spark | spark 机器学习chapter3 数据的获取、处理与准备

阅读spark机器学习这本书来学习在spark上做机器学习

注意：数据集是电影评分等数据，下载链接：http://files.grouplens.org/datasets/movielens/ml-100k.zip

数据集包括：用户属性文件、电影元素、用户对电影的评级
１、将数据解压到某个目录下，并切换到该目录

unzip ml-100k.zipcd ml-100k

２、查看上述三种数据

用户

电影

评分

３、启动python，分析数据

启动

/home/hadoop/spark/bin/pyspark

４、读数据

from pyspark import SparkContextuser_data = sc.textFile("u.user")user_data.first()

u’1|24|M|technician|85711’

５、基本的分析

＃分割数据，函数splituser_fields=user_data.map(lambda line:line.split("|"))#用户数量num_users=user_fields.map(lambda fields:fields[0]).count()＃性别数num_genders = user_fields.map(lambda fields:fields[2]).distinct().count()＃职业种数num_occupations = user_fields.map(lambda fields:fields[3]).distinct().count()＃其他num_zipcodes=user_fields.map(lambda fields:fields[4]).distinct().count()＃结果打印print "Users:%d ,genders:%d ,occupations:%d ,ZIP codes:%d" % (num_users,num_genders,num_occupations,num_zipcodes)

Users:943 ,genders:2 ,occupations:21 ,ZIP codes:795

6、画图
对ages这个属性做直方图。
由于在终端下没法画图，这里给出代码

ages = user_fields.map(lambda x: int(x[1])).collect()hist(ages, bins=20, color='lightblue', normed=True)fig = matplotlib.pyplot.gcf()fig.set_size_inches(16, 10)

７、统计职业的种类和数量

import numpy as npcount_by_occupation = user_fields.map(lambda fields:(fields[3],1)).reduceByKey(lambda x,y:x+y).collect()x_axis1 = np.array([c[0] for c in count_by_occupation])y_axis1 = np.array([c[1] for c in count_by_occupation])

打印出结果

print x_axis1

[u’administrator’ u’retired’ u’lawyer’ u’none’ u’student’ u’technician’
u’programmer’ u’salesman’ u’homemaker’ u’executive’ u’doctor’
u’entertainment’ u’marketing’ u’writer’ u’scientist’ u’educator’
u’healthcare’ u’librarian’ u’artist’ u’other’ u’engineer’]

print y_axis1

[ 79 14 12 9 196 27 66 12 7 32 7 18 26 45 31 95 16 51
28 105 67]

y_axis = y_axis1[np.argsort(y_axis1)]

array([ 7, 7, 9, 12, 12, 14, 16, 18, 26, 27, 28, 31, 32,
45, 51, 66, 67, 79, 95, 105, 196])

np.argsort() : 得到升序的下标

画图：

pos = np.arange(len(x_axis))width = 1.0ax = plt.axes()ax.set_xticks(pos + (width / 2))ax.set_xticklabels(x_axis)plt.bar(pos, y_axis, width, color='lightblue')plt.xticks(rotation=30)fig = matplotlib.pyplot.gcf()fig.set_size_inches(16, 10)

计算各个值出现不同次数的方法：

count_by_occupation2 = user_fields.map(lambda fields: fields[3]).countByValue()print "Map-reduce approach:"print dict(count_by_occupation2)print ""print "countByValue approach:"print dict(count_by_occupation)

Map-reduce approach
print(dict(count_by_occupation2))
{u’administrator’: 79, u’retired’: 14, u’lawyer’: 12, u’healthcare’: 16, u’marketing’: 26, u’executive’: 32, u’scientist’: 31, u’student’: 196, u’technician’: 27, u’librarian’: 51, u’programmer’: 66, u’salesman’: 12, u’homemaker’: 7, u’engineer’: 67, u’none’: 9, u’doctor’: 7, u’writer’: 45, u’entertainment’: 18, u’other’: 105, u’educator’: 95, u’artist’: 28}

countByValue approach
{u’administrator’: 79, u’writer’: 45, u’retired’: 14, u’lawyer’: 12, u’doctor’: 7, u’marketing’: 26, u’executive’: 32, u’none’: 9, u’entertainment’: 18, u’healthcare’: 16, u’scientist’: 31, u’student’: 196, u’educator’: 95, u’technician’: 27, u’librarian’: 51, u’programmer’: 66, u’artist’: 28, u’salesman’: 12, u’other’: 105, u’homemaker’: 7, u’engineer’: 67}

探索电影数据
解析电影分类数据的特征

读数据和查看数据

读数据

movie_data = sc.textFile("u.item")

查看数据

＃第一行print movie_data.first()

1|Toy Story (1995)|01-Jan-1995||http://us.imdb.com/M/title-exact?Toy%20Story%20(1995)|0|0|0|1|1|1|0|0|0|0|0|0|0|0|0|0|0|0|0

电影总数

print "Movies:%d" % num_movies

Movies:1682

对电影发型的时间做处理
先过虑掉缺失值，定义函数，缺失值取为1900

def convert_year(x):　　try:　　　　return int(x[-4:])　　except:　　　　return 1900 

第３列为时间，格式为：01-Jan-1995　，－４：得到年

数据处理

movie_fields = movie_data.map(lambda lines: lines.split("|"))years = movie_fields.map(lambda fields: fields[2]).map(lambda x: convert_year(x))

过滤掉为1900的

years_filtered = years.filter(lambda x: x != 1900)

计算电影的年龄，该数据发生在1998年，要得到发行时间，需要1998减去时间

movie_ages = years_filtered.map(lambda yr: 1998-yr).countByValue()values = movie_ages.values()bins = movie_ages.keys()print valuesprint bins

[65, 286, 355, 219, 214, 126, 37, 22, 24, 15, 11, 13, 15, 7, 8, 5, 13, 12, 8, 9, 4, 4, 5, 6, 8, 4, 3, 7, 3, 4, 6, 5, 2, 5, 2, 6, 5, 3, 5, 4, 9, 8, 4, 5, 7, 2, 3, 5, 7, 4, 3, 5, 5, 4, 5, 4, 2, 5, 8, 7, 3, 4, 2, 4, 4, 2, 1, 1, 1, 1, 1]

[0, 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, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 72, 76]

画图

hist(values, bins=bins, color='lightblue', normed=True)fig = matplotlib.pyplot.gcf()fig.set_size_inches(16,10)

探索评级数据

读数据、数据量

>>> #评级数据... rating_data = sc.textFile("u.data")print rating_data.first()196 242 3   881250949num_ratings = rating_data.count()print "Ratings: %d" % num_ratingsRatings: 100000

196 242 3 881250949
Ratings: 100000
总共10万条数据

一些基本统计

#数据分割rating_data1 = rating_data.map(lambda line:line.split("\t"))#评分ratings = rating_data1.map(lambda fields:int(fields[2]))#最高得分max_rating = ratings.reduce(lambda x,y:max(x,y))#最低得分min_rating = ratings.reduce(lambda x,y:min(x,y))#评价得分mean_rating = ratings.reduce(lambda x,y:x+y) / num_ratings#中位数median_rating = np.median(ratings.collect())#平均每个用户打分数ratings_per_user = num_ratings / num_users＃平均每部电影有多少评分ratings_per_movie = num_ratings / num_moviesprint "Min ratings: %d" % min_ratingprint "Max rating: %d" % max_ratingprint "Average rating: %2.2f" % mean_ratingprint "Median rating: %d" % median_ratingprint "Average # of rating per user:%2.2f" % ratings_per_userprint "Average # of ratings per movie: %2.2f" % ratings_per_movie

Min ratings: 1
Max rating: 5
Average rating: 3.00
Median rating: 4
Average # of rating per user:106.00
Average # of ratings per movie: 59.00

类似功能的函数stats()

ratings.stats()

(count: 100000, mean: 3.52986, stdev: 1.12566797076, max: 5.0, min: 1.0)

计算每个用户打分次数：

user_ratings_grouped = rating_data.map(lambda fields:(int(fields[0]),int(fields[2]))).groupByKey()user_ratings_byuser = user_ratings_grouped.map(lambda (k,v):(k,int(len(v))))user_ratings_byuser.take(5)　　＃这里在spark2.1下报错，后续探究

绘图

user_ratings_byuser_local = user_ratings_byuser.map(lambda (k, v): v).collect()hist(user_ratings_byuser_local, bins=200, color='lightblue', normed=True)fig = matplotlib.pyplot.gcf()fig.set_size_inches(16,10)