,not_present .,WEKA only deals with “flat” files,numeric attribute,nominal attribute,6/28/2019,University of Waikato,7,6/28/2019,University of Waikato,8,6/28/2019,University of Waikato,9,6/28/2019,University of Waikato,10,Ex
8、plorer: pre-processing the data,Data can be imported from a file in various formats: ARFF, CSV, C4.5, binary Data can also be read from a URL or from an SQL database (using JDBC) Pre-processing tools in WEKA are called “filters” WEKA contains filters for: Discretization, normalization, resampling, a
9、ttribute selection, transforming and combining attributes, ,6/28/2019,University of Waikato,11,6/28/2019,University of Waikato,12,6/28/2019,University of Waikato,13,6/28/2019,University of Waikato,14,6/28/2019,University of Waikato,15,6/28/2019,University of Waikato,16,6/28/2019,University of Waikat
10、o,17,6/28/2019,University of Waikato,18,6/28/2019,University of Waikato,19,6/28/2019,University of Waikato,20,6/28/2019,University of Waikato,21,6/28/2019,University of Waikato,22,6/28/2019,University of Waikato,23,6/28/2019,University of Waikato,24,6/28/2019,University of Waikato,25,6/28/2019,Unive
11、rsity of Waikato,26,6/28/2019,University of Waikato,27,6/28/2019,University of Waikato,28,6/28/2019,University of Waikato,29,6/28/2019,University of Waikato,30,6/28/2019,University of Waikato,31,6/28/2019,University of Waikato,32,Explorer: building “classifiers”,Classifiers in WEKA are models for pr
12、edicting nominal or numeric quantities Implemented learning schemes include: Decision trees and lists, instance-based classifiers, support vector machines, multi-layer perceptrons, logistic regression, Bayes nets, “Meta”-classifiers include: Bagging, boosting, stacking, error-correcting output codes
13、, locally weighted learning, ,6/28/2019,University of Waikato,33,6/28/2019,University of Waikato,34,6/28/2019,University of Waikato,35,6/28/2019,University of Waikato,36,6/28/2019,University of Waikato,37,6/28/2019,University of Waikato,38,6/28/2019,University of Waikato,39,6/28/2019,University of W
14、aikato,40,6/28/2019,University of Waikato,41,6/28/2019,University of Waikato,42,6/28/2019,University of Waikato,43,6/28/2019,University of Waikato,44,6/28/2019,University of Waikato,45,6/28/2019,University of Waikato,46,6/28/2019,University of Waikato,47,6/28/2019,University of Waikato,48,6/28/2019,
15、University of Waikato,49,6/28/2019,University of Waikato,50,6/28/2019,University of Waikato,51,6/28/2019,University of Waikato,52,6/28/2019,University of Waikato,53,6/28/2019,University of Waikato,54,6/28/2019,University of Waikato,55,6/28/2019,University of Waikato,56,6/28/2019,University of Waikat
16、o,57,6/28/2019,University of Waikato,58,6/28/2019,University of Waikato,59,6/28/2019,University of Waikato,60,6/28/2019,University of Waikato,61,6/28/2019,University of Waikato,62,6/28/2019,University of Waikato,63,6/28/2019,University of Waikato,64,6/28/2019,University of Waikato,65,6/28/2019,Unive
17、rsity of Waikato,66,6/28/2019,University of Waikato,67,6/28/2019,University of Waikato,68,6/28/2019,University of Waikato,69,6/28/2019,University of Waikato,70,6/28/2019,University of Waikato,71,6/28/2019,University of Waikato,72,6/28/2019,University of Waikato,73,6/28/2019,University of Waikato,74,
18、6/28/2019,University of Waikato,75,6/28/2019,University of Waikato,76,6/28/2019,University of Waikato,77,6/28/2019,University of Waikato,78,6/28/2019,University of Waikato,79,6/28/2019,University of Waikato,80,6/28/2019,University of Waikato,81,6/28/2019,University of Waikato,82,6/28/2019,University
19、 of Waikato,83,6/28/2019,University of Waikato,84,6/28/2019,University of Waikato,85,6/28/2019,University of Waikato,86,6/28/2019,University of Waikato,87,6/28/2019,University of Waikato,88,6/28/2019,University of Waikato,89,6/28/2019,University of Waikato,90,6/28/2019,University of Waikato,91,6/28/
20、2019,University of Waikato,92,Explorer: clustering data,WEKA contains “clusterers” for finding groups of similar instances in a dataset Implemented schemes are: k-Means, EM, Cobweb, X-means, FarthestFirst Clusters can be visualized and compared to “true” clusters (if given) Evaluation based on logli
21、kelihood if clustering scheme produces a probability distribution,6/28/2019,University of Waikato,93,6/28/2019,University of Waikato,94,6/28/2019,University of Waikato,95,6/28/2019,University of Waikato,96,6/28/2019,University of Waikato,97,6/28/2019,University of Waikato,98,6/28/2019,University of
22、Waikato,99,6/28/2019,University of Waikato,100,6/28/2019,University of Waikato,101,6/28/2019,University of Waikato,102,6/28/2019,University of Waikato,103,6/28/2019,University of Waikato,104,6/28/2019,University of Waikato,105,6/28/2019,University of Waikato,106,6/28/2019,University of Waikato,107,6
23、/28/2019,University of Waikato,108,Explorer: finding associations,WEKA contains an implementation of the Apriori algorithm for learning association rules Works only with discrete data Can identify statistical dependencies between groups of attributes: milk, butter bread, eggs (with confidence 0.9 an
24、d support 2000) Apriori can compute all rules that have a given minimum support and exceed a given confidence,6/28/2019,University of Waikato,109,6/28/2019,University of Waikato,110,6/28/2019,University of Waikato,111,6/28/2019,University of Waikato,112,6/28/2019,University of Waikato,113,6/28/2019,
25、University of Waikato,114,6/28/2019,University of Waikato,115,6/28/2019,University of Waikato,116,Explorer: attribute selection,Panel that can be used to investigate which (subsets of) attributes are the most predictive ones Attribute selection methods contain two parts: A search method: best-first,
26、 forward selection, random, exhaustive, genetic algorithm, ranking An evaluation method: correlation-based, wrapper, information gain, chi-squared, Very flexible: WEKA allows (almost) arbitrary combinations of these two,6/28/2019,University of Waikato,117,6/28/2019,University of Waikato,118,6/28/201
27、9,University of Waikato,119,6/28/2019,University of Waikato,120,6/28/2019,University of Waikato,121,6/28/2019,University of Waikato,122,6/28/2019,University of Waikato,123,6/28/2019,University of Waikato,124,6/28/2019,University of Waikato,125,Explorer: data visualization,Visualization very useful i
28、n practice: e.g. helps to determine difficulty of the learning problem WEKA can visualize single attributes (1-d) and pairs of attributes (2-d) To do: rotating 3-d visualizations (Xgobi-style) Color-coded class values “Jitter” option to deal with nominal attributes (and to detect “hidden” data point
29、s) “Zoom-in” function,6/28/2019,University of Waikato,126,6/28/2019,University of Waikato,127,6/28/2019,University of Waikato,128,6/28/2019,University of Waikato,129,6/28/2019,University of Waikato,130,6/28/2019,University of Waikato,131,6/28/2019,University of Waikato,132,6/28/2019,University of Wa
30、ikato,133,6/28/2019,University of Waikato,134,6/28/2019,University of Waikato,135,6/28/2019,University of Waikato,136,6/28/2019,University of Waikato,137,6/28/2019,University of Waikato,138,Performing experiments,Experimenter makes it easy to compare the performance of different learning schemes For
31、 classification and regression problems Results can be written into file or database Evaluation options: cross-validation, learning curve, hold-out Can also iterate over different parameter settings Significance-testing built in!,6/28/2019,University of Waikato,139,6/28/2019,University of Waikato,14
32、0,6/28/2019,University of Waikato,141,6/28/2019,University of Waikato,142,6/28/2019,University of Waikato,143,6/28/2019,University of Waikato,144,6/28/2019,University of Waikato,145,6/28/2019,University of Waikato,146,6/28/2019,University of Waikato,147,6/28/2019,University of Waikato,148,6/28/2019,
33、University of Waikato,149,6/28/2019,University of Waikato,150,6/28/2019,University of Waikato,151,6/28/2019,University of Waikato,152,The Knowledge Flow GUI,New graphical user interface for WEKA Java-Beans-based interface for setting up and running machine learning experiments Data sources, classifi
34、ers, etc. are beans and can be connected graphically Data “flows” through components: e.g., “data source” - “filter” - “classifier” - “evaluator” Layouts can be saved and loaded again later,6/28/2019,University of Waikato,153,6/28/2019,University of Waikato,154,6/28/2019,University of Waikato,155,6/
35、28/2019,University of Waikato,156,6/28/2019,University of Waikato,157,6/28/2019,University of Waikato,158,6/28/2019,University of Waikato,159,6/28/2019,University of Waikato,160,6/28/2019,University of Waikato,161,6/28/2019,University of Waikato,162,Can continue this.,6/28/2019,University of Waikato
36、,163,6/28/2019,University of Waikato,164,6/28/2019,University of Waikato,165,6/28/2019,University of Waikato,166,6/28/2019,University of Waikato,167,6/28/2019,University of Waikato,168,6/28/2019,University of Waikato,169,6/28/2019,University of Waikato,170,6/28/2019,University of Waikato,171,6/28/20
37、19,University of Waikato,172,6/28/2019,University of Waikato,173,Conclusion: try it yourself!,WEKA is available at http:/www.cs.waikato.ac.nz/ml/weka Also has a list of projects based on WEKA WEKA contributors: Abdelaziz Mahoui, Alexander K. Seewald, Ashraf M. Kibriya, Bernhard Pfahringer , Brent Martin, Peter Flach, Eibe Frank ,Gabi Schmidberger ,Ian H. Witten , J. Lindgren, Janice Boughton, Jason Wells, Len Trigg, Lucio de Souza Coelho, Malcolm Ware, Mark Hall ,Remco Bouckaert , Richard Kirkby, Shane Butler, Shane Legg, Stuart Inglis, Sylvain Roy, Tony Voyle, Xin Xu, Yong Wang, Zhihai Wang,