(作者：玛瑙河，转载请注明作者或出处，
)
以RBFNetwork为例，简要说明Jython + Weka 协同工作的方法。
1. install weka (into /opt/weka/ or elsewhere) & Jython
2. export CLASSPATH="$CLASSPATH:/opt/weka/weka.jar"
3. jython rbfnetwork.py traindata.arff testdata1.arff testdata2.arff ...
下面是rbfnetwork.py的代码
rbfnetwork.py
1 #!/usr/bin/env jython
2 import sys,os
3
4 import java.io.FileReader as FileReader
5 import java.lang.StringBuffer as StringBuffer
6 import java.lang.Boolean as Boolean
7 import java.io.ObjectOutputStream as ObjectOutputStream
8 import java.io.ObjectInputStream as ObjectInputStream
9 import java.io.FileOutputStream as FileOutputStream
10 import java.io.FileInputStream as FileInputStream
11 import weka.core.Instances as Instances
12 import weka.classifiers.functions.RBFNetwork as RBFNetwork
13 import weka.classifiers.Evaluation as Evaluation
14 import weka.core.Range as Range
15 import weka.core.Utils.splitOptions as splitOptions
16 import weka.classifiers.evaluation.output.prediction.PlainText as PlainText
17
18
19 # check commandline parameters
20 if ( (len(sys.argv) < 3)):
21 print "Usage: rbfnetwork.py <TRAIN-ARFF-file> <TEST-ARFF-file> ..."
22 sys.exit()
23
24 # load data file
25 print >>sys.stdout,"Loading data..."
26 train_file = FileReader(sys.argv[1])
27
28 train_data = Instances(train_file)
29
30 # set the class Index - the index of the dependent variable
31 train_data.setClassIndex(train_data.numAttributes() - 1)
32
33 model=sys.argv[1]+".model"
34 if os.path.exists(model):
35 #load existed model
36 f_in = FileInputStream (model);
37 obj_in = ObjectInputStream (f_in);
38 rbfnetwork = obj_in.readObject ();
39
40 print "--> Use exsisted model: %s" % model
41 else:
42 # create the model
43 options=splitOptions("-B 2 -S 1 -R 1.0E-8 -M -1 -W 0.1")
44 rbfnetwork = RBFNetwork()
45 rbfnetwork.setOptions(options)
46 rbfnetwork.buildClassifier(train_data) # only a trained classifier can be evaluated
47
48 #save model
49 f_out = FileOutputStream (model);
50 obj_out = ObjectOutputStream (f_out);
51 obj_out.writeObject (rbfnetwork);
52
53 # print out the built model
54 print "--> Generated model:\n"
55
56 print rbfnetwork
57 print "="*80
58
59 for i in range(len(sys.argv)-2):
60 test_file = FileReader(sys.argv[i+2])
61 test_data = Instances(test_file)
62 test_data.setClassIndex(train_data.numAttributes() - 1)
63 buffer = StringBuffer() # buffer for the predictions
64 output=PlainText()
65 output.setHeader(test_data)
66 output.setBuffer(buffer)
67
68 attRange = Range() # attributes to output
69 outputDistribution = Boolean(True) # we want distribution
70 evaluation = Evaluation(train_data)
71 evaluation.evaluateModel(rbfnetwork, test_data, [output, attRange, outputDistribution])
72
73 print "--> Evaluation for %s:\n" % sys.argv[i+2]
74 print evaluation.toSummaryString()
75 print evaluation.toMatrixString()
76 print "-"*80
77 print "--> Predictions for %s:\n" % sys.argv[i+2]
78 print buffer
79 print "="*80
80
81
2 import sys,os
3
4 import java.io.FileReader as FileReader
5 import java.lang.StringBuffer as StringBuffer
6 import java.lang.Boolean as Boolean
7 import java.io.ObjectOutputStream as ObjectOutputStream
8 import java.io.ObjectInputStream as ObjectInputStream
9 import java.io.FileOutputStream as FileOutputStream
10 import java.io.FileInputStream as FileInputStream
11 import weka.core.Instances as Instances
12 import weka.classifiers.functions.RBFNetwork as RBFNetwork
13 import weka.classifiers.Evaluation as Evaluation
14 import weka.core.Range as Range
15 import weka.core.Utils.splitOptions as splitOptions
16 import weka.classifiers.evaluation.output.prediction.PlainText as PlainText
17
18
19 # check commandline parameters
20 if ( (len(sys.argv) < 3)):
21 print "Usage: rbfnetwork.py <TRAIN-ARFF-file> <TEST-ARFF-file> ..."
22 sys.exit()
23
24 # load data file
25 print >>sys.stdout,"Loading data..."
26 train_file = FileReader(sys.argv[1])
27
28 train_data = Instances(train_file)
29
30 # set the class Index - the index of the dependent variable
31 train_data.setClassIndex(train_data.numAttributes() - 1)
32
33 model=sys.argv[1]+".model"
34 if os.path.exists(model):
35 #load existed model
36 f_in = FileInputStream (model);
37 obj_in = ObjectInputStream (f_in);
38 rbfnetwork = obj_in.readObject ();
39
40 print "--> Use exsisted model: %s" % model
41 else:
42 # create the model
43 options=splitOptions("-B 2 -S 1 -R 1.0E-8 -M -1 -W 0.1")
44 rbfnetwork = RBFNetwork()
45 rbfnetwork.setOptions(options)
46 rbfnetwork.buildClassifier(train_data) # only a trained classifier can be evaluated
47
48 #save model
49 f_out = FileOutputStream (model);
50 obj_out = ObjectOutputStream (f_out);
51 obj_out.writeObject (rbfnetwork);
52
53 # print out the built model
54 print "--> Generated model:\n"
55
56 print rbfnetwork
57 print "="*80
58
59 for i in range(len(sys.argv)-2):
60 test_file = FileReader(sys.argv[i+2])
61 test_data = Instances(test_file)
62 test_data.setClassIndex(train_data.numAttributes() - 1)
63 buffer = StringBuffer() # buffer for the predictions
64 output=PlainText()
65 output.setHeader(test_data)
66 output.setBuffer(buffer)
67
68 attRange = Range() # attributes to output
69 outputDistribution = Boolean(True) # we want distribution
70 evaluation = Evaluation(train_data)
71 evaluation.evaluateModel(rbfnetwork, test_data, [output, attRange, outputDistribution])
72
73 print "--> Evaluation for %s:\n" % sys.argv[i+2]
74 print evaluation.toSummaryString()
75 print evaluation.toMatrixString()
76 print "-"*80
77 print "--> Predictions for %s:\n" % sys.argv[i+2]
78 print buffer
79 print "="*80
80
81
如果你有大量的类似任务需要运行，你还可以将这些任务提交到任务管理系统如Torque等，然后你就不用管了，等着任务运行完毕后系统给你发送通知邮件吧。
