我正在试验Weka，我正试图用它来预测赛马的结果。我有很多数据要使用，我现有的属性是：
我使用数据进行训练和评估，其中所有属性都已知，然后我提供了一个数据集，其中place-attribute未知。但我无法弄清楚如何使用weka来获取place属性的预测值。下面的代码来自我的小型测试应用程序，非常感谢帮助让我朝着正确的方向前进。
使用classifier.classifyInstance（...）时得到的值总是0.0
我从运行这个程序得到的输出是：
Relation Name: QueryResult
Num Instances: 2475
Num Attributes: 7
Name Type Nom Int Real Missing Unique Dist
1 hbnHorse_id Num 0% 100% 0% 0 / 0% 1514 / 61% 1873
2 hbnDriver_id Num 0% 100% 0% 0 / 0% 274 / 11% 645
3 hbnTrackCondition_id Num 0% 97% 0% 84 / 3% 0 / 0% 3
4 raceDistance Nom 100% 0% 0% 0 / 0% 7 / 0% 26
5 hbnTrack_id Num 0% 98% 0% 53 / 2% 2 / 0% 26
6 kmTime Num 0% 100% 0% 0 / 0% 74 / 3% 246
7 place Num 0% 100% 0% 0 / 0% 1 / 0% 10
Results
======
Correctly Classified Instances 909 36.7421 %
Incorrectly Classified Instances 1565 63.2579 %
Kappa statistic 0.1341
Mean absolute error 0.1313
Root mean squared error 0.3162
Relative absolute error 84.6014 %
Root relative squared error 113.7708 %
Total Number of Instances 2474
Ignored Class Unknown Instances 1
Relation Name: QueryResult-weka.filters.unsupervised.attribute.NumericToNominal-R7-7
Num Instances: 1643
Num Attributes: 7
Name Type Nom Int Real Missing Unique Dist
1 hbnHorse_id Num 0% 100% 0% 0 / 0% 1593 / 97% 1618
2 hbnDriver_id Num 0% 100% 0% 0 / 0% 238 / 14% 508
3 hbnTrackCondition_id Num 0% 100% 0% 0 / 0% 0 / 0% 2
4 raceDistance Nom 100% 0% 0% 0 / 0% 0 / 0% 5
5 startMethod Nom 100% 0% 0% 0 / 0% 0 / 0% 2
6 kmTime Num 0% 100% 0% 0 / 0% 0 / 0% 1
7 place Nom 100% 0% 0% 1 / 0% 0 / 0% 1
源代码：
package org.peanuts.weka.core;
import weka.classifiers.Classifier;
import weka.classifiers.bayes.NaiveBayes;
import weka.classifiers.evaluation.Evaluation;
import weka.classifiers.trees.J48;
import weka.core.Instances;
import weka.experiment.InstanceQuery;
import weka.filters.Filter;
import weka.filters.unsupervised.attribute.NumericToNominal;
import java.io.File;
public class WekaTest {
public static void main(String args[]) {
try {
InstanceQuery query = new InstanceQuery();
query.setCustomPropsFile(new File("C:\\codez\\WekaPlayground\\src\\main\\resources\\props\\DatabaseUtils.props"));
query.setUsername("wekaplayground");
query.setPassword("password");
// query.setQuery("select s.hbnHorse_id, s.hbnDriver_id, r.hbnTrackCondition_id, r.raceDistance, r.startMethod, r.hbnTrack_id, s.kmTime, s.place\n" +
query.setQuery("select s.hbnHorse_id, s.hbnDriver_id, r.hbnTrackCondition_id, r.raceDistance, r.hbnTrack_id, s.kmTime, s.place\n" +
"from hbnstart s, hbnrace r\n" +
"where s.hbnRace_id = r.id\n" +
"and r.hasResults = 1 " +
// "and s.place > 0 " +
"and s.id < 22000");
// You can declare that your train set is sparse
// query.setSparseData(true);
Instances train = query.retrieveInstances();
if (train.classIndex() == -1)
train.setClassIndex(train.numAttributes() - 1);
query.setQuery("select s.hbnHorse_id, s.hbnDriver_id, r.hbnTrackCondition_id, r.raceDistance, r.hbnTrack_id, s.kmTime, s.place\n" +
"from hbnstart s, hbnrace r\n" +
"where s.hbnRace_id = r.id\n" +
"and r.hasResults = 1 " +
// "and s.place > 0 " +
"and s.id > 22000");
Instances test = query.retrieveInstances();
if (test.classIndex() == -1)
test.setClassIndex(test.numAttributes() - 1);
System.out.println(test.toSummaryString());
Classifier classifier = createClassifier();
classifier.buildClassifier(convertToNominal(train));
// evaluate classifier and print some statistics
Evaluation eval = new Evaluation(convertToNominal(train));
eval.evaluateModel(classifier, convertToNominal(test));
System.out.println(eval.toSummaryString("\nResults\n======\n", false));
if (train.classIndex() == -1)
train.setClassIndex(train.numAttributes() - 1);
query.setQuery("select s.hbnHorse_id, s.hbnDriver_id, r.hbnTrackCondition_id, r.raceDistance, r.startMethod, s.kmTime, s.place\n" +
"from hbnstart s, hbnrace r\n" +
"where s.hbnRace_id = r.id\n" +
"and r.hasResults = 0");
Instances unlabeled = query.retrieveInstances();
// set class attribute
unlabeled.setClassIndex(unlabeled.numAttributes() - 1);
unlabeled = convertToNominal(unlabeled);
// create copy
Instances labeled = new Instances(unlabeled);
System.out.println(labeled.toSummaryString());
// label instances
for (int i = 0; i < unlabeled.numInstances(); i++) {
double clsLabel = classifier.classifyInstance(unlabeled.instance(i));
labeled.instance(i).setClassValue(clsLabel);
}
} catch (Exception e) {
e.printStackTrace();
}
}
public static Instances convertToNominal(Instances instances) throws Exception {
instances.get(0).setClassMissing();
NumericToNominal convert = new NumericToNominal();
String[] numToNomOptions = new String[2];
numToNomOptions[0] = "-R";
numToNomOptions[1] = "7-7"; //range of variables to make numeric
convert.setOptions(numToNomOptions);
convert.setInputFormat(instances);
return Filter.useFilter(instances, convert);
}
private static Classifier createClassifier() throws Exception {
String[] options = new String[1];
options[0] = "-U"; // unpruned classifier
J48 classifier = new J48(); // new instance of classifier
classifier.setOptions(options);
return classifier;
}
}