{"id":320,"date":"2012-12-13T11:42:50","date_gmt":"2012-12-13T02:42:50","guid":{"rendered":"https:\/\/blog.id774.net\/entry\/?p=320"},"modified":"2012-12-13T12:35:45","modified_gmt":"2012-12-13T03:35:45","slug":"%e6%96%87%e6%9b%b8%e3%82%92%e3%82%af%e3%83%a9%e3%82%b9%e3%82%bf%e3%83%aa%e3%83%b3%e3%82%b0%e3%81%99%e3%82%8b","status":"publish","type":"post","link":"https:\/\/blog.id774.net\/entry\/2012\/12\/13\/320\/","title":{"rendered":"\u6587\u66f8\u3092\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0\u3059\u308b"},"content":{"rendered":"

\u4eba\u9593\u3001\u751f\u304d\u3066\u308b\u3068\u3044\u308d\u3093\u306a\u60c5\u5831\u3068\u51fa\u4f1a\u3044\u3001\u305d\u308c\u3089\u3092\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0\u3057\u305f\u304f\u306a\u308b\u306e\u304c\u3053\u306e\u4e16\u306e\u5e38\u3067\u3042\u308b\u3002\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u306f\u4e00\u304b\u3089\u81ea\u5206\u3067\u5b9f\u88c5\u3059\u308b\u3088\u308a\u306f\u3059\u3067\u306b\u5b9f\u7e3e\u306e\u3042\u308b\u3082\u306e\u3092\u5229\u7528\u3059\u308b\u306e\u304c\u826f\u3044\u3060\u308d\u3046\u3002\u307e\u305a\u306f Mahout \u3092\u4f7f\u3063\u3066\u307f\u308b\u3002<\/p>\n

\u30c6\u30ad\u30b9\u30c8\u30b3\u30fc\u30d1\u30b9\u306e\u53d6\u5f97<\/strong><\/p>\n

\u307e\u305a\u306f\u30c6\u30ad\u30b9\u30c8\u30b3\u30fc\u30d1\u30b9\u3092\u7528\u610f\u3059\u308b\u3002\u5b9f\u9a13\u306a\u306e\u3067\u7121\u96e3\u306b\u9752\u7a7a\u6587\u5eab\u3068\u304b\u3067\u3082\u3044\u3044\u306e\u3060\u3051\u3069\u3001\u3044\u307e\u3059\u3050\u5206\u6790\u3057\u305f\u3044\u6587\u66f8\u304c\u3042\u308b\u306a\u3089\u305d\u308c\u3092\u4f7f\u3048\u3070\u826f\u3044\u3002\u541b\u306f\u65e5\u9803\u53ce\u96c6\u3057\u3066\u3044\u308b\u30bd\u30fc\u30b7\u30e3\u30eb\u30cd\u30c3\u30c8\u30ef\u30fc\u30af\u306e\u60c5\u5831\u3092\u4f7f\u3063\u3066\u3082\u3044\u3044\u3057\u3001\u904b\u55b6\u3057\u3066\u3044\u308b\u30b5\u30fc\u30d3\u30b9\u306b\u5bc4\u305b\u3089\u308c\u305f\u30e6\u30fc\u30b6\u30fc\u306e\u58f0\u3092\u5229\u7528\u3057\u3066\u3082\u3044\u3044\u3002<\/p>\n

\u3061\u306a\u307f\u306b\u3044\u3044\u8a71\u5224\u5b9a\u5668<\/a>\u306b\u683c\u7d0d\u3055\u308c\u305f\u3044\u3044\u8a71<\/a>\u3092\u4f7f\u3046\u306a\u3093\u3066\u306e\u3082\u4e00\u8208\u3067\u3042\u308b\u3002<\/p>\n

\u5f62\u614b\u7d20\u89e3\u6790<\/strong><\/p>\n

\u6211\u3005\u306f\u65e5\u672c\u4eba\u3067\u3042\u308b\u3002\u65e5\u672c\u8a9e\u6587\u66f8\u306f\u82f1\u8a9e\u306e\u3088\u3046\u306b\u5358\u8a9e\u3069\u3046\u3057\u304c\u7a7a\u767d\u3067\u533a\u5207\u3089\u308c\u3066\u3044\u306a\u3044\u3002\u305d\u3053\u3067\u5f62\u614b\u7d20\u89e3\u6790\u30a8\u30f3\u30b8\u30f3\u3092\u5229\u7528\u3059\u308b\u5fc5\u8981\u304c\u3042\u308b\u3057\u3001\u3053\u3053\u3067\u306f\u3044\u308f\u3086\u308b\u5206\u304b\u3061\u66f8\u304d\u3092\u3059\u308b\u3002<\/p>\n

\u30b7\u30a7\u30eb\u30b9\u30af\u30ea\u30d7\u30c8\u3067\u306e\u5358\u7d14\u306a\u4f8b\u3002<\/p>\n

1
2
3
4
<\/div><\/td>
for file in `find -type f -name "*.txt"`
\ndo
\n  mecab -O wakati -o $file.wakati.txt $file
\ndone<\/div><\/td><\/tr><\/tbody><\/table><\/div>\n

MeCab \u306f Ruby \u7b49\u306e\u5404\u7a2e\u8a00\u8a9e\u306e\u30d0\u30a4\u30f3\u30c7\u30a3\u30f3\u30b0\u306a\u3069\u3082\u3042\u308b\u306e\u3067\u305d\u308c\u3089\u3092\u5229\u7528\u3057\u305f\u308a MapReduce \u3067\u3084\u3063\u3066\u3082\u3044\u3044\u3002<\/p>\n

\u30b7\u30fc\u30b1\u30f3\u30b9\u3068\u30d9\u30af\u30c8\u30eb\u306e\u751f\u6210<\/strong><\/p>\n

\u5206\u304b\u3061\u66f8\u304d\u3092\u3057\u305f\u30d5\u30a1\u30a4\u30eb\u3092 HDFS \u306b\u30a2\u30c3\u30d7\u30ed\u30fc\u30c9\u3057\u3001\u30b7\u30fc\u30b1\u30f3\u30b9\u3068\u30d9\u30af\u30c8\u30eb\u3092\u751f\u6210\u3059\u308b\u3002<\/p>\n

1
2
3
4
5
6
7
8
9
10
11
<\/div><\/td>
hadoop fs -put corpus_examples corpus_examples
\nmahout seqdirectory -i corpus_examples -o corpus_seq
\nmahout seq2sparse -i corpus_seq -o corpus_sparse -a org.apache.lucene.analysis.WhitespaceAnalyzer
\n
\n# \u4e2d\u8eab\u306e\u30c1\u30a7\u30c3\u30af
\nhadoop fs -lsr corpus_seq
\nhadoop fs -lsr corpus_sparse
\n
\n# \u983b\u51fa\u8a9e\u53e5\u306e\u30c1\u30a7\u30c3\u30af
\nmahout seqdumper -s corpus_sparse\/wordcount\/part-r-00000 -o wordcount
\nsort -nrk4 -t: wordcount | head<\/div><\/td><\/tr><\/tbody><\/table><\/div>\n

\u3053\u308c\u3067 Mahout \u3092\u4f7f\u3046\u6e96\u5099\u306f\u3067\u304d\u305f\u3002<\/p>\n

\u30ad\u30e3\u30ce\u30d4\u30fc\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0<\/strong><\/p>\n

\u5358\u7d14\u306b K \u5e73\u5747\u6cd5\u3067\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0\u3057\u3066\u3082\u3044\u3044\u3051\u3069\u30af\u30e9\u30b9\u30bf\u306e\u6570\u3092\u4e8b\u524d\u306b\u898b\u7a4d\u3082\u308c\u306a\u3044\u3088\u306d\u3001\u3068\u3044\u3046\u3053\u3068\u3067 TF-IDF \u3067\u5358\u8a9e\u3054\u3068\u306e\u91cd\u307f\u4ed8\u3051\u3092\u3057\u305f\u30d9\u30af\u30c8\u30eb\u3092\u5229\u7528\u3057\u30ad\u30e3\u30ce\u30d4\u30fc\u91cd\u5fc3\u3092\u6c42\u3081\u308b\u3002<\/p>\n

1
<\/div><\/td>
mahout canopy -i corpus_sparse\/tfidf-vectors -o corpus_canopy -t1 0.4 -t2 0.6 -dm org.apache.mahout.common.distance.CosineDistanceMeasure -ow<\/div><\/td><\/tr><\/tbody><\/table><\/div>\n

\u6b21\u306b\u30ad\u30e3\u30ce\u30d4\u30fc\u91cd\u5fc3\u3092\u5229\u7528\u3057\u3066 K \u5e73\u5747\u6cd5\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0\u3002<\/p>\n

1
2
3
<\/div><\/td>
mahout kmeans -i corpus_sparse\/tfidf-vectors -c corpus_canopy\/clusters-0\/part-r-00000 -o corpus_kmeans --maxIter 10 -cl -ow
\n# \u30af\u30e9\u30b9\u30bf\u306e\u30c0\u30f3\u30d7\u3092\u53d6\u5f97\u3057\u5185\u5bb9\u3092\u7cbe\u67fb
\nmahout clusterdump -d corpus_sparse\/dictionary.file-0 -dt sequencefile -s corpus_kmeans\/clusters-1 -p corpus_kmeans\/clusterdPoints -o dump<\/div><\/td><\/tr><\/tbody><\/table><\/div>\n

\u3042\u3068\u306f dump \u306e\u4e2d\u8eab\u3092\u898b\u305f\u308a\u30af\u30e9\u30b9\u30bf\u9593\u53ca\u3073\u30af\u30e9\u30b9\u30bf\u5185\u90e8\u70b9\u9593\u306e\u8ddd\u96e2\u3092\u8a08\u6e2c\u3057\u305f\u308a\u3057\u3066\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0\u306e\u8cea\u3092\u5411\u4e0a\u3059\u308c\u3070\u826f\u3044\u3002\u30b3\u30b5\u30a4\u30f3\u8ddd\u96e2\u4ee5\u5916\u306e\u985e\u4f3c\u6027\u6307\u6a19\u3092\u4f7f\u3063\u3066\u3082\u826f\u3044\u3002\u3053\u3046\u3044\u3046\u306e\u306f\u30e9\u30a4\u30d6\u30e9\u30ea\u304c\u81ea\u52d5\u3067\u3084\u3063\u3066\u304f\u308c\u308b\u308f\u3051\u3067\u306f\u306a\u3044\u306e\u3067\u81ea\u529b\u3067\u3072\u305f\u3059\u3089\u6570\u5b57\u3068\u306b\u3089\u3081\u3063\u3053\u3059\u308b\u3057\u304b\u306a\u3055\u305d\u3046\u3067\u3059\u306d\u3002<\/p>\n

Ruby \u3067\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0\u3059\u308b<\/strong><\/p>\n

Mahout \u306f\u74b0\u5883\u3092\u7528\u610f\u3057\u305f\u308a\u4f7f\u3046\u307e\u3067\u306e\u4e8b\u524d\u6e96\u5099\u304c\u610f\u5916\u3068\u3060\u308b\u3044\u3002\u305d\u3053\u3067\u3082\u3063\u3068\u30ab\u30b8\u30e5\u30a2\u30eb\u306b Ruby \u3067\u3084\u3063\u3066\u307f\u308b\u3002<\/p>\n

Ruby \u306e\u4e16\u754c\u306b\u306f\u4ee5\u524d\u306b\u3082\u66f8\u3044\u305f\u901a\u308a<\/a> AI4R<\/a> \u3084 K-Means<\/a> \u3001\u305d\u308c\u306b\u62d9\u4f5c\u306e kmeans<\/a> \u3068\u3044\u3063\u305f\u30e9\u30a4\u30d6\u30e9\u30ea\u304c\u3042\u308b\u3002<\/p>\n

\u4e0a\u8a18\u306e\u3046\u3061 K-Means \u3068\u3044\u3046\u306e\u304c\u9ad8\u901f\u304b\u3064\u4fbf\u5229\u3067\u3001\u5185\u90e8\u7684\u306b Distance Measures<\/a> \u30e9\u30a4\u30d6\u30e9\u30ea\u3092\u547c\u3093\u3067\u3044\u308b\u3002\u985e\u4f3c\u6027\u6307\u6a19\u3068\u3057\u3066\u30b3\u30b5\u30a4\u30f3\u8ddd\u96e2\u3084\u30e6\u30fc\u30af\u30ea\u30c3\u30c9\u8ddd\u96e2\u306a\u3069\u81ea\u7531\u306b\u6307\u5b9a\u3059\u308b\u3053\u3068\u304c\u3067\u304d\u308b\u3002\u3061\u306a\u307f\u306b\u62d9\u4f5c\u306e kmeans \u3067\u306f\u96c6\u5408\u9593\u306e\u8ddd\u96e2\u3092\u30d4\u30a2\u30bd\u30f3\u76f8\u95a2\u4fc2\u6570\u3067\u6c42\u3081\u3066\u3044\u308b\u3002<\/p>\n

\u516c\u5f0f\u306e\u4f8b\u306b\u3057\u305f\u304c\u3063\u3066\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u3092\u53d6\u3063\u3066\u307f\u308b\u3002<\/p>\n

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
<\/div><\/td>
require 'rubygems'
\nrequire 'benchmark'
\nrequire 'k_means'
\nrequire 'ai4r'
\n
\n# 1000 records with 50 dimensions
\ndata = Array.new(1000) {Array.new(50) {rand(10)}}
\nai4r_data = Ai4r::Data::DataSet.new(:data_items=> data)
\n
\n# Clustering can happen in magical ways
\n# so lets do it over multiple times
\nn = 5
\n
\nBenchmark.bm do |x|
\n  x.report('euclidean') do
\n    n.times { KMeans.new(data, :distance_measure => :euclidean_distance) }
\n  end
\n  x.report('cosine') do
\n    n.times { KMeans.new(data, :distance_measure => :cosine_similarity) }
\n  end
\n  x.report('j_index') do
\n    n.times { KMeans.new(data, :distance_measure => :jaccard_index) }
\n  end
\n  x.report('j_distance') do
\n    n.times { KMeans.new(data, :distance_measure => :jaccard_distance) }
\n  end
\n  x.report('b_j_index') do
\n    n.times { KMeans.new(data, :distance_measure => :binary_jaccard_index) }
\n  end
\n  x.report('b_j_distance') do
\n    n.times { KMeans.new(data, :distance_measure => :binary_jaccard_distance) }
\n  end
\n  x.report('tanimoto') do
\n    n.times { KMeans.new(data, :distance_measure => :tanimoto_coefficient) }
\n  end
\n  x.report("Ai4R") do
\n    n.times do
\n      b = Ai4r::Clusterers::KMeans.new
\n      b.build(ai4r_data, 4)
\n    end
\n  end
\nend<\/div><\/td><\/tr><\/tbody><\/table><\/div>\n

\u3067\u3001\u3053\u308c\u3070\u30d9\u30f3\u30c1\u30de\u30fc\u30af\u7d50\u679c\u3002\u985e\u4f3c\u5ea6\u306f\u4f55\u3092\u4f7f\u3063\u3066\u3082\u3044\u3044\u3051\u3069\u30a6\u30fc\u30e0 AI4R \u9045\u3059\u304e\u3060\u308d\u3068\u3044\u3046\u611f\u3058\u3067\u3059\u306d\u3002<\/p>\n

1
2
3
4
5
6
7
8
9
<\/div><\/td>
               user     system      total        real
\neuclidean     13.250000   0.090000  13.340000 ( 15.467818)
\ncosine         4.780000   0.040000   4.820000 (  5.913864)
\nj_index        7.940000   0.080000   8.020000 (  9.106817)
\nj_distance     8.400000   0.080000   8.480000 (  9.889458)
\nb_j_index      8.040000   0.050000   8.090000 (  9.884315)
\nb_j_distance   5.030000   0.060000   5.090000 (  5.860338)
\ntanimoto       4.630000   0.020000   4.650000 (  5.422322)
\nAi4R         121.700000   1.140000 122.840000 (159.453762)<\/div><\/td><\/tr><\/tbody><\/table><\/div>\n

\u307e\u3068\u307e\u3063\u3066\u3044\u306a\u3044\u3051\u3069\u4ee5\u4e0a\u3067\u3059\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"

\u4eba\u9593\u3001\u751f\u304d\u3066\u308b\u3068\u3044\u308d\u3093\u306a\u60c5\u5831\u3068\u51fa\u4f1a\u3044\u3001\u305d\u308c\u3089\u3092\u30af\u30e9\u30b9\u30bf\u30ea\u30f3\u30b0\u3057\u305f\u304f\u306a\u308b\u306e\u304c\u3053\u306e\u4e16\u306e\u5e38\u3067\u3042\u308b\u3002\u6a5f\u68b0\u5b66\u7fd2\u30e9\u30a4\u30d6\u30e9\u30ea\u306f\u4e00\u304b\u3089\u81ea\u5206\u3067\u5b9f\u88c5\u3059\u308b\u3088\u308a\u306f\u3059\u3067\u306b\u5b9f\u7e3e\u306e\u3042\u308b\u3082\u306e\u3092\u5229\u7528\u3059\u308b\u306e\u304c\u826f\u3044\u3060\u308d\u3046\u3002\u307e\u305a\u306f Mahout \u3092\u4f7f\u3063\u3066\u307f\u308b\u3002 … \u7d9a\u304d\u3092\u8aad\u3080<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[14],"tags":[],"class_list":["post-320","post","type-post","status-publish","format-standard","hentry","category-tech"],"_links":{"self":[{"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/posts\/320","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/comments?post=320"}],"version-history":[{"count":0,"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/posts\/320\/revisions"}],"wp:attachment":[{"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/media?parent=320"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/categories?post=320"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blog.id774.net\/entry\/wp-json\/wp\/v2\/tags?post=320"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}