Spark机器学习（Chapter 04）推荐系统

代码：

#coding:utf-8
from pyspark import SparkContext

sc = SparkContext("local[4]", "MovieLens Spark App")
rawData=sc.textFile("../02/ml-100k/u.data")
print rawData.first()
rawRating=rawData.map(lambda x:x.split('\t'))
print rawRating.take(5)

#导入Rating、ALS模块，用于后面建模
from pyspark.mllib.recommendation import Rating,ALS
ratings=rawRating.map(lambda x:Rating(int(x[0]),int(x[1]),int(x[2])))
print ratings.take(5)

#模型训练
model=ALS.train(ratings,50,10,0.01)
userFeatures=model.userFeatures()
print userFeatures.take(2)

print model.userFeatures().count()
print model.productFeatures().count()

#进行预测
print len(userFeatures.first()[1])
predictRating = model.predict(789,123)
print predictRating

#要为某个用户生成前K个推荐物品,可借助MatrixFactorizationModel所提供的recommendProducts函数来实现。该函数需两个输入参数:user和num。其中user
#是用户ID,而num是要推荐的物品个数。
topRecs=model.recommendProducts(789,10)  #给789用户推荐10部电影
print '给用户userId推荐其喜欢的item：'
for rec in topRecs:
	print rec

#检验推荐内容要直观地检验推荐的效果,可以简单比对下用户所评级过的电影和被推荐的那些电影
moviesForUser=ratings.groupBy(lambda x:x.user).mapValues(list).lookup(789)
print '用户对%d部电影进行了评级'%len(moviesForUser[0])
print '源数据中用户(userId=789)喜欢的电影(item)：'
for i in sorted(moviesForUser[0],key=lambda x: x.rating,reverse=True):
	print i.product

movies=sc.textFile("../02/ml-100k/u.item")
titles=movies.map(lambda line: (int(line.split