mahout算法源码分析之Itembased Collaborative Filtering(一)PreparePreferenceMatrixJob
Mahout版本:0.7,hadoop版本:1.0.4,jdk:1.7.0_25 64bit。
本篇分析RecommenderJob的源码,这个类也是继承了AbstractJob,所以也会覆写其run方法,点开这个run方法,可以看到和其他的job类都一样,刚开始都是基本参数的默认值设置和获取;然后到了第一个job,在这个job之前有一个shouldRunNextPhase()函数,点开这个函数看到下面的源码:
protected static boolean shouldRunNextPhase(Map<String, List<String>> args, AtomicInteger currentPhase) { int phase = currentPhase.getAndIncrement(); String startPhase = getOption(args, "--startPhase"); String endPhase = getOption(args, "--endPhase"); boolean phaseSkipped = (startPhase != null && phase < Integer.parseInt(startPhase)) || (endPhase != null && phase > Integer.parseInt(endPhase)); if (phaseSkipped) { log.info("Skipping phase {}", phase); } return !phaseSkipped; }
其中phase是获取当前的phase值的,关于phase的相关概念可以参考: mahout中phase的含义,这里可以看到主要是根据phase和startPhase、endPhase的值做比较,然后返回true或者false,因为在实战中是按默认值的(startPhase和endPhase都没有设置),所以RecommenderJob中的这个函数都是返回true的。
看第一个job的调用:
if (shouldRunNextPhase(parsedArgs, currentPhase)) { ToolRunner.run(getConf(), new PreparePreferenceMatrixJob(), new String[]{ "--input", getInputPath().toString(), "--output", prepPath.toString(), "--maxPrefsPerUser", String.valueOf(maxPrefsPerUserInItemSimilarity), "--minPrefsPerUser", String.valueOf(minPrefsPerUser), "--booleanData", String.valueOf(booleanData), "--tempDir", getTempPath().toString()}); numberOfUsers = HadoopUtil.readInt(new Path(prepPath, PreparePreferenceMatrixJob.NUM_USERS), getConf()); }
这里看到调用的job主类是PreparePreferenceMatrixJob,然后这个job的输入参数有输入、出、maxPrefsPerUser、minPrefsPerUser、booleanData、tempDir。那么就打开主类PreparePreferenceMatrixJob,来看看。这个PreparePreferenceMatrixJob同样实现了AbstractJob类,那么直接看run方法吧。在run中的参数设置里有一个ratingShift,这个在调用的时候没有使用,所以按照默认,设置为0.0。大致浏览一下发现一共有三个prepareJob,所以这个主类会产生3个job。下面来一个个来看:
(1)//convert items to an internal index
Job itemIDIndex = prepareJob(getInputPath(), getOutputPath(ITEMID_INDEX), TextInputFormat.class, ItemIDIndexMapper.class, VarIntWritable.class, VarLongWritable.class, ItemIDIndexReducer.class, VarIntWritable.class, VarLongWritable.class, SequenceFileOutputFormat.class);
输入格式:userid,itemid,value
先看mapper:
protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException { String[] tokens = TasteHadoopUtils.splitPrefTokens(value.toString()); long itemID = Long.parseLong(tokens[transpose ? 0 : 1]); int index = TasteHadoopUtils.idToIndex(itemID); context.write(new VarIntWritable(index), new VarLongWritable(itemID)); }
在map中,首先获得itemID,在tokens中tokens[1]即是itemID了,至于当transpose为true的时候就要选择tokens[0]作为itemID这个应该是其他的应用吧,由于在调用的时候没有设置这个参数,所以这里按照默认值为false,所以选择tokens[1]作为itemID。然后看到index和itemID的转换使用的是TasteHadoopUtils.idToIndex()函数,看到这个函数返回的是return 0x7FFFFFFF & Longs.hashCode(id);所以当这个数在int可以表示的数范围内(小于2147483647)时候就会返回这个数本身了,比如实战中的项目101,返回的index也是101。
再看reducer:
protected void reduce(VarIntWritable index, Iterable<VarLongWritable> possibleItemIDs, Context context) throws IOException, InterruptedException { long minimumItemID = Long.MAX_VALUE; for (VarLongWritable varLongWritable : possibleItemIDs) { long itemID = varLongWritable.get(); if (itemID < minimumItemID) { minimumItemID = itemID; } } if (minimumItemID != Long.MAX_VALUE) { context.write(index, new VarLongWritable(minimumItemID)); } }
总感觉这里没啥必要,reducer返回的还是101-->101,或者这里应该有什么说法的?
输出文件是ITEMID_INDEX,输出格式<key,value> : VarintWritable-->VarLongWritable
所以这个job就分析完了。
(2)//convert user preferences into a vector per user
Job toUserVectors = prepareJob(getInputPath(), getOutputPath(USER_VECTORS), TextInputFormat.class, ToItemPrefsMapper.class, VarLongWritable.class, booleanData ? VarLongWritable.class : EntityPrefWritable.class, ToUserVectorsReducer.class, VarLongWritable.class, VectorWritable.class, SequenceFileOutputFormat.class);
输入格式:userid,itemid,value
看mapper:(ToItemPrefsMapper继承ToEntityPrefsMapper,而ToItemPrefsMapper是空的,所以看ToEntityPrefsMapper)
public void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException { String[] tokens = DELIMITER.split(value.toString()); long userID = Long.parseLong(tokens[0]); long itemID = Long.parseLong(tokens[1]); if (itemKey ^ transpose) { // If using items as keys, and not transposing items and users, then users are items! // Or if not using items as keys (users are, as usual), but transposing items and users, // then users are items! Confused? long temp = userID; userID = itemID; itemID = temp; } if (booleanData) { context.write(new VarLongWritable(userID), new VarLongWritable(itemID)); } else { float prefValue = tokens.length > 2 ? Float.parseFloat(tokens[2]) + ratingShift : 1.0f; context.write(new VarLongWritable(userID), new EntityPrefWritable(itemID, prefValue)); } }
这么些代码,最主要的就是最后两句了,一句是求评分值,但是这里的加上ratingShift不知道是干啥的?虽然ratingShift是0.0。最后输出就是userID-->[itemID,prefValue]
再看reducer:
protected void reduce(VarLongWritable userID, Iterable<VarLongWritable> itemPrefs, Context context) throws IOException, InterruptedException { Vector userVector = new RandomAccessSparseVector(Integer.MAX_VALUE, 100); for (VarLongWritable itemPref : itemPrefs) { int index = TasteHadoopUtils.idToIndex(itemPref.get()); float value = itemPref instanceof EntityPrefWritable ? ((EntityPrefWritable) itemPref).getPrefValue() : 1.0f; userVector.set(index, value); } if (userVector.getNumNondefaultElements() >= minPreferences) { VectorWritable vw = new VectorWritable(userVector); vw.setWritesLaxPrecision(true); context.getCounter(Counters.USERS).increment(1); context.write(userID, vw); } }
首先说下为啥mapper输出的value是EntityPrefWritable,但是这里的Iterable接收的时候使用的是VarLongWritable,因为前者继承后者。然后就是用户所有的评分都写入一个vecotr,使用itemid作为vector的下标,prefValue作为值;最后判断一下,如果vector含有的item个数大于或等于minPreference(这里看出这个参数的意义了吧)就输出,否则不输出。另外,就是设置了一个Counters.USERS计数器,用来统计用户的个数。
这个job的输出为:USER_VECTORS,格式为:<key,value> : userid-->vector[itemid:prefValue,itemid:prefValue,...]
随后代码获得了用户的个数:
int numberOfUsers = (int) toUserVectors.getCounters().findCounter(ToUserVectorsReducer.Counters.USERS).getValue(); HadoopUtil.writeInt(numberOfUsers, getOutputPath(NUM_USERS), getConf());
(3)//build the rating matrix
Job toItemVectors = prepareJob(getOutputPath(USER_VECTORS), getOutputPath(RATING_MATRIX), ToItemVectorsMapper.class, IntWritable.class, VectorWritable.class, ToItemVectorsReducer.class, IntWritable.class, VectorWritable.class);
输入是第二个job的输出,格式为:<key,value> : userid-->vector[itemid:prefValue,itemid:prefValue,...]
先看mapper:
protected void map(VarLongWritable rowIndex, VectorWritable vectorWritable, Context ctx) throws IOException, InterruptedException { Vector userRatings = vectorWritable.get(); int numElementsBeforeSampling = userRatings.getNumNondefaultElements(); userRatings = Vectors.maybeSample(userRatings, sampleSize); int numElementsAfterSampling = userRatings.getNumNondefaultElements(); int column = TasteHadoopUtils.idToIndex(rowIndex.get()); VectorWritable itemVector = new VectorWritable(new RandomAccessSparseVector(Integer.MAX_VALUE, 1)); itemVector.setWritesLaxPrecision(true); Iterator<Vector.Element> iterator = userRatings.iterateNonZero(); while (iterator.hasNext()) { Vector.Element elem = iterator.next(); itemVector.get().setQuick(column, elem.get()); ctx.write(new IntWritable(elem.index()), itemVector); } ctx.getCounter(Elements.USER_RATINGS_USED).increment(numElementsAfterSampling); ctx.getCounter(Elements.USER_RATINGS_NEGLECTED).increment(numElementsBeforeSampling - numElementsAfterSampling); }
其中的userRatings = Vectors.maybeSample(userRatings, sampleSize);函数,由于sampleSize没有设置,所以取到的数是Integer的最大值,那么maybeSample就会返回原始值,vector中的非默认项的个数肯定是小于Integer的最大值的:
public static Vector maybeSample(Vector original, int sampleSize) { if (original.getNumNondefaultElements() <= sampleSize) { return original; } Vector sample = original.like(); Iterator<Vector.Element> sampledElements = new FixedSizeSamplingIterator<Vector.Element>(sampleSize, original.iterateNonZero()); while (sampledElements.hasNext()) { Vector.Element elem = sampledElements.next(); sample.setQuick(elem.index(), elem.get()); } return sample; }
map函数中column就是userid,然后输出是elem.index()就是itemID,而itemVector.get().setQuick(column, elem.get())其实就是设置itemVecotor为[userID:prefValue]的格式,这样的话mapper输出就是 itemID-->vector[userID:prefValue];同时还有两个计数器,因为numElementsBeforeSampling - numElementsAfterSampling=0,所以计数器Elements.USER_RATINGS_NEGLECTED就一直是零。
再看reducer:
protected void reduce(IntWritable row, Iterable<VectorWritable> vectors, Context ctx) throws IOException, InterruptedException { VectorWritable vectorWritable = VectorWritable.merge(vectors.iterator()); vectorWritable.setWritesLaxPrecision(true); ctx.write(row, vectorWritable); }
merge函数就是把mapper的输出变换成下面的形式:itemID-->vector[userID:prefValue,userID:prefVlaue,...];
所以这个job的输出是:RATING_MATRIX,格式为:<key,value> : itemID-->vector[userID:prefValue,userID:prefVlaue,...];
额 ,好吧,那个sampleSize是有值的,而非默认的Integer的最大值:
if (hasOption("maxPrefsPerUser")) { int samplingSize = Integer.parseInt(getOption("maxPrefsPerUser")); toItemVectors.getConfiguration().setInt(ToItemVectorsMapper.SAMPLE_SIZE, samplingSize); }
这个值也是可以设置的,所以现在你知道maxPrefsPerUser的值的用处了。但是这个值的默认是100,实战总的item才7,所以numElementsBeforeSampling - numElementsAfterSampling=0不变。
好了,这个job也分析完了。
分享,成长,快乐
转载请注明blog地址:http://blog.csdn.net/fansy1990