引言
接前一篇文章《【甘道夫】Mahout0.9 打patch使其支持 Hadoop2.2.0》
http://blog.csdn.net/u010967382/article/details/39088035，
为Mahout0.9打过Patch编译成功后，使用贝叶斯文本分类来测试Mahout0.9对Hadoop2.2.0的兼容性。
欢迎转载，转载请注明出处：
http://blog.csdn.net/u010967382/article/details/39088285
步骤一：将20news的文件都上传到hdfs
[email protected]:~/Mahout/mahout-distribution-0.7$ hadoop fs -ls /workspace/mahout/week4/data/20news
Found 2 items
drwxr-xr-x - yarn supergroup 0 2014-09-04 21:52 /workspace/mahout/week4/data/20news/20news-bydate-test
drwxr-xr-x - yarn supergroup 0 2014-09-04 21:57 /workspace/mahout/week4/data/20news/20news-bydate-train
步骤二：对数据创建序列文件
[email protected]:~/Mahout/mahout-distribution-0.7/bin$ ./mahout seqdirectory -i /workspace/mahout/week4/data/20news -o /workspace/mahout/week4/data/20news_seq
[email protected]:~/Mahout/mahout-distribution-0.7/bin$ hadoop fs -ls /workspace/mahout/week4/data/20news_seq
Found 1 items
-rw-r--r-- 1 yarn supergroup 37064977 2014-09-04 22:12 /workspace/mahout/week4/data/20news_seq/chunk-0
第三步：将序列文件转化成向量
[email protected]:~/Mahout/mahout-distribution-0.7/bin$ ./mahout seq2sparse -i /workspace/mahout/week4/data/20news_seq/ -o /workspace/mahout/week4/data/20news_vectors -lnorm
-nv -wt tfidf
[email protected]:~/Mahout/mahout-distribution-0.7/bin$ hadoop fs -ls /workspace/mahout/week4/data/20news_vectors
Found 7 items
drwxr-xr-x - yarn supergroup 0 2014-09-04 22:20 /workspace/mahout/week4/data/20news_vectors/df-count
-rw-r--r-- 1 yarn supergroup 1937084 2014-09-04 22:18 /workspace/mahout/week4/data/20news_vectors/dictionary.file-0
-rw-r--r-- 1 yarn supergroup 1890053 2014-09-04 22:20 /workspace/mahout/week4/data/20news_vectors/frequency.file-0
drwxr-xr-x - yarn supergroup 0 2014-09-04 22:19 /workspace/mahout/week4/data/20news_vectors/tf-vectors
drwxr-xr-x - yarn supergroup 0 2014-09-04 22:21 /workspace/mahout/week4/data/20news_vectors/tfidf-vectors
drwxr-xr-x - yarn supergroup 0 2014-09-04 22:18 /workspace/mahout/week4/data/20news_vectors/tokenized-documents
drwxr-xr-x - yarn supergroup 0 2014-09-04 22:18 /workspace/mahout/week4/data/20news_vectors/wordcount
第四步：将向量集分为训练集和测试数据
参数：
- -tr训练集
- -te测试集
- -rp参数设定的是测试数据集占总数据集的百分比，以下代码设定为20%！
[email protected]:~/Mahout/mahout-distribution-0.7/bin$ ./mahout split -i /workspace/mahout/week4/data/20news_vectors/tfidf-vectors -tr /workspace/mahout/week4/data/train-vectors -te /workspace/mahout/week4/data/test-vectors -rp
20 -ow -seq -xm sequential
第五步：训练模型
[email protected]:~/Mahout/mahout-distribution-0.9/bin$ ./mahout trainnb -i /workspace/mahout/week4/data/train-vectors -el
-o /workspace/mahout/week4/nbmodel -li /workspace/mahout/week4/labindex -ow -c
查看生成的索引：
[email protected]:~$ hadoop fs -text /workspace/mahout/week4/labindex
20news-bydate-test 0
20news-bydate-train 1
查看训练出来的模型：
[email protected]:~$ hadoop fs -ls /workspace/mahout/week4/nbmodel
Found 1 items
-rw-r--r-- 1 yarn supergroup 2437874 2014-09-05 23:09 /workspace/mahout/week4/nbmodel/naiveBayesModel.bin
第六步：测试
[email protected]:~/Mahout/mahout-distribution-0.9/bin$ ./mahout testnb -i /workspace/mahout/week4/data/test-vectors -m /workspace/mahout/week4/nbmodel -l /workspace/mahout/week4/labindex -ow
-o /workspace/mahout/week4/20news-test-result -c
注意：测试时的-i跟着的输入路径是第四步拆分出来的测试集。
测试结果：
14/09/05 23:18:09 INFO test.TestNaiveBayesDriver: Complementary Results:
=======================================================
Summary
-------------------------------------------------------
Correctly Classified Instances : 2887 74.9675%
Incorrectly Classified Instances : 964 25.0325%
Total Classified Instances : 3851
=======================================================
Confusion Matrix
-------------------------------------------------------
a b <--Classified as
1131 413 | 1544 a = 20news-bydate-test
551 1756 | 2307 b = 20news-bydate-train
=======================================================
Statistics
-------------------------------------------------------
Kappa 0.486
Accuracy 74.9675%
Reliability 49.7892%
Reliability (standard deviation) 0.4314
14/09/05 23:18:09 INFO driver.MahoutDriver: Program took 17504 ms (Minutes: 0.29173333333333334)