Cash on Delivery Rejection Prediction
Liam
2018/7/26
R Markdown
货到付款是一种是一种非常受用户青睐的支付方式，对于客户而言，货到付款更加安全，特别是对于一些电商不发达的的确，货到付款能够有效的打消用户对于网购的不信任。
对于商家而言，货到付款这种支付方式不利于现金的流动，并且，有一部分人会在货到了之后不付款，也就是拒收。拒收的原因很多，很简单的就是不想要了。
一般而言，货到付款的拒收率可以高达20%，这将造成很大的运营成本。因此，本文利用机器学习的方法，对用户是否回拒收进行预测。
1 介绍
1.1 初步了解使用到的数据
kable(as.data.frame(colnames(Model_data)))| colnames(Model_data) |
|---|
| 发货方式 |
| 州 |
| 原始来单金额 |
| 修改后金额 |
| 发货件数 |
| 原始来单件数 |
| cod运费 |
| 用户性别 |
| 用户设备 |
| app1 |
| 用户类型 |
| 地址种类 |
| label |
| 下单小时 |
| 付款小时 |
| 下单与付款时间间隔 |
| 金额差异 |
| 件数差异 |
| 确认小时 |
| 付款到派送 |
1.1.2 查看数据的结构
str(Model_data)## Classes 'data.table' and 'data.frame': 322715 obs. of 20 variables:
## $ 发货方式 : chr "Delhivery" "Delhivery" "Ecom" "Ecom" ...
## $ 州 : chr "Telangana" "Telangana" "Maharashtra" "Maharashtra" ...
## $ 原始来单金额 : num 5.6 6.92 10.32 4.67 10.26 ...
## $ 修改后金额 : num 5.6 6.92 10.32 4.67 10.26 ...
## $ 发货件数 : num 1 1 1 1 1 6 1 3 1 1 ...
## $ 原始来单件数 : num 1 1 1 1 1 6 1 3 1 1 ...
## $ cod运费 : num 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 ...
## $ 用户性别 : chr "women" "women" "men" "women" ...
## $ 用户设备 : chr "ios" "android" "android" "android" ...
## $ app1 : chr "iOS_4.1.0" "android_4.1.1" "android_4.2.2" "android_4.0.3" ...
## $ 用户类型 : chr "new_cod" "new_cod" "new_cod" "old_cod" ...
## $ 地址种类 : chr "Valid Address" "Valid Address" "Missing Rooftop" "Valid Address" ...
## $ label : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
## $ 下单小时 : num 15 9 10 15 16 16 16 6 12 17 ...
## $ 付款小时 : num 14 11 15 16 16 17 11 6 6 9 ...
## $ 下单与付款时间间隔: num 19.5 16.9 17.4 16.9 19.6 ...
## $ 金额差异 : num 0 0 0 0 0 0 0 0 0 0 ...
## $ 件数差异 : num 0 0 0 0 0 0 0 0 0 0 ...
## $ 确认小时 : num 4 3 3 13 4 4 3 9 3 4 ...
## $ 付款到派送 : num 2.71 -0.477 -0.151 -0.127 -0.17 ...
## - attr(*, ".internal.selfref")=<externalptr>1.1.3 查看数据的基本信息
summary(Model_data)## 发货方式 州 原始来单金额 修改后金额
## Length:322715 Length:322715 Min. : 0.72 Min. : 0.390
## Class :character Class :character 1st Qu.: 3.22 1st Qu.: 3.210
## Mode :character Mode :character Median : 7.29 Median : 7.190
## Mean :10.12 Mean : 9.784
## 3rd Qu.:13.54 3rd Qu.:13.210
## Max. :51.55 Max. :51.550
##
## 发货件数 原始来单件数 cod运费 用户性别
## Min. : 0.000 Min. : 1.0 Min. :0.000 Length:322715
## 1st Qu.: 1.000 1st Qu.: 1.0 1st Qu.:0.770 Class :character
## Median : 1.000 Median : 1.0 Median :1.550 Mode :character
## Mean : 1.743 Mean : 1.8 Mean :1.197
## 3rd Qu.: 1.000 3rd Qu.: 2.0 3rd Qu.:1.550
## Max. :48.000 Max. :50.0 Max. :1.550
##
## 用户设备 app1 用户类型
## Length:322715 Length:322715 Length:322715
## Class :character Class :character Class :character
## Mode :character Mode :character Mode :character
##
##
##
##
## 地址种类 label 下单小时 付款小时
## Length:322715 0:285297 Min. : 0.00 Min. : 0.00
## Class :character 1: 37418 1st Qu.: 7.00 1st Qu.: 7.00
## Mode :character Median :11.00 Median :11.00
## Mean :10.85 Mean :10.87
## 3rd Qu.:15.00 3rd Qu.:15.00
## Max. :23.00 Max. :23.00
##
## 下单与付款时间间隔 金额差异 件数差异 确认小时
## Min. :-0.08387 Min. :-45.6000 Min. :-31.00000 Min. : 0.00
## 1st Qu.:-0.08260 1st Qu.: 0.0000 1st Qu.: 0.00000 1st Qu.: 7.00
## Median :-0.08166 Median : 0.0000 Median : 0.00000 Median : 9.00
## Mean : 0.00000 Mean : -0.3403 Mean : -0.05702 Mean : 9.37
## 3rd Qu.:-0.07987 3rd Qu.: 0.0000 3rd Qu.: 0.00000 3rd Qu.:12.00
## Max. :19.57188 Max. : 5.1600 Max. : 1.00000 Max. :23.00
##
## 付款到派送
## Min. :-4.011
## 1st Qu.:-0.669
## Median :-0.063
## Mean : 0.000
## 3rd Qu.: 0.611
## Max. :10.935
## NA's :34412 初步的数据分析
2.1 修改数据的类型
2.2 查看标签的比例
pct(Model_data$label)| Count | Percentage | |
|---|---|---|
| 0 | 285297 | 88.41 |
| 1 | 37418 | 11.59 |
3 单变量分析
WOE(Weight of Evidence):WOE显示了自变量对因变量的预测能力
WOE=ln(Distribution of Non-Events(Good)Distribution of Events(Bad))
其通过更基本的比率计算而来:
(Distribution of Good Credit Outcomes) / (Distribution of Bad Credit Outcomes)
Information Value(IV):
信息值有利于通过变量的重要性进行筛选变量
IV=∑(%Non-Events - %Events)∗WOE
Efficiency:
Efficiency=Abs(%Non-Events - %Events)/2
3.1 发货方式
A1 <- gbpct(Model_data$发货方式)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$发货方式), Model_data$label,
ylab="Good-Bad", xlab="发货方式",
main="发货方式对标签的影响")
barplot(A1$WOE, col="brown", names.arg=c(A1$Levels),
main="发货方式的WOE",
xlab="发货方式",
ylab="WOE"
)
3.2 州
A1 <- gbpct(Model_data$州)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$州), Model_data$label,
ylab="Good-Bad", xlab="州",
main="州对于标签的关系 ")
3.3 用户性别
A1 <- gbpct(Model_data$用户性别)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$用户性别), Model_data$label,
ylab="Good-Bad", xlab="用户性别",
main="用户性别对于标签的关系 ")
barplot(A1$WOE, col="brown", names.arg=c(A1$Levels),
main="用户性别的WOE",
xlab="用户性别",
ylab="WOE"
)
### 3.4 用户设备
A1 <- gbpct(Model_data$用户设备)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$用户设备), Model_data$label,
ylab="Good-Bad", xlab="用户设备",
main="用户设别与标签的关系")
barplot(A1$WOE, col="brown", names.arg=c(A1$Levels),
main="用户设备的WOE",
xlab="用户设备",
ylab="WOE"
)
3.5 操作系统版本
A1 <- gbpct(Model_data$app1)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$app1), Model_data$label,
ylab="Good-Bad", xlab="操作系统",
main="操作系统版本与标签的关系 ")
# barplot(A1$WOE, col="brown", names.arg=c(A1$Levels),
# main="Score:Checking Shipping method Status",
# xlab="Category",
# ylab="WOE"
# )
3.6 用户类别
A1 <- gbpct(Model_data$用户类型)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$用户类型), Model_data$label,
ylab="Good-Bad", xlab="用户类型",
main="用户类型与标签的关系")
barplot(A1$WOE, col="brown", names.arg=c(A1$Levels),
main="用户类型的WOE",
xlab="用户类型",
ylab="WOE"
)
3.7 地址类型
A1 <- gbpct(Model_data$地址种类)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$地址种类), Model_data$label,
ylab="Good-Bad", xlab="地址类型",
main="地址类型与标签的关系")
barplot(A1$WOE, col="brown", names.arg=c(A1$Levels),
main="地址类型WOE",
xlab="地址类型",
ylab="WOE"
)
3.8 下单时间(小时)
A1 <- gbpct(Model_data$下单小时)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$下单小时), Model_data$label,
ylab="Good-Bad", xlab="下单时间(小时)",
main="下单时间(小时)与标签的关系 ")
barplot(A1$WOE, col="brown", names.arg=c(A1$Levels),
main="下单时间(小时)WOE",
xlab="下单时间(小时)",
ylab="WOE"
)
3.9 付款时间(小时)
A1 <- gbpct(Model_data$付款小时)
op1<-par(mfrow=c(1,2), new=TRUE)## Warning in par(mfrow = c(1, 2), new = TRUE): 不绘图就不能调用par(new=TRUE)par(family='STKaiti')
plot(as.factor(Model_data$付款小时), Model_data$label,
ylab="Good-Bad", xlab="付款时间(小时)",
main="付款时间(小时)与标签的关系")
barplot(A1$WOE, col="brown", names.arg=c(A1$Levels),
main="付款时间(小时)WOE",
xlab="Category",
ylab="WOE"
)
计算信息值(Information Value) 和 WOE (Weight of Evidence)
kable(iv)| variable | info_value |
|---|---|
| 地址种类 | 0.4482661 |
| app1 | 0.3126790 |
| 下单与付款时间间隔 | 0.2858385 |
| cod运费 | 0.2818102 |
| 修改后金额 | 0.1986989 |
| 原始来单金额 | 0.1946768 |
| 金额差异 | 0.1632335 |
| 付款到派送 | 0.1379788 |
| 发货方式 | 0.1256872 |
| 用户性别 | 0.1238769 |
| 州 | 0.1158185 |
| 发货件数 | 0.0954921 |
| 原始来单件数 | 0.0929052 |
| 用户类型 | 0.0274259 |
| 确认小时 | 0.0205682 |
| 用户设备 | 0.0140496 |
| 付款小时 | 0.0119562 |
| 下单小时 | 0.0118502 |
| 件数差异 | 0.0073371 |
bins## $发货方式
## variable bin count count_distr good bad
## 1: 发货方式 XpressBees%,%Delhivery 172229 0.5336876 156606 15623
## 2: 发货方式 Ecom 150486 0.4663124 128691 21795
## badprob woe bin_iv total_iv breaks
## 1: 0.09071062 -0.2736100 0.03595137 0.06954223 XpressBees%,%Delhivery
## 2: 0.14483075 0.2556453 0.03359086 0.06954223 Ecom
## is_special_values
## 1: FALSE
## 2: FALSE
##
## $州
## variable
## 1: 州
## 2: 州
## 3: 州
## 4: 州
## bin
## 1: West bengal%,%UTTAR PRADESH%,%madhya pradesh%,%west bengal%,%Uttar pradesh%,%new delhi%,%New Delhi%,%andhra pradesh%,%maharashtra%,%WEST BENGAL%,%uttar pardesh%,%MADHYA PRADESH%,%palakkad%,%Kheda%,%haryana%,%Andhra pradesh%,%Maharashtara%,%Pondicherry%,%RAJSTHAN%,%Tamil nadu%,%Tamilnadu%,%Jammu & Kashmir%,%J&K%,%maharasta%,%Hyderabad%,%daman%,%GUJARAT%,%Haryana,%,%Jharkhan%,%Chattisgarh%,%karnataka%,%kerala%,%West Bangal%,%Meghalaya%,%Mizoram%,%Nagaland%,%Goa%,%Arunachal Pradesh%,%Assam%,%Daman and Diu%,%Puducherry%,%Kerala
## 2: West Bengal%,%Tamil Nadu%,%Chandigarh%,%Karnataka%,%Sikkim%,%Chhattisgarh%,%Himachal Pradesh%,%Andhra Pradesh
## 3: Telangana%,%Manipur%,%Odisha%,%Tripura%,%Gujarat%,%Uttarakhand
## 4: Jammu and Kashmir%,%Haryana%,%Madhya Pradesh%,%Uttar Pradesh%,%Punjab%,%Rajasthan%,%Maharashtra%,%Jharkhand%,%Delhi%,%Bihar%,%punjab%,%Andaman and Nicobar Islands%,%tamil nadu%,%Hariyana
## count count_distr good bad badprob woe bin_iv
## 1: 21778 0.06748369 20513 1265 0.05808614 -0.75460784 0.0287454667
## 2: 87476 0.27106270 80294 7182 0.08210252 -0.38273813 0.0342551910
## 3: 62240 0.19286367 55233 7007 0.11258033 -0.03327208 0.0002107931
## 4: 151221 0.46858993 129257 21964 0.14524438 0.25898095 0.0346850655
## total_iv
## 1: 0.09789652
## 2: 0.09789652
## 3: 0.09789652
## 4: 0.09789652
## breaks
## 1: West bengal%,%UTTAR PRADESH%,%madhya pradesh%,%west bengal%,%Uttar pradesh%,%new delhi%,%New Delhi%,%andhra pradesh%,%maharashtra%,%WEST BENGAL%,%uttar pardesh%,%MADHYA PRADESH%,%palakkad%,%Kheda%,%haryana%,%Andhra pradesh%,%Maharashtara%,%Pondicherry%,%RAJSTHAN%,%Tamil nadu%,%Tamilnadu%,%Jammu & Kashmir%,%J&K%,%maharasta%,%Hyderabad%,%daman%,%GUJARAT%,%Haryana,%,%Jharkhan%,%Chattisgarh%,%karnataka%,%kerala%,%West Bangal%,%Meghalaya%,%Mizoram%,%Nagaland%,%Goa%,%Arunachal Pradesh%,%Assam%,%Daman and Diu%,%Puducherry%,%Kerala
## 2: West Bengal%,%Tamil Nadu%,%Chandigarh%,%Karnataka%,%Sikkim%,%Chhattisgarh%,%Himachal Pradesh%,%Andhra Pradesh
## 3: Telangana%,%Manipur%,%Odisha%,%Tripura%,%Gujarat%,%Uttarakhand
## 4: Jammu and Kashmir%,%Haryana%,%Madhya Pradesh%,%Uttar Pradesh%,%Punjab%,%Rajasthan%,%Maharashtra%,%Jharkhand%,%Delhi%,%Bihar%,%punjab%,%Andaman and Nicobar Islands%,%tamil nadu%,%Hariyana
## is_special_values
## 1: FALSE
## 2: FALSE
## 3: FALSE
## 4: FALSE
##
## $原始来单金额
## variable bin count count_distr good bad badprob
## 1: 原始来单金额 [-Inf,2) 20298 0.06289760 18324 1974 0.09725096
## 2: 原始来单金额 [2,4) 82577 0.25588213 73354 9223 0.11168970
## 3: 原始来单金额 [4,6) 43843 0.13585672 39918 3925 0.08952398
## 4: 原始来单金额 [6,10) 55017 0.17048169 48312 6705 0.12187142
## 5: 原始来单金额 [10,18) 72926 0.22597648 62305 10621 0.14564079
## 6: 原始来单金额 [18,28) 29221 0.09054739 25891 3330 0.11395914
## 7: 原始来单金额 [28, Inf) 18833 0.05835799 17193 1640 0.08708119
## woe bin_iv total_iv breaks is_special_values
## 1: -0.19677086 0.0022574432 0.03578174 2 FALSE
## 2: -0.04221780 0.0004487274 0.03578174 4 FALSE
## 3: -0.28808213 0.0100890135 0.03578174 6 FALSE
## 4: 0.05655241 0.0005571833 0.03578174 10 FALSE
## 5: 0.26217036 0.0171619061 0.03578174 18 FALSE
## 6: -0.01954424 0.0000343283 0.03578174 28 FALSE
## 7: -0.31842721 0.0052331416 0.03578174 Inf FALSE
##
## $修改后金额
## variable bin count count_distr good bad badprob
## 1: 修改后金额 [-Inf,2) 20384 0.06316409 18401 1983 0.09728218
## 2: 修改后金额 [2,4) 83151 0.25766078 73874 9277 0.11156811
## 3: 修改后金额 [4,6) 44200 0.13696295 40226 3974 0.08990950
## 4: 修改后金额 [6,10) 56409 0.17479510 49630 6779 0.12017586
## 5: 修改后金额 [10,18) 73848 0.22883349 63115 10733 0.14533907
## 6: 修改后金额 [18,25) 23496 0.07280728 20784 2712 0.11542390
## 7: 修改后金额 [25, Inf) 21227 0.06577630 19267 1960 0.09233523
## woe bin_iv total_iv breaks is_special_values
## 1: -0.196415291 2.259132e-03 0.03378337 2 FALSE
## 2: -0.043443850 4.782461e-04 0.03378337 4 FALSE
## 3: -0.283361538 9.858532e-03 0.03378337 6 FALSE
## 4: 0.040612980 2.928368e-04 0.03378337 10 FALSE
## 5: 0.259743519 1.704306e-02 0.03378337 18 FALSE
## 6: -0.005118219 1.903526e-06 0.03378337 25 FALSE
## 7: -0.254070444 3.849656e-03 0.03378337 Inf FALSE
##
## $发货件数
## variable bin count count_distr good bad badprob woe
## 1: 发货件数 [-Inf,2) 242394 0.75110856 210330 32064 0.13228050 0.1504351
## 2: 发货件数 [2,3) 30017 0.09301396 27705 2312 0.07702302 -0.4521211
## 3: 发货件数 [3, Inf) 50304 0.15587748 47262 3042 0.06047233 -0.7118125
## bin_iv total_iv breaks is_special_values
## 1: 0.01800438 0.09402303 2 FALSE
## 2: 0.01596932 0.09402303 3 FALSE
## 3: 0.06004934 0.09402303 Inf FALSE
##
## $原始来单件数
## variable bin count count_distr good bad badprob
## 1: 原始来单件数 [-Inf,2) 239614 0.74249415 207870 31744 0.13247974
## 2: 原始来单件数 [2,3) 29493 0.09139024 27176 2317 0.07856101
## 3: 原始来单件数 [3, Inf) 53608 0.16611561 50251 3357 0.06262125
## woe bin_iv total_iv breaks is_special_values
## 1: 0.1521697 0.01822272 0.09087764 2 FALSE
## 2: -0.4306821 0.01435595 0.09087764 3 FALSE
## 3: -0.6746039 0.05829897 0.09087764 Inf FALSE
##
## $cod运费
## variable bin count count_distr good bad badprob
## 1: cod运费 [-Inf,1.5) 143652 0.4451358 129995 13657 0.09507003
## 2: cod运费 [1.5, Inf) 179063 0.5548642 155302 23761 0.13269631
## woe bin_iv total_iv breaks is_special_values
## 1: -0.2218649 0.02011498 0.03408191 1.5 FALSE
## 2: 0.1540528 0.01396692 0.03408191 Inf FALSE
##
## $用户性别
## variable bin count count_distr good bad badprob
## 1: 用户性别 missing 1972 0.006110655 1694 278 0.14097363
## 2: 用户性别 not set%,%women 228872 0.709207815 207855 21017 0.09182862
## 3: 用户性别 men 91871 0.284681530 75748 16123 0.17549608
## woe bin_iv total_iv breaks is_special_values
## 1: 0.2241521 0.0003344142 0.1238244 missing TRUE
## 2: -0.2601302 0.0434092333 0.1238244 not set%,%women FALSE
## 3: 0.4842137 0.0800807579 0.1238244 men FALSE
##
## $用户设备
## variable bin count count_distr good bad badprob
## 1: 用户设备 missing 2467 0.007644516 2109 358 0.14511552
## 2: 用户设备 pc%,%mobile%,%ios 35046 0.108597369 32068 2978 0.08497403
## 3: 用户设备 android 285202 0.883758115 251120 34082 0.11950127
## woe bin_iv total_iv breaks is_special_values
## 1: 0.25794268 0.0005611005 0.01293809 missing TRUE
## 2: -0.34522784 0.0113285823 0.01293809 pc%,%mobile%,%ios FALSE
## 3: 0.03421734 0.0010484026 0.01293809 android FALSE
##
## $app1
## variable
## 1: app1
## 2: app1
## 3: app1
## 4: app1
## bin
## 1: missing
## 2: android_2.45%,%android_3.7.1%,%android_2.49%,%android_3.3.3%,%android_4.2.0%,%android_2.34%,%android_2.48%,%android_2.33%,%android_4.0.1%,%iOS_1.6.1%,%iOS_1.5.9%,%android_3.2.0%,%iOS_1.5.8%,%android_2.38%,%android_3.3.0%,%android_3.4.0%,%android_null%,%android_4.3.4%,%pc%,%android_4.3.5%,%iOS_1.6.2%,%android_4.0.2%,%android_4.3.3%,%iOS_1.9.1%,%iOS_2.0.0
## 3: iOS_1.8.0%,%iOS_4.2.0
## 4: iOS_2.0.1%,%android_4.3.0%,%iOS_1.7.0%,%iOS_2.1.0%,%iOS_1.9.0%,%android_3.6.2%,%android_3.7.3%,%android_4.2.1%,%iOS_4.0.0%,%android_4.3.2%,%android_3.8.0%,%android_3.4.3%,%android_3.2.1%,%android_3.9.1%,%android_3.3.1%,%android_3.8.1%,%android_4.0.3%,%android_3.1.1%,%android_3.5.5%,%iOS_4.1.0%,%android_3.4.2%,%android_3.4.1%,%android_4.1.1%,%android_4.2.3%,%android_4.1.0%,%android_3.9.0%,%mobile-pwa%,%iOS_1.6.0%,%mobile%,%android_3.0.2%,%android_3.5.2%,%android_2.42%,%android_3.6.1%,%android_3.0.1%,%android_4.2.2%,%android_3.7.0%,%android_4.0.0%,%android_3.5.1%,%android_4.3.1%,%android_2.44%,%android_2.50%,%android_2.40%,%android_2.46%,%android_2.47
## count count_distr good bad badprob woe bin_iv
## 1: 2467 0.007644516 2109 358 0.14511552 0.2579427 0.0005611005
## 2: 95468 0.295827588 90854 4614 0.04833033 -0.9487798 0.1851491250
## 3: 29467 0.091309670 27136 2331 0.07910544 -0.4231850 0.0138883801
## 4: 195313 0.605218227 165198 30115 0.15418841 0.3292575 0.0743423496
## total_iv
## 1: 0.273941
## 2: 0.273941
## 3: 0.273941
## 4: 0.273941
## breaks
## 1: missing
## 2: android_2.45%,%android_3.7.1%,%android_2.49%,%android_3.3.3%,%android_4.2.0%,%android_2.34%,%android_2.48%,%android_2.33%,%android_4.0.1%,%iOS_1.6.1%,%iOS_1.5.9%,%android_3.2.0%,%iOS_1.5.8%,%android_2.38%,%android_3.3.0%,%android_3.4.0%,%android_null%,%android_4.3.4%,%pc%,%android_4.3.5%,%iOS_1.6.2%,%android_4.0.2%,%android_4.3.3%,%iOS_1.9.1%,%iOS_2.0.0
## 3: iOS_1.8.0%,%iOS_4.2.0
## 4: iOS_2.0.1%,%android_4.3.0%,%iOS_1.7.0%,%iOS_2.1.0%,%iOS_1.9.0%,%android_3.6.2%,%android_3.7.3%,%android_4.2.1%,%iOS_4.0.0%,%android_4.3.2%,%android_3.8.0%,%android_3.4.3%,%android_3.2.1%,%android_3.9.1%,%android_3.3.1%,%android_3.8.1%,%android_4.0.3%,%android_3.1.1%,%android_3.5.5%,%iOS_4.1.0%,%android_3.4.2%,%android_3.4.1%,%android_4.1.1%,%android_4.2.3%,%android_4.1.0%,%android_3.9.0%,%mobile-pwa%,%iOS_1.6.0%,%mobile%,%android_3.0.2%,%android_3.5.2%,%android_2.42%,%android_3.6.1%,%android_3.0.1%,%android_4.2.2%,%android_3.7.0%,%android_4.0.0%,%android_3.5.1%,%android_4.3.1%,%android_2.44%,%android_2.50%,%android_2.40%,%android_2.46%,%android_2.47
## is_special_values
## 1: TRUE
## 2: FALSE
## 3: FALSE
## 4: FALSE
##
## $用户类型
## variable
## 1: 用户类型
## 2: 用户类型
## 3: 用户类型
## bin count
## 1: old_prepaid_old_cod%,%old_prepaid_new_cod%,%new_prepaid_old_cod 31057
## 2: old_cod 47207
## 3: new_cod%,%new_prepaid_new_cod 244451
## count_distr good bad badprob woe bin_iv total_iv
## 1: 0.09623662 28600 2457 0.0791126 -0.42308675 0.014631535 0.02626036
## 2: 0.14628077 42704 4503 0.0953884 -0.21816988 0.006400987 0.02626036
## 3: 0.75748261 213993 30458 0.1245976 0.08178425 0.005227836 0.02626036
## breaks
## 1: old_prepaid_old_cod%,%old_prepaid_new_cod%,%new_prepaid_old_cod
## 2: old_cod
## 3: new_cod%,%new_prepaid_new_cod
## is_special_values
## 1: FALSE
## 2: FALSE
## 3: FALSE
##
## $地址种类
## variable bin count
## 1: 地址种类 missing 32963
## 2: 地址种类 Valid Address 211036
## 3: 地址种类 Missing Rooftop with POI 27599
## 4: 地址种类 Missing Rooftop%,%Inappropriate%,%Incomplete%,%Junk 51117
## count_distr good bad badprob woe bin_iv total_iv
## 1: 0.10214276 32836 127 0.003852805 -3.52371480 3.935990e-01 0.4450414
## 2: 0.65393923 186341 24695 0.117017950 0.01040134 7.103135e-05 0.4450414
## 3: 0.08552128 24204 3395 0.123011703 0.06716472 3.958573e-04 0.4450414
## 4: 0.15839673 41916 9201 0.179998826 0.51502343 5.097554e-02 0.4450414
## breaks is_special_values
## 1: missing TRUE
## 2: Valid Address FALSE
## 3: Missing Rooftop with POI FALSE
## 4: Missing Rooftop%,%Inappropriate%,%Incomplete%,%Junk FALSE
##
## $下单小时
## variable bin count count_distr good bad badprob
## 1: 下单小时 [-Inf,5) 37569 0.11641541 32658 4911 0.1307195
## 2: 下单小时 [5,17) 231021 0.71586694 205342 25679 0.1111544
## 3: 下单小时 [17,19) 34948 0.10829370 30943 4005 0.1145988
## 4: 下单小时 [19, Inf) 19177 0.05942395 16354 2823 0.1472076
## woe bin_iv total_iv breaks is_special_values
## 1: 0.13676661 0.0022945074 0.008885842 5 FALSE
## 2: -0.04762447 0.0015941915 0.008885842 17 FALSE
## 3: -0.01322435 0.0000188428 0.008885842 19 FALSE
## 4: 0.27470650 0.0049783008 0.008885842 Inf FALSE
##
## $付款小时
## variable bin count count_distr good bad badprob
## 1: 付款小时 [-Inf,5) 37095 0.11494662 32224 4871 0.1313115
## 2: 付款小时 [5,17) 230946 0.71563454 205269 25677 0.1111818
## 3: 付款小时 [17,19) 35187 0.10903429 31172 4015 0.1141046
## 4: 付款小时 [19, Inf) 19487 0.06038455 16632 2855 0.1465079
## woe bin_iv total_iv breaks is_special_values
## 1: 0.14196661 2.445947e-03 0.008901763 5 FALSE
## 2: -0.04734679 1.575312e-03 0.008901763 17 FALSE
## 3: -0.01810404 3.548894e-05 0.008901763 19 FALSE
## 4: 0.26912216 4.845015e-03 0.008901763 Inf FALSE
##
## $下单与付款时间间隔
## variable bin count count_distr good bad
## 1: 下单与付款时间间隔 [-Inf,-0.083) 38240 0.1184946 33125 5115
## 2: 下单与付款时间间隔 [-0.083,-0.0814) 137899 0.4273089 120792 17107
## 3: 下单与付款时间间隔 [-0.0814,-0.0774) 106184 0.3290334 94815 11369
## 4: 下单与付款时间间隔 [-0.0774, Inf) 40392 0.1251631 36565 3827
## badprob woe bin_iv total_iv breaks is_special_values
## 1: 0.13376046 0.16326799 0.003361988 0.01435376 -0.083 FALSE
## 2: 0.12405456 0.07679655 0.002595418 0.01435376 -0.0814 FALSE
## 3: 0.10706886 -0.08965840 0.002555278 0.01435376 -0.0774 FALSE
## 4: 0.09474648 -0.22563142 0.005841080 0.01435376 Inf FALSE
##
## $金额差异
## variable bin count count_distr good bad badprob woe
## 1: 金额差异 [-Inf, Inf) 322715 1 285297 37418 0.1159475 0
## bin_iv total_iv breaks is_special_values
## 1: 0 0 Inf FALSE
##
## $件数差异
## variable bin count count_distr good bad badprob woe
## 1: 件数差异 [-Inf, Inf) 322715 1 285297 37418 0.1159475 0
## bin_iv total_iv breaks is_special_values
## 1: 0 0 Inf FALSE
##
## $确认小时
## variable bin count count_distr good bad badprob
## 1: 确认小时 [-Inf,5) 20463 0.06340889 18528 1935 0.09456091
## 2: 确认小时 [5,12) 220152 0.68218707 193376 26776 0.12162506
## 3: 确认小时 [12,13) 29449 0.09125389 26112 3337 0.11331454
## 4: 确认小时 [13, Inf) 52651 0.16315015 47281 5370 0.10199236
## woe bin_iv total_iv breaks is_special_values
## 1: -0.22779690 3.013701e-03 0.00832062 5 FALSE
## 2: 0.05424835 2.049801e-03 0.00832062 12 FALSE
## 3: -0.02594391 6.081242e-05 0.00832062 13 FALSE
## 4: -0.14390174 3.196306e-03 0.00832062 Inf FALSE
##
## $付款到派送
## variable bin count count_distr good bad badprob
## 1: 付款到派送 missing 3441 0.01066266 1327 2114 0.61435629
## 2: 付款到派送 [-Inf,0.2) 190833 0.59133601 168293 22540 0.11811374
## 3: 付款到派送 [0.2,1) 81786 0.25343105 73034 8752 0.10701098
## 4: 付款到派送 [1,1.4) 20867 0.06466077 18904 1963 0.09407198
## 5: 付款到派送 [1.4, Inf) 25788 0.07990952 23739 2049 0.07945556
## woe bin_iv total_iv breaks is_special_values
## 1: 2.49704000 0.1294604871 0.1468417 missing TRUE
## 2: 0.02096387 0.0002619823 0.1468417 0.2 FALSE
## 3: -0.09026397 0.0019943602 0.1468417 1 FALSE
## 4: -0.23352075 0.0032224443 0.1468417 1.4 FALSE
## 5: -0.41838853 0.0119024356 0.1468417 Inf FALSE- 下面这些变量是没有预测能力或者预测能力非常弱的一些变量 (IV< 2%), 因此可以直接将这些变量筛选掉
library(tidyverse)
kable(iv %>% filter(info_value<0.02))## Warning: package 'bindrcpp' was built under R version 3.4.4| variable | info_value |
|---|---|
| 用户设备 | 0.0140496 |
| 付款小时 | 0.0119562 |
| 下单小时 | 0.0118502 |
| 件数差异 | 0.0073371 |
- 下面这一部分变量只是有非常弱的预测变量 (2%<=IV< 10%), 因此可以考虑加上这一部分变量，也可以不加上这些变量
library(tidyverse)
kable(iv %>% filter(info_value>=0.02,info_value<0.1))| variable | info_value |
|---|---|
| 发货件数 | 0.0954921 |
| 原始来单件数 | 0.0929052 |
| 用户类型 | 0.0274259 |
| 确认小时 | 0.0205682 |
- 这些变量有一定的预测能力 (10%<=IV< 30%), 可以考虑选取其中一些变量加入到模型里面去
library(tidyverse)
kable(iv %>% filter(info_value>=0.1,info_value<0.3))| variable | info_value |
|---|---|
| 下单与付款时间间隔 | 0.2858385 |
| cod运费 | 0.2818102 |
| 修改后金额 | 0.1986989 |
| 原始来单金额 | 0.1946768 |
| 金额差异 | 0.1632335 |
| 付款到派送 | 0.1379788 |
| 发货方式 | 0.1256872 |
| 用户性别 | 0.1238769 |
| 州 | 0.1158185 |
- 这些变量有比较强的预测能力 (IV 30% to 50%),模型选取这一部分变量进行建模
library(tidyverse)
kable(iv %>% filter(info_value>=0.3,info_value<0.5))| variable | info_value |
|---|---|
| 地址种类 | 0.4482661 |
| app1 | 0.3126790 |
4 选取进行建模的变量
var_list_1 <- iv %>% filter(info_value>0.1) %>% select(variable) # 15 variables
Model_data1 <- Model_data %>% select(var_list_1$variable,label) #12 variables
head(Model_data1)## 地址种类 app1 下单与付款时间间隔 cod运费 修改后金额
## 1: Valid Address iOS_4.1.0 19.45732 1.55 5.60
## 2: Valid Address android_4.1.1 16.93115 1.55 6.92
## 3: Missing Rooftop android_4.2.2 17.41311 1.55 10.32
## 4: Valid Address android_4.0.3 16.85653 1.55 4.67
## 5: Missing Rooftop android_4.1.1 19.56840 1.55 10.26
## 6: Valid Address iOS_4.1.0 16.91516 1.55 16.02
## 原始来单金额 金额差异 付款到派送 发货方式 用户性别 州 label
## 1: 5.60 0 2.7096488 Delhivery women Telangana 0
## 2: 6.92 0 -0.4770722 Delhivery women Telangana 0
## 3: 10.32 0 -0.1513002 Ecom men Maharashtra 0
## 4: 4.67 0 -0.1274765 Ecom women Maharashtra 0
## 5: 10.26 0 -0.1704649 Delhivery men Karnataka 0
## 6: 16.02 0 0.2219836 Delhivery women Karnataka 05 多元数据分析 - 聚类，降维
对变量的聚类可以讲含有相同信息的变量聚为同一个族类
当我们有大量的变量的时候，这种方法可以很好的用于进行降维。同样可以用于降维的方法还有主成分分析和因子分析。
Model_data1$app1 <- as.factor(Model_data1$app1)
Model_data1$label <- as.factor(Model_data1$label)
Model_data1$地址种类 <- as.factor(Model_data1$地址种类)
Model_data1$发货方式 <- as.factor(Model_data1$发货方式)
Model_data1$用户性别 <- as.factor(Model_data1$用户性别)
Model_data1$州 <- as.factor(Model_data1$州)factors <- sapply(Model_data1, is.factor)
#subset Qualitative variables
vars_quali <- Model_data1 %>% select(names(Model_data1)[factors])
#vars_quali$good_bad_21<-vars_quali$good_bad_21[drop=TRUE] # remove empty factors
str(vars_quali)## Classes 'data.table' and 'data.frame': 322715 obs. of 6 variables:
## $ 地址种类: Factor w/ 6 levels "Inappropriate",..: 6 6 4 6 4 6 6 6 4 6 ...
## $ app1 : Factor w/ 71 levels "android_2.33",..: 67 42 45 40 42 67 42 32 42 29 ...
## $ 发货方式: Factor w/ 3 levels "Delhivery","Ecom",..: 1 1 2 2 1 1 1 2 2 1 ...
## $ 用户性别: Factor w/ 3 levels "men","not set",..: 3 3 1 3 1 3 3 3 3 1 ...
## $ 州 : Factor w/ 70 levels "Andaman and Nicobar Islands",..: 59 59 38 38 29 29 29 14 29 29 ...
## $ label : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
## - attr(*, ".internal.selfref")=<externalptr>#subset Quantitative variables
vars_quanti <- Model_data1 %>% select(names(Model_data1)[!factors])
str(vars_quanti)## Classes 'data.table' and 'data.frame': 322715 obs. of 6 variables:
## $ 下单与付款时间间隔: num 19.5 16.9 17.4 16.9 19.6 ...
## $ cod运费 : num 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 ...
## $ 修改后金额 : num 5.6 6.92 10.32 4.67 10.26 ...
## $ 原始来单金额 : num 5.6 6.92 10.32 4.67 10.26 ...
## $ 金额差异 : num 0 0 0 0 0 0 0 0 0 0 ...
## $ 付款到派送 : num 2.71 -0.477 -0.151 -0.127 -0.17 ...
## - attr(*, ".internal.selfref")=<externalptr>## 6 变量的层次聚类
tree <- hclustvar(X.quanti=vars_quanti,X.quali=vars_quali)
par(family='STKaiti')
plot(tree, main="variable clustering")
rect.hclust(tree, k=8, border = 1:8)
summary(tree)## Length Class Mode
## call 3 -none- call
## rec 16 -none- list
## init 12 -none- numeric
## merge 22 -none- numeric
## height 11 -none- numeric
## order 12 -none- numeric
## labels 12 -none- character
## clusmat 144 -none- numeric
## X.quanti 6 data.table list
## X.quali 6 data.table list# Phylogenetic trees
# require library("ape")
par(family='STKaiti')
plot(as.phylo(tree), type = "fan",
tip.color = hsv(runif(15, 0.65, 0.95), 1, 1, 0.7),
edge.color = hsv(runif(10, 0.65, 0.75), 1, 1, 0.7),
edge.width = runif(20, 0.5, 3), use.edge.length = TRUE, col = "gray80")
summary.phylo(as.phylo(tree))##
## Phylogenetic tree: as.phylo(tree)
##
## Number of tips: 12
## Number of nodes: 11
## Branch lengths:
## mean: 0.2498154
## variance: 0.02762882
## distribution summary:
## Min. 1st Qu. Median 3rd Qu. Max.
## 0.01203149 0.11483605 0.24931255 0.40189405 0.49995107
## No root edge.
## First ten tip labels: 下单与付款时间间隔
## cod运费
## 修改后金额
## 原始来单金额
## 金额差异
## 付款到派送
## 地址种类
## app1
## 发货方式
## 用户性别
## No node labels.part<-cutreevar(tree,8)
print(part)##
## Call:
## cutreevar(obj = tree, k = 8)
##
##
##
## name
## "$var"
## "$sim"
## "$cluster"
## "$wss"
## "$E"
## "$size"
## "$scores"
## "$coef"
## description
## "list of variables in each cluster"
## "similarity matrix in each cluster"
## "cluster memberships"
## "within-cluster sum of squares"
## "gain in cohesion (in %)"
## "size of each cluster"
## "synthetic score of each cluster"
## "coef of the linear combinations defining the synthetic scores of each cluster"summary(part)##
## Call:
## cutreevar(obj = tree, k = 8)
##
##
##
## Data:
## number of observations: 322715
## number of variables: 12
## number of numerical variables: 6
## number of categorical variables: 6
## number of clusters: 8
##
## Cluster 1 :
## squared loading correlation
## 1 1
##
##
## Cluster 2 :
## squared loading correlation
## 修改后金额 0.93 -0.96
## 原始来单金额 0.92 -0.96
## cod运费 0.65 -0.81
##
##
## Cluster 3 :
## squared loading correlation
## 1 1
##
##
## Cluster 4 :
## squared loading correlation
## 州 0.68 NA
## 付款到派送 0.56 -0.75
## 发货方式 0.44 NA
##
##
## Cluster 5 :
## squared loading correlation
## 1 NA
##
##
## Cluster 6 :
## squared loading correlation
## 1 NA
##
##
## Cluster 7 :
## squared loading correlation
## 1 NA
##
##
## Cluster 8 :
## squared loading correlation
## 1 NA
##
##
## Gain in cohesion (in %): 80.387 通过聚类选取部分变量
# cod运费
# 付款到派送
# keep<- c(1,2,3,4,7,8,10,12)
cdata_reduced_2 <- Model_data1 # %>% select(keep)
str(cdata_reduced_2)## Classes 'data.table' and 'data.frame': 322715 obs. of 12 variables:
## $ 地址种类 : Factor w/ 6 levels "Inappropriate",..: 6 6 4 6 4 6 6 6 4 6 ...
## $ app1 : Factor w/ 71 levels "android_2.33",..: 67 42 45 40 42 67 42 32 42 29 ...
## $ 下单与付款时间间隔: num 19.5 16.9 17.4 16.9 19.6 ...
## $ cod运费 : num 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 1.55 ...
## $ 修改后金额 : num 5.6 6.92 10.32 4.67 10.26 ...
## $ 原始来单金额 : num 5.6 6.92 10.32 4.67 10.26 ...
## $ 金额差异 : num 0 0 0 0 0 0 0 0 0 0 ...
## $ 付款到派送 : num 2.71 -0.477 -0.151 -0.127 -0.17 ...
## $ 发货方式 : Factor w/ 3 levels "Delhivery","Ecom",..: 1 1 2 2 1 1 1 2 2 1 ...
## $ 用户性别 : Factor w/ 3 levels "men","not set",..: 3 3 1 3 1 3 3 3 3 1 ...
## $ 州 : Factor w/ 70 levels "Andaman and Nicobar Islands",..: 59 59 38 38 29 29 29 14 29 29 ...
## $ label : Factor w/ 2 levels "0","1": 1 1 1 1 1 1 1 1 1 1 ...
## - attr(*, ".internal.selfref")=<externalptr>8 划分选来集合测试集合
bins <- scorecard::woebin(cdata_reduced_2,y = 'label')## Binning on 322715 rows and 12 columns in 00:00:12dt_woe <- scorecard::woebin_ply(cdata_reduced_2,bins)## Woe transformating on 322715 rows and 11 columns in 00:00:14dt_woe$label <- as.factor(dt_woe$label)
div_part_1 <- createDataPartition(y = dt_woe$label, p = 0.7, list = F)
# Training Sample
train_1 <- dt_woe[div_part_1,] # 70% here
pct(train_1$label)| Count | Percentage | |
|---|---|---|
| 0 | 199708 | 88.41 |
| 1 | 26193 | 11.59 |
# Test Sample
test_1 <- dt_woe[-div_part_1,] # rest of the 30% data goes here
pct(test_1$label)| Count | Percentage | |
|---|---|---|
| 0 | 85589 | 88.41 |
| 1 | 11225 | 11.59 |
9 训练模型以及模型选择
9.1 逻辑回归以及逐步回归
m1 <- glm(label~.,data=train_1,family=binomial())
m1 <- step(m1)## Start: AIC=141024.5
## label ~ 地址种类_woe + app1_woe + 下单与付款时间间隔_woe +
## cod运费_woe + 修改后金额_woe + 原始来单金额_woe +
## 金额差异_woe + 付款到派送_woe + 发货方式_woe +
## 用户性别_woe + 州_woe
##
##
## Step: AIC=141024.5
## label ~ 地址种类_woe + app1_woe + 下单与付款时间间隔_woe +
## cod运费_woe + 修改后金额_woe + 原始来单金额_woe +
## 付款到派送_woe + 发货方式_woe + 用户性别_woe +
## 州_woe
##
## Df Deviance AIC
## - 修改后金额_woe 1 141003 141023
## <none> 141003 141025
## - 原始来单金额_woe 1 141009 141029
## - cod运费_woe 1 141135 141155
## - 下单与付款时间间隔_woe 1 141299 141319
## - 发货方式_woe 1 141412 141432
## - 州_woe 1 141435 141455
## - 用户性别_woe 1 142696 142716
## - 付款到派送_woe 1 145948 145968
## - app1_woe 1 146169 146189
## - 地址种类_woe 1 146308 146328
##
## Step: AIC=141022.6
## label ~ 地址种类_woe + app1_woe + 下单与付款时间间隔_woe +
## cod运费_woe + 原始来单金额_woe + 付款到派送_woe +
## 发货方式_woe + 用户性别_woe + 州_woe
##
## Df Deviance AIC
## <none> 141003 141023
## - 原始来单金额_woe 1 141065 141083
## - cod运费_woe 1 141139 141157
## - 下单与付款时间间隔_woe 1 141299 141317
## - 发货方式_woe 1 141412 141430
## - 州_woe 1 141435 141453
## - 用户性别_woe 1 142696 142714
## - 付款到派送_woe 1 145949 145967
## - app1_woe 1 146169 146187
## - 地址种类_woe 1 146308 146326summary(m1)##
## Call:
## glm(formula = label ~ 地址种类_woe + app1_woe + 下单与付款时间间隔_woe +
## cod运费_woe + 原始来单金额_woe + 付款到派送_woe +
## 发货方式_woe + 用户性别_woe + 州_woe, family = binomial(),
## data = train_1)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.6032 -0.5345 -0.3971 -0.2335 3.8450
##
## Coefficients:
## Estimate Std. Error z value Pr(>|z|)
## (Intercept) -2.032089 0.007589 -267.759 < 2e-16 ***
## 地址种类_woe 0.989096 0.020734 47.704 < 2e-16 ***
## app1_woe 1.005629 0.015343 65.544 < 2e-16 ***
## 下单与付款时间间隔_woe 1.005120 0.058818 17.089 < 2e-16 ***
## cod运费_woe 0.541428 0.046333 11.686 < 2e-16 ***
## 原始来单金额_woe 0.349058 0.044593 7.828 4.97e-15 ***
## 付款到派送_woe 1.408441 0.021893 64.332 < 2e-16 ***
## 发货方式_woe 0.750572 0.037196 20.179 < 2e-16 ***
## 用户性别_woe 0.825245 0.019827 41.622 < 2e-16 ***
## 州_woe 0.624494 0.030162 20.705 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 162095 on 225900 degrees of freedom
## Residual deviance: 141003 on 225891 degrees of freedom
## AIC: 141023
##
## Number of Fisher Scoring iterations: 7significant.variables <- summary(m1)$coeff[-1,4] < 0.01
names(significant.variables)[significant.variables == TRUE]## [1] "地址种类_woe" "app1_woe"
## [3] "下单与付款时间间隔_woe" "cod运费_woe"
## [5] "原始来单金额_woe" "付款到派送_woe"
## [7] "发货方式_woe" "用户性别_woe"
## [9] "州_woe"dt_pred = predict(m1, type='response', test_1)
perf_eva(test_1$label, dt_pred, type = c("ks","lift","roc","pr"))## Warning: Removed 1 rows containing missing values (geom_path).
## $KS
## [1] 0.3402
##
## $AUC
## [1] 0.7434
##
## $Gini
## [1] 0.4868
##
## $pic
## TableGrob (2 x 2) "arrange": 4 grobs
## z cells name grob
## pks 1 (1-1,1-1) arrange gtable[layout]
## plift 2 (1-1,2-2) arrange gtable[layout]
## proc 3 (2-2,1-1) arrange gtable[layout]
## ppr 4 (2-2,2-2) arrange gtable[layout]9.2 随即森林
m3 <- randomForest(label ~ ., data = train_1)
par(family='STKaiti')
varImpPlot(m3, main="Random Forest: Variable Importance")
dt_pred = predict(m3, type='prob', test_1)[,1]
perf_eva(test_1$label, dt_pred, type = c("ks","lift","roc","pr"))## Warning: Removed 1 rows containing missing values (geom_path).
## $KS
## [1] 0.1661
##
## $AUC
## [1] 0.4041
##
## $Gini
## [1] -0.1919
##
## $pic
## TableGrob (2 x 2) "arrange": 4 grobs
## z cells name grob
## pks 1 (1-1,1-1) arrange gtable[layout]
## plift 2 (1-1,2-2) arrange gtable[layout]
## proc 3 (2-2,1-1) arrange gtable[layout]
## ppr 4 (2-2,2-2) arrange gtable[layout]不平衡的数据会造成非常低AUC，需要尝试解决样本不平衡的问题
欠抽样
load('/Users/milin/COD\ 建模/model_rf_under.RData')
load('/Users/milin/COD\ 建模/dt_woe.RData')
require(scorecard)
dt_pred = predict(model_rf_under, type = 'prob', dt_woe)
perf_eva(dt_woe$label, dt_pred$`1`)
## $KS
## [1] 0.3986
##
## $AUC
## [1] 0.7641
##
## $Gini
## [1] 0.5281
##
## $pic
## TableGrob (1 x 2) "arrange": 2 grobs
## z cells name grob
## pks 1 (1-1,1-1) arrange gtable[layout]
## proc 2 (1-1,2-2) arrange gtable[layout]重抽样
load('/Users/milin/COD\ 建模/model_rf_under1.RData')
dt_pred = predict(model_rf_under, type = 'prob', dt_woe)
perf_eva(dt_woe$label, dt_pred$`1`)
## $KS
## [1] 0.3986
##
## $AUC
## [1] 0.7641
##
## $Gini
## [1] 0.5281
##
## $pic
## TableGrob (1 x 2) "arrange": 2 grobs
## z cells name grob
## pks 1 (1-1,1-1) arrange gtable[layout]
## proc 2 (1-1,2-2) arrange gtable[layout]附录 A: 使用到的包
A.1 library(lattice)
The lattice add-on of Trellis graphics for R. Ref:https://cran.r-project.org/web/packages/lattice/lattice.pdf
A.2 library(knitr)
For Dynamic Report Generation in R. Ref: https://cran.r-project.org/web/packages/knitr/knitr.pdf
A.3 library(gplots)
Various R Programming Tools for Plotting Data. Ref:https://cran.r-project.org/web/packages/gplots/gplots.pdf
A.4 library(ggplot2)
An Implementation of the Grammar of Graphics. Ref: https://cran.rstudio.com/web/packages/ggplot2/ggplot2.pdf
A.5 library(ClustOfVar)
Clustering of variables. Ref: https://cran.r-project.org/web/packages/ClustOfVar/ClustOfVar.pdf
A.6 library(ape)
Analyses of Phylogenetics and Evolution (as.phylo). Ref: https://cran.r-project.org/web/packages/ape/ape.pdf
A.7 library(Information)
Data Exploration with Information Theory (Weight-of-Evidence and Information Value). Ref: https://cran.r-project.org/web/packages/Information/Information.pdf
A.8 library(ROCR)
Visualizing the Performance of Scoring Classifiers. Ref: https://cran.r-project.org/web/packages/ROCR/ROCR.pdf
A.9 library(caret)
Classification and Regression Training - for any machine learning algorithms. Ref: ftp://cran.r-project.org/pub/R/web/packages/caret/caret.pdf
A.10 library(rpart)
Recursive partitioning for classification, regression and survival trees. Ref: https://cran.r-project.org/web/packages/rpart/rpart.pdf
A.10.1 library(rpart.utils)
Tools for parsing and manipulating rpart objects, including generating machine readable rules. Ref: https://cran.r-project.org/web/packages/rpart.utils/rpart.utils.pdf
A.10.2 library(rpart.plot)
Plot ‘rpart’ Models: An Enhanced Version of ‘plot.rpart’. Ref: https://cran.r-project.org/web/packages/knitr/knitr.pdf
A.11 library(randomForest)
Leo Breiman and Cutler’s Random Forests for Classification and Regression. Ref: https://cran.r-project.org/web/packages/randomForest/randomForest.pdf
A.12 library(party)
A computational toolbox for recursive partitioning - Conditional inference Trees. Ref: https://cran.r-project.org/web/packages/party/party.pdf
A.13 library(bnlearn)
Bayesian Network Structure Learning, Parameter Learning and Inference. Ref: https://cran.r-project.org/web/packages/bnlearn/bnlearn.pdf
A.14 library(DAAG)
Data Analysis and Graphics Data and Functions. Ref: https://cran.r-project.org/web/packages/DAAG/DAAG.pdf
A.15 library(vcd)
Visualizing Categorical Data. Ref: https://cran.r-project.org/web/packages/vcd/vcd.pdf
A.16 library(neuralnet)
Neural Network implementation. Ref: https://cran.r-project.org/web/packages/neuralnet/neuralnet.pdf
A.17 library(kernlab)
Kernel-Based Machine Learning Lab. Ref: https://cran.r-project.org/web/packages/kernlab/kernlab.pdf
A.18 library(glmnet)
Lasso and Elastic-Net Regularized Generalized Linear Models. Ref: https://cran.r-project.org/web/packages/glmnet/glmnet.pdf
A.19 library(lars)
Least Angle Regression, Lasso and Forward Stagewise. Ref: ftp://cran.r-project.org/pub/R/web/packages/lars/lars.pdf
参考
WOE变换与信息值
- http://www.ponssard.net/wp-content/uploads/2011/02/on-the-concept-of-the-value-of-information.pdf
- http://research.microsoft.com/en-us/um/people/horvitz/gev.pdf
- http://ucanalytics.com/blogs/information-value-and-weight-of-evidencebanking-case/
- http://www.listendata.com/2015/03/weight-of-evidence-woe-and-information.html
- https://github.com/klarsen1/gampost/blob/master/compare_models.r
- https://www.r-bloggers.com/r-credit-scoring-woe-information-value-in-woe-package/
- http://www.mwsug.org/proceedings/2013/AA/MWSUG-2013-AA14.pdf