EM算法：
在Eclipse中写出读取文件的代码然后调用EM算法计算输出结果：
package EMAlg; import java.io.*; import weka.core.*; import weka.filters.Filter; import weka.filters.unsupervised.attribute.Remove; import weka.clusterers.*; public class EMAlg { public EMAlg() { // TODO Auto-generated constructor stub System.out.println("this is the EMAlg"); } public static void main(String[] args) throws Exception { // TODO Auto-generated method stub String file="C:\Program Files/DataMining/Weka-3-6-10/data/labor.arff"; FileReader FReader=new FileReader(file); BufferedReader Reader= new BufferedReader(FReader); Instances data=new Instances(Reader); data.setClassIndex(data.numAttributes()-1);//设置最后一个属性作为分类属性 Remove filter=new Remove(); System.out.println("''+data.classIndex()的输出内容是："+""+data.classIndex()); System.out.println("读取数据的属性个数一共有："+data.numAttributes()+"个."); filter.setAttributeIndices(""+(data.classIndex()+1)); /*filter.setAttributeIndices(); * Set which attributes are to be deleted (or kept if invert is true) * 用来设置哪一个属性应该被删除的方法。 * Parameters: * rangeList - a string representing the list of attributes. Since the string will typically come from a user, attributes are indexed from 1. * eg: first-3,5,6-last */ filter.setInputFormat(data); /* * public boolean setInputFormat(Instances instanceInfo)throws java.lang.Exception * Sets the format of the input instances(设置输入数据的格式). If the filter is able to determine the output format before seeing any input instances, it does so here（如果过滤器在查看任何输入文件之前可以决定 输入文件的格式，那么这个函数就放在这里）. * This default implementation clears the output format and output queue, and the new batch flag is set. * Overriders should call super.setInputFormat(Instances) * Parameters: * instanceInfo - an Instances object containing the input instance structure (any instances contained in the object are ignored - only the structure is required). * Returns: * true if the outputFormat may be collected immediately * Throws: * java.lang.Exception - if the inputFormat can't be set successfully */ Instances dataCluster=Filter.useFilter(data, filter); /*public static Instances useFilter(Instances data,Filter filter)throws java.lang.Exception * Filters an entire set of instances through a filter and returns the new set. * 传入两个参数，第一个是需要进行过滤的数据，第二个是使用的过滤器，返回只为新的数据集。 * Parameters: * data - the data to be filtered * filter - the filter to be used * Returns: * the filtered set of data * Throws: * java.lang.Exception - if the filter can't be used successfully */ EM clusterer=new EM(); /* * public class EM * extends RandomizableDensityBasedClusterer * implements NumberOfClustersRequestable, WeightedInstancesHandler * Simple EM (expectation maximisation) class. * EM assigns a probability distribution to each instance which indicates the probability of it belonging to each of the clusters. EM can decide how many clusters to create by cross validation, or you may specify apriori how many clusters to generate. * The cross validation performed to determine the number of clusters is done in the following steps: * 1. the number of clusters is set to 1 * 2. the training set is split randomly into 10 folds. * 3. EM is performed 10 times using the 10 folds the usual CV way. * 4. the loglikelihood is averaged over all 10 results. * 5. if loglikelihood has increased the number of clusters is increased by 1 and the program continues at step 2. * The number of folds is fixed to 10, as long as the number of instances in the training set is not smaller 10. If this is the case the number of folds is set equal to the number of instances. * Valid options are: * -N <num> * number of clusters. If omitted or -1 specified, then cross validation is used to select the number of clusters. * -I <num> * max iterations.(default 100) * -V * verbose. * -M <num> * minimum allowable standard deviation for normal density computation * (default 1e-6) * -O * Display model in old format (good when there are many clusters) * -S <num> * Random number seed.(default 100) */ String [] options=new String[4]; // max. iterations //最大迭代次数 options[0] = "-I"; options[1] = "100"; //set cluster numbers,设置簇的个数 options[2]="-N"; options[3]="2"; clusterer.setOptions(options); clusterer.buildClusterer(dataCluster); //clusterer.buildClusterer(dataClusterer); // evaluate clusterer ClusterEvaluation eval = new ClusterEvaluation(); eval.setClusterer(clusterer); eval.evaluateClusterer(data); // print results System.out.println("数据总数："+data.numInstances()+"属性个数为："+data.numAttributes()); System.out.println(eval.clusterResultsToString()); } }
使用的数据是Weka安装目录下data文件夹中的labor.arff文件。
输出的结果是：
''+data.classIndex()的输出内容是：16 读取数据的属性个数一共有：17个. 数据总数：57属性个数为：17 EM == Number of clusters: 2 Cluster Attribute 0 1 (0.14) (0.86) ================================================= duration mean 1.5702 2.2532 std. dev. 0.4953 0.6764 wage-increase-first-year mean 3.0708 3.9184 std. dev. 1.0028 1.3571 wage-increase-second-year mean 3.8141 3.9964 std. dev. 0.8153 1.0624 wage-increase-third-year mean 3.9133 3.9133 std. dev. 0.6522 0.6952 cost-of-living-adjustment none 7.0614 36.9386 tcf 1.3707 8.6293 tc 2.2872 6.7128 [total] 10.7192 52.2808 working-hours mean 39.4412 37.8196 std. dev. 0.8911 2.4268 pension none 6.4515 6.5485 ret_allw 2.3211 3.6789 empl_contr 1.9466 42.0534 [total] 10.7192 52.2808 standby-pay mean 6.7945 7.5462 std. dev. 1.4912 1.918 shift-differential mean 3.4074 5.1002 std. dev. 1.6629 3.4277 education-allowance yes 3.167 8.833 no 6.5522 42.4478 [total] 9.7192 51.2808 statutory-holidays mean 10.555 11.1788 std. dev. 0.572 1.2533 vacation below_average 4.657 21.343 average 4.0313 14.9687 generous 2.0309 15.9691 [total] 10.7192 52.2808 longterm-disability-assistance yes 2.9977 48.0023 no 6.7215 3.2785 [total] 9.7192 51.2808 contribution-to-dental-plan none 7.1218 3.8782 half 2.5419 34.4581 full 1.0556 13.9444 [total] 10.7192 52.2808 bereavement-assistance yes 5.7192 50.2808 no 4 1 [total] 9.7192 51.2808 contribution-to-health-plan none 6.2887 3.7113 half 1.8752 9.1248 full 2.5554 39.4446 [total] 10.7192 52.2808 Clustered Instances 0 8 ( 14%) 1 49 ( 86%) Log likelihood: -18.37167 Class attribute: class Classes to Clusters: 0 1 <-- assigned to cluster 8 12 | bad 0 37 | good Cluster 0 <-- bad Cluster 1 <-- good Incorrectly clustered instances : 12.0 21.0526 %