List of usage examples for org.apache.mahout.math.jet.random.sampling RandomSampler sample
public static void sample(long n, long N, int count, long low, long[] values, int fromIndex, Random randomGenerator)
From source file:com.netease.news.utils.SplitInput.java
License:Apache License
/** * Perform a split on the specified input file. Results will be written to files of the same name in the specified * training and test output directories. The {@link #validate()} method is called prior to executing the split. *///from w w w . jav a 2 s . c o m public void splitFile(Path inputFile) throws IOException { Configuration conf = getConf(); FileSystem fs = inputFile.getFileSystem(conf); if (fs.getFileStatus(inputFile) == null) { throw new IOException(inputFile + " does not exist"); } if (fs.getFileStatus(inputFile).isDir()) { throw new IOException(inputFile + " is a directory"); } validate(); Path testOutputFile = new Path(testOutputDirectory, inputFile.getName()); Path trainingOutputFile = new Path(trainingOutputDirectory, inputFile.getName()); int lineCount = countLines(fs, inputFile, charset); log.info("{} has {} lines", inputFile.getName(), lineCount); int testSplitStart = 0; int testSplitSize = this.testSplitSize; // don't modify state BitSet randomSel = null; if (testRandomSelectionPct > 0 || testRandomSelectionSize > 0) { testSplitSize = this.testRandomSelectionSize; if (testRandomSelectionPct > 0) { testSplitSize = Math.round(lineCount * testRandomSelectionPct / 100.0f); } log.info("{} test split size is {} based on random selection percentage {}", inputFile.getName(), testSplitSize, testRandomSelectionPct); long[] ridx = new long[testSplitSize]; RandomSampler.sample(testSplitSize, lineCount - 1, testSplitSize, 0, ridx, 0, RandomUtils.getRandom()); randomSel = new BitSet(lineCount); for (long idx : ridx) { randomSel.set((int) idx + 1); } } else { if (testSplitPct > 0) { // calculate split size based on percentage testSplitSize = Math.round(lineCount * testSplitPct / 100.0f); log.info("{} test split size is {} based on percentage {}", inputFile.getName(), testSplitSize, testSplitPct); } else { log.info("{} test split size is {}", inputFile.getName(), testSplitSize); } if (splitLocation > 0) { // calculate start of split based on percentage testSplitStart = Math.round(lineCount * splitLocation / 100.0f); if (lineCount - testSplitStart < testSplitSize) { // adjust split start downwards based on split size. testSplitStart = lineCount - testSplitSize; } log.info("{} test split start is {} based on split location {}", inputFile.getName(), testSplitStart, splitLocation); } if (testSplitStart < 0) { throw new IllegalArgumentException( "test split size for " + inputFile + " is too large, it would produce an " + "empty training set from the initial set of " + lineCount + " examples"); } else if (lineCount - testSplitSize < testSplitSize) { log.warn( "Test set size for {} may be too large, {} is larger than the number of " + "lines remaining in the training set: {}", inputFile, testSplitSize, lineCount - testSplitSize); } } int trainCount = 0; int testCount = 0; if (!useSequence) { BufferedReader reader = new BufferedReader(new InputStreamReader(fs.open(inputFile), charset)); Writer trainingWriter = new OutputStreamWriter(fs.create(trainingOutputFile), charset); Writer testWriter = new OutputStreamWriter(fs.create(testOutputFile), charset); try { String line; int pos = 0; while ((line = reader.readLine()) != null) { pos++; Writer writer; if (testRandomSelectionPct > 0) { // Randomly choose writer = randomSel.get(pos) ? testWriter : trainingWriter; } else { // Choose based on location writer = pos > testSplitStart ? testWriter : trainingWriter; } if (writer == testWriter) { if (testCount >= testSplitSize) { writer = trainingWriter; } else { testCount++; } } if (writer == trainingWriter) { trainCount++; } writer.write(line); writer.write('\n'); } } finally { Closeables.close(reader, true); Closeables.close(trainingWriter, false); Closeables.close(testWriter, false); } } else { SequenceFileIterator<Writable, Writable> iterator = new SequenceFileIterator<Writable, Writable>( inputFile, false, fs.getConf()); SequenceFile.Writer trainingWriter = SequenceFile.createWriter(fs, fs.getConf(), trainingOutputFile, iterator.getKeyClass(), iterator.getValueClass()); SequenceFile.Writer testWriter = SequenceFile.createWriter(fs, fs.getConf(), testOutputFile, iterator.getKeyClass(), iterator.getValueClass()); try { int pos = 0; while (iterator.hasNext()) { pos++; SequenceFile.Writer writer; if (testRandomSelectionPct > 0) { // Randomly choose writer = randomSel.get(pos) ? testWriter : trainingWriter; } else { // Choose based on location writer = pos > testSplitStart ? testWriter : trainingWriter; } if (writer == testWriter) { if (testCount >= testSplitSize) { writer = trainingWriter; } else { testCount++; } } if (writer == trainingWriter) { trainCount++; } Pair<Writable, Writable> pair = iterator.next(); writer.append(pair.getFirst(), pair.getSecond()); } } finally { Closeables.close(iterator, true); Closeables.close(trainingWriter, false); Closeables.close(testWriter, false); } } log.info("file: {}, input: {} train: {}, test: {} starting at {}", inputFile.getName(), lineCount, trainCount, testCount, testSplitStart); // testing; if (callback != null) { callback.splitComplete(inputFile, lineCount, trainCount, testCount, testSplitStart); } }
From source file:com.tamingtext.util.SplitInput.java
License:Apache License
/** Perform a split on the specified input file. Results will be written to files of the same name in the specified * training and test output directories. The {@link #validate()} method is called prior to executing the split. *//*from w ww . j av a 2 s.c o m*/ public void splitFile(Path inputFile) throws IOException { if (fs.getFileStatus(inputFile) == null) { throw new IOException(inputFile + " does not exist"); } else if (fs.getFileStatus(inputFile).isDir()) { throw new IOException(inputFile + " is a directory"); } validate(); Path testOutputFile = new Path(testOutputDirectory, inputFile.getName()); Path trainingOutputFile = new Path(trainingOutputDirectory, inputFile.getName()); int lineCount = countLines(fs, inputFile, charset); log.info("{} has {} lines", inputFile.getName(), lineCount); int testSplitStart = 0; int testSplitSize = this.testSplitSize; // don't modify state BitSet randomSel = null; if (testRandomSelectionPct > 0 || testRandomSelectionSize > 0) { testSplitSize = this.testRandomSelectionSize; if (testRandomSelectionPct > 0) { testSplitSize = Math.round(lineCount * (testRandomSelectionPct / 100.0f)); } log.info("{} test split size is {} based on random selection percentage {}", new Object[] { inputFile.getName(), testSplitSize, testRandomSelectionPct }); long[] ridx = new long[testSplitSize]; RandomSampler.sample(testSplitSize, lineCount - 1, testSplitSize, 0, ridx, 0, RandomUtils.getRandom()); randomSel = new BitSet(lineCount); for (long idx : ridx) { randomSel.set((int) idx + 1); } } else { if (testSplitPct > 0) { // calculate split size based on percentage testSplitSize = Math.round(lineCount * (testSplitPct / 100.0f)); log.info("{} test split size is {} based on percentage {}", new Object[] { inputFile.getName(), testSplitSize, testSplitPct }); } else { log.info("{} test split size is {}", inputFile.getName(), testSplitSize); } if (splitLocation > 0) { // calculate start of split based on percentage testSplitStart = Math.round(lineCount * (splitLocation / 100.0f)); if (lineCount - testSplitStart < testSplitSize) { // adjust split start downwards based on split size. testSplitStart = lineCount - testSplitSize; } log.info("{} test split start is {} based on split location {}", new Object[] { inputFile.getName(), testSplitStart, splitLocation }); } if (testSplitStart < 0) { throw new IllegalArgumentException( "test split size for " + inputFile + " is too large, it would produce an " + "empty training set from the initial set of " + lineCount + " examples"); } else if ((lineCount - testSplitSize) < testSplitSize) { log.warn( "Test set size for {} may be too large, {} is larger than the number of " + "lines remaining in the training set: {}", new Object[] { inputFile, testSplitSize, lineCount - testSplitSize }); } } BufferedReader reader = new BufferedReader(new InputStreamReader(fs.open(inputFile), charset)); Writer trainingWriter = new OutputStreamWriter(fs.create(trainingOutputFile), charset); Writer testWriter = new OutputStreamWriter(fs.create(testOutputFile), charset); int pos = 0; int trainCount = 0; int testCount = 0; String line; while ((line = reader.readLine()) != null) { pos++; Writer writer; if (testRandomSelectionPct > 0) { // Randomly choose writer = randomSel.get(pos) ? testWriter : trainingWriter; } else { // Choose based on location writer = pos > testSplitStart ? testWriter : trainingWriter; } if (writer == testWriter) { if (testCount >= testSplitSize) { writer = trainingWriter; } else { testCount++; } } if (writer == trainingWriter) { trainCount++; } writer.write(line); writer.write('\n'); } IOUtils.close(Collections.singleton(trainingWriter)); IOUtils.close(Collections.singleton(testWriter)); log.info("file: {}, input: {} train: {}, test: {} starting at {}", new Object[] { inputFile.getName(), lineCount, trainCount, testCount, testSplitStart }); // testing; if (callback != null) { callback.splitComplete(inputFile, lineCount, trainCount, testCount, testSplitStart); } }