推荐算法之矩阵分解实例

其他 2018-07-06 10:40:25 阅读次数: 0

矩阵分解的数据利用的上篇文章的数据，协同过滤

用到的知识

python的surprise
k折交叉验证
SVD
SVDpp
NMF

算法与结果可视化

# 可以使用上面提到的各种推荐系统算法
from surprise import SVD,SVDpp,NMF
from surprise import Dataset
from surprise import  print_perf
import os
from surprise import Reader, Dataset
from surprise.model_selection import cross_validate
from pandas import DataFrame 
import numpy as np
import pandas as pd

##################SVD_noBiased 
## 指定文件路径
file_path = os.path.expanduser('./python_data.txt')
## 指定文件格式\n",
reader = Reader(line_format='user item rating timestamp', sep=',')
## 从文件读取数据
data = Dataset.load_from_file(file_path, reader=reader)

# 在数据集上测试一下效果
#perf = evaluate(algo, data, measures=['RMSE', 'MAE'])
# Run 5-fold cross-validation and print results.
perf1 = cross_validate(SVD(n_factors=1,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf2 = cross_validate(SVD(n_factors=3,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf3 = cross_validate(SVD(n_factors=5,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf4 = cross_validate(SVD(n_factors=7,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf5 = cross_validate(SVD(n_factors=9,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf6 = cross_validate(SVD(n_factors=11,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf7 = cross_validate(SVD(n_factors=13,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf8 = cross_validate(SVD(n_factors=15,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf9 = cross_validate(SVD(n_factors=17,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf10= cross_validate(SVD(n_factors=19,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf11= cross_validate(SVD(n_factors=21,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf12= cross_validate(SVD(n_factors=23,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf13= cross_validate(SVD(n_factors=25,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf14= cross_validate(SVD(n_factors=27,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf15= cross_validate(SVD(n_factors=29,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)

perf_result=[]
for i in range(1,16):
  perf_result.append('perf'+ str(i)) 
MAE=[]
for perf in perf_result:
  MAE.append(round(np.mean(DataFrame(eval(perf)).iloc[:,1]),4))

RMSE=[]
for perf in perf_result:
  RMSE.append(round(np.mean(DataFrame(eval(perf)).iloc[:,2]),4))


FIT_TIME=[]
for perf in perf_result:
  FIT_TIME.append(round(np.mean(DataFrame(eval(perf)).iloc[:,0]),4))


TEST_TIME=[]
for perf in perf_result:
  TEST_TIME.append(round(np.mean(DataFrame(eval(perf)).iloc[:,3]),4))

MAE = DataFrame(MAE,columns=['MAE'])
RMSE = DataFrame(RMSE,columns=['RMSE'])
FIT_TIME = DataFrame(FIT_TIME,columns=['FIT_TIME'])
TEST_TIME = DataFrame(TEST_TIME,columns=['TEST_TIME'])
Factors = DataFrame([1,3,5,7,9,11,13,15,17,19,21,23,25,27,29],columns=['Factors'])

SVD_noBaised_result = pd.concat([Factors,MAE,RMSE,FIT_TIME,TEST_TIME],axis=1)
SVD_noBaised_result.to_csv('./result_data/SVD_noBaised_result.csv',header=True,encoding='utf-8')
##################SVD_biased
# Run 5-fold cross-validation and print results.
perf01 = cross_validate(SVD(n_factors=1,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf02 = cross_validate(SVD(n_factors=3,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf03 = cross_validate(SVD(n_factors=5,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf04 = cross_validate(SVD(n_factors=7,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf05 = cross_validate(SVD(n_factors=9,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf06 = cross_validate(SVD(n_factors=11,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf07 = cross_validate(SVD(n_factors=13,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf08 = cross_validate(SVD(n_factors=15,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf09 = cross_validate(SVD(n_factors=17,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf010= cross_validate(SVD(n_factors=19,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf011= cross_validate(SVD(n_factors=21,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf012= cross_validate(SVD(n_factors=23,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf013= cross_validate(SVD(n_factors=25,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf014= cross_validate(SVD(n_factors=27,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf015= cross_validate(SVD(n_factors=29,biased=True),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)

perf_result1=[]
for i in range(1,16):
  perf_result1.append('perf0'+ str(i)) 
MAE=[]
for perf in perf_result1:
  MAE.append(round(np.mean(DataFrame(eval(perf)).iloc[:,1]),4))

RMSE=[]
for perf in perf_result1:
  RMSE.append(round(np.mean(DataFrame(eval(perf)).iloc[:,2]),4))


FIT_TIME=[]
for perf in perf_result1:
  FIT_TIME.append(round(np.mean(DataFrame(eval(perf)).iloc[:,0]),4))


TEST_TIME=[]
for perf in perf_result1:
  TEST_TIME.append(round(np.mean(DataFrame(eval(perf)).iloc[:,3]),4))

MAE = DataFrame(MAE,columns=['MAE'])
RMSE = DataFrame(RMSE,columns=['RMSE'])
FIT_TIME = DataFrame(FIT_TIME,columns=['FIT_TIME'])
TEST_TIME = DataFrame(TEST_TIME,columns=['TEST_TIME'])
Factors = DataFrame([1,3,5,7,9,11,13,15,17,19,21,23,25,27,29],columns=['Factors'])

SVD_baised_result = pd.concat([Factors,MAE,RMSE,FIT_TIME,TEST_TIME],axis=1)
SVD_baised_result.to_csv('./result_data/SVD_baised_result.csv',header=True,encoding='utf-8')

##############SVD++

# Run 5-fold cross-validation and print results.
perf001 = cross_validate(SVDpp(n_factors=1),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf002 = cross_validate(SVDpp(n_factors=3),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf003 = cross_validate(SVDpp(n_factors=5),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf004 = cross_validate(SVDpp(n_factors=7),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf005 = cross_validate(SVDpp(n_factors=9),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf006 = cross_validate(SVDpp(n_factors=11),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf007 = cross_validate(SVDpp(n_factors=13),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf008 = cross_validate(SVDpp(n_factors=15),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf009 = cross_validate(SVDpp(n_factors=17),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0010= cross_validate(SVDpp(n_factors=19),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0011= cross_validate(SVDpp(n_factors=21),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0012= cross_validate(SVDpp(n_factors=23),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0013= cross_validate(SVDpp(n_factors=25),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0014= cross_validate(SVDpp(n_factors=27),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0015= cross_validate(SVDpp(n_factors=29),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)

perf_result2=[]
for i in range(1,16):
  perf_result2.append('perf00'+ str(i)) 
MAE=[]
for perf in perf_result2:
  MAE.append(round(np.mean(DataFrame(eval(perf)).iloc[:,1]),4))

RMSE=[]
for perf in perf_result2:
  RMSE.append(round(np.mean(DataFrame(eval(perf)).iloc[:,2]),4))


FIT_TIME=[]
for perf in perf_result2:
  FIT_TIME.append(round(np.mean(DataFrame(eval(perf)).iloc[:,0]),4))


TEST_TIME=[]
for perf in perf_result2:
  TEST_TIME.append(round(np.mean(DataFrame(eval(perf)).iloc[:,3]),4))

MAE = DataFrame(MAE,columns=['MAE'])
RMSE = DataFrame(RMSE,columns=['RMSE'])
FIT_TIME = DataFrame(FIT_TIME,columns=['FIT_TIME'])
TEST_TIME = DataFrame(TEST_TIME,columns=['TEST_TIME'])
Factors = DataFrame([1,3,5,7,9,11,13,15,17,19,21,23,25,27,29],columns=['Factors'])

SVDpp_result = pd.concat([Factors,MAE,RMSE,FIT_TIME,TEST_TIME],axis=1)
SVDpp_result.to_csv('./result_data/SVDpp_result.csv',header=True,encoding='utf-8')

######## NMF

# Run 5-fold cross-validation and print results.
perf0001 = cross_validate(NMF(n_factors=1,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0002 = cross_validate(NMF(n_factors=3,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0003 = cross_validate(NMF(n_factors=5,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0004 = cross_validate(NMF(n_factors=7,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0005 = cross_validate(NMF(n_factors=9,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0006 = cross_validate(NMF(n_factors=11,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0007 = cross_validate(NMF(n_factors=13,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0008 = cross_validate(NMF(n_factors=15,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf0009 = cross_validate(NMF(n_factors=17,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf00010= cross_validate(NMF(n_factors=19,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf00011= cross_validate(NMF(n_factors=21,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf00012= cross_validate(NMF(n_factors=23,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf00013= cross_validate(NMF(n_factors=25,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf00014= cross_validate(NMF(n_factors=27,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)
perf00015= cross_validate(NMF(n_factors=29,biased=False),data, measures=['RMSE', 'MAE'], cv=5, verbose=True)


perf_result3=[]
for i in range(1,16):
  perf_result3.append('perf000'+ str(i)) 
MAE=[]
for perf in perf_result3:
  MAE.append(round(np.mean(DataFrame(eval(perf)).iloc[:,1]),4))

RMSE=[]
for perf in perf_result3:
  RMSE.append(round(np.mean(DataFrame(eval(perf)).iloc[:,2]),4))


FIT_TIME=[]
for perf in perf_result3:
  FIT_TIME.append(round(np.mean(DataFrame(eval(perf)).iloc[:,0]),4))


TEST_TIME=[]
for perf in perf_result3:
  TEST_TIME.append(round(np.mean(DataFrame(eval(perf)).iloc[:,3]),4))

MAE = DataFrame(MAE,columns=['MAE'])
RMSE = DataFrame(RMSE,columns=['RMSE'])
FIT_TIME = DataFrame(FIT_TIME,columns=['FIT_TIME'])
TEST_TIME = DataFrame(TEST_TIME,columns=['TEST_TIME'])
Factors = DataFrame([1,3,5,7,9,11,13,15,17,19,21,23,25,27,29],columns=['Factors'])

NMF_result = pd.concat([Factors,MAE,RMSE,FIT_TIME,TEST_TIME],axis=1)
NMF_result.to_csv('./result_data/NMF_result.csv',header=True,encoding='utf-8')
#########################################################################################################################

####################################对SVD的可视化

SVD_noBaised_result <- read.csv('SVD_noBaised_result1.csv',encoding = 'utf-8')
SVD_baised_result <- read.csv('SVD_baised_result1.csv',encoding = 'utf-8')
SVDpp_result <- read.csv('SVDpp_result1.csv',encoding = 'utf-8')
NMF_result <- read.csv('NMF_result1.csv',encoding = 'utf-8')

SVD_noBiased_result <- as.data.table(SVD_noBaised_result)
SVD_biased_result <- as.data.table(SVD_baised_result)
SVDpp_result <- as.data.table(SVDpp_result)
NMF_result <- as.data.table(NMF_result)

SVD_noBiased_result <- SVD_noBiased_result[,SVD_class:='SVD_noBiased']
SVD_biased_result <- SVD_biased_result[,SVD_class:='SVD_biased']
SVDpp_result <- SVDpp_result[,SVD_class:='SVD++']
NMF_result <- NMF_result[,SVD_class:='NMF']

merge_SVD_result <- rbind(SVD_noBiased_result,SVD_biased_result,SVDpp_result,NMF_result)
merge_SVD_result <- merge_SVD_result[,-1]

# plot result

colour <- c('#34495e','#3498db','#2ecc71','#f1c40f','#e74c3c','#9b59b6','#1abc9c')
mycol <- define_palette(swatch = colour,gradient = c(lower=colour[1L],upper=colour[2L]))
ggthemr(mycol)



p01 <- ggplot(data= merge_SVD_result, aes(x=Factors, y= MAE,group=SVD_class, shape=SVD_class, color=SVD_class)) +
  geom_line()+
  geom_point()+
  theme(legend.position="top",axis.text.x = element_text(angle = 50, hjust = 0.5, vjust = 0.5),
        text = element_text(color = "black", size = 12))

p02 <- ggplot(data= merge_SVD_result, aes(x=Factors, y= RMSE,group=SVD_class, shape=SVD_class, color=SVD_class)) +
  geom_line()+
  geom_point()+
  theme(legend.position="top",axis.text.x = element_text(angle = 50, hjust = 0.5, vjust = 0.5),
        text = element_text(color = "black", size = 12))


p1 <- ggplot(data= merge_SVD_result, aes(x=Factors, y= FIT_TIME,group=SVD_class, shape=SVD_class, color=SVD_class)) +
  geom_line()+
  geom_point()+
  theme(legend.position="top",axis.text.x = element_text(angle = 50, hjust = 0.5, vjust = 0.5),
        text = element_text(color = "black", size = 12))


p2 <- ggplot(data= merge_SVD_result, aes(x=Factors, y= TEST_TIME,group=SVD_class, shape=SVD_class, color=SVD_class)) +
  geom_line()+
  geom_point()+
  theme(legend.position="top",axis.text.x = element_text(angle = 50, hjust = 0.5, vjust = 0.5),
        text = element_text(color = "black", size = 12))

p3 <- ggplot(data= filter(merge_SVD_result,SVD_class!='SVD++'), aes(x=Factors, y= FIT_TIME,group=SVD_class, shape=SVD_class, color=SVD_class)) +
  geom_line()+
  geom_point()+
  theme(legend.position="none",axis.text.x = element_text(angle = 50, hjust = 0.5, vjust = 0.5),
        text = element_text(color = "black", size = 12))


p4 <- ggplot(data= filter(merge_SVD_result,SVD_class!='SVD++'), aes(x=Factors, y= TEST_TIME,group=SVD_class, shape=SVD_class, color=SVD_class)) +
  geom_line()+
  geom_point()+
  theme(legend.position="none",axis.text.x = element_text(angle = 50, hjust = 0.5, vjust = 0.5),
        text = element_text(color = "black", size = 12))

x11()
ggplot2.multiplot(p01,p02,cols = 2)
x11()
ggplot2.multiplot(p1,p2,p3,p4,cols = 2)

这里写图片描述

这里写图片描述

猜你喜欢

转载自blog.csdn.net/u012111465/article/details/80306252
今日推荐
基于大语言模型的开源知识库问答系统 MaxKB GitHub Star 数量突破 5,000 个！
美国拟限制 AI 大模型出口中国和俄罗斯
苹果将与 OpenAI 达成协议，将 ChatGPT 应用于 iPhone
openKylin 社区生态委员会第六次会议圆满召开
阿里云正式发布通义千问 2.5
Python 3.13 发布首个 Beta：实验性自由线程模式和 JIT、改进交互式解释器
Stack Overflow 拿我的代码去训练 AI 大模型，还封了我的账号
Pop!_OS 的 COSMIC 桌面完成 App Store 上架工作
《2024 年一季度互联网投融资运行情况》研究报告
报告：Django 仍然是 74% 开发者的首选
15 年前上了“FFmpeg 耻辱柱”，今天他还得谢谢咱——腾讯QQPlayer一雪前耻？
TIOBE 5 月榜单：Fortran “复活”进入 Top 10
周排行
记一下去大梅沙的准备（2018-05-26）
Spring 注解 事务
基于HTTP协议的客户端缓存
阿里云rds 备份和还原
[PHP] 几个拖慢 PHP 程序/API 运行速度的点
python 代码风格------------PEP8规则
js控制json生成菜单——自制菜单（一）
将字符串: 'k:1|k1:2|k2:3|k3:4 ' ,处理成 python 字典: {'k':1, 'k1':2, ...}
微信小程序转支付宝小程序
Qt551.窗口滚动条
每日归档
更多
2024-05-13(18)
2024-05-12(0)
2024-05-11(38)
2024-05-10(38)
2024-05-09(35)
2024-05-08(42)
2024-05-07(14)
2024-05-06(40)
2024-05-05(0)
2024-05-04(7)

代码天地 © 2018 codetd.com

境外服务器   最新电视剧2022