O objetivo deste script é demonstrar a criação de um modelo preditivo no formato de árvore de decisão para identificar spam. Para tanto, será utilizado o dataset disponibilizado em http://archive.ics.uci.edu/ml/datasets/Spambase
Carga dos dados:
spambase <- read.csv("../data/spambase.data", header=F)
names(spambase) <- c('word_freq_make',
'word_freq_address',
'word_freq_all',
'word_freq_3d',
'word_freq_our',
'word_freq_over',
'word_freq_remove',
'word_freq_internet',
'word_freq_order',
'word_freq_mail',
'word_freq_receive',
'word_freq_will',
'word_freq_people',
'word_freq_report',
'word_freq_addresses',
'word_freq_free',
'word_freq_business',
'word_freq_email',
'word_freq_you',
'word_freq_credit',
'word_freq_your',
'word_freq_font',
'word_freq_000',
'word_freq_money',
'word_freq_hp',
'word_freq_hpl',
'word_freq_george',
'word_freq_650',
'word_freq_lab',
'word_freq_labs',
'word_freq_telnet',
'word_freq_857',
'word_freq_data',
'word_freq_415',
'word_freq_85',
'word_freq_technology',
'word_freq_1999',
'word_freq_parts',
'word_freq_pm',
'word_freq_direct',
'word_freq_cs',
'word_freq_meeting',
'word_freq_original',
'word_freq_project',
'word_freq_re',
'word_freq_edu',
'word_freq_table',
'word_freq_conference',
'char_freq_;',
'char_freq_(',
'char_freq_[',
'char_freq_!',
'char_freq_$',
'char_freq_#',
'capital_run_length_average',
'capital_run_length_longest',
'capital_run_length_total', 'spam')
sapply(spambase,class)
## word_freq_make word_freq_address
## "numeric" "numeric"
## word_freq_all word_freq_3d
## "numeric" "numeric"
## word_freq_our word_freq_over
## "numeric" "numeric"
## word_freq_remove word_freq_internet
## "numeric" "numeric"
## word_freq_order word_freq_mail
## "numeric" "numeric"
## word_freq_receive word_freq_will
## "numeric" "numeric"
## word_freq_people word_freq_report
## "numeric" "numeric"
## word_freq_addresses word_freq_free
## "numeric" "numeric"
## word_freq_business word_freq_email
## "numeric" "numeric"
## word_freq_you word_freq_credit
## "numeric" "numeric"
## word_freq_your word_freq_font
## "numeric" "numeric"
## word_freq_000 word_freq_money
## "numeric" "numeric"
## word_freq_hp word_freq_hpl
## "numeric" "numeric"
## word_freq_george word_freq_650
## "numeric" "numeric"
## word_freq_lab word_freq_labs
## "numeric" "numeric"
## word_freq_telnet word_freq_857
## "numeric" "numeric"
## word_freq_data word_freq_415
## "numeric" "numeric"
## word_freq_85 word_freq_technology
## "numeric" "numeric"
## word_freq_1999 word_freq_parts
## "numeric" "numeric"
## word_freq_pm word_freq_direct
## "numeric" "numeric"
## word_freq_cs word_freq_meeting
## "numeric" "numeric"
## word_freq_original word_freq_project
## "numeric" "numeric"
## word_freq_re word_freq_edu
## "numeric" "numeric"
## word_freq_table word_freq_conference
## "numeric" "numeric"
## char_freq_; char_freq_(
## "numeric" "numeric"
## char_freq_[ char_freq_!
## "numeric" "numeric"
## char_freq_$ char_freq_#
## "numeric" "numeric"
## capital_run_length_average capital_run_length_longest
## "numeric" "integer"
## capital_run_length_total spam
## "integer" "integer"
spambase$spam <- as.factor(paste(spambase$spam))
Divisão entre conjunto de treinamento e conjunto de teste.
set.seed(1234)
ind <- sample(2, nrow(spambase), replace = TRUE, prob = c(0.7, 0.3))
trainData <- spambase[ind == 1, ]
testData <- spambase[ind == 2, ]
table(trainData$spam)
##
## 0 1
## 1944 1297
Criação do modelo:
library(RWeka)
spam_model_j48 <- J48(spam ~ ., data = trainData)
print(spam_model_j48)
## J48 pruned tree
## ------------------
##
## word_freq_remove <= 0
## | char_freq_$ <= 0.058
## | | word_freq_000 <= 0.25
## | | | char_freq_! <= 0.355
## | | | | word_freq_money <= 0.06
## | | | | | word_freq_free <= 0.1
## | | | | | | word_freq_george <= 0
## | | | | | | | word_freq_hp <= 0
## | | | | | | | | word_freq_business <= 0.05
## | | | | | | | | | capital_run_length_longest <= 10: 0 (505.0/20.0)
## | | | | | | | | | capital_run_length_longest > 10
## | | | | | | | | | | word_freq_our <= 0.71
## | | | | | | | | | | | word_freq_re <= 0.05
## | | | | | | | | | | | | word_freq_you <= 3.67
## | | | | | | | | | | | | | capital_run_length_average <= 7.836
## | | | | | | | | | | | | | | word_freq_you <= 0.03
## | | | | | | | | | | | | | | | word_freq_font <= 0.07
## | | | | | | | | | | | | | | | | word_freq_address <= 0.16
## | | | | | | | | | | | | | | | | | char_freq_! <= 0.139
## | | | | | | | | | | | | | | | | | | word_freq_650 <= 0.32: 0 (50.0/4.0)
## | | | | | | | | | | | | | | | | | | word_freq_650 > 0.32: 1 (9.0/2.0)
## | | | | | | | | | | | | | | | | | char_freq_! > 0.139: 1 (4.0/1.0)
## | | | | | | | | | | | | | | | | word_freq_address > 0.16: 1 (5.0/1.0)
## | | | | | | | | | | | | | | | word_freq_font > 0.07: 1 (4.0)
## | | | | | | | | | | | | | | word_freq_you > 0.03
## | | | | | | | | | | | | | | | word_freq_over <= 0.54: 0 (72.0/1.0)
## | | | | | | | | | | | | | | | word_freq_over > 0.54
## | | | | | | | | | | | | | | | | word_freq_over <= 0.78: 0 (2.0)
## | | | | | | | | | | | | | | | | word_freq_over > 0.78: 1 (2.0)
## | | | | | | | | | | | | | capital_run_length_average > 7.836: 1 (7.0/1.0)
## | | | | | | | | | | | | word_freq_you > 3.67
## | | | | | | | | | | | | | word_freq_edu <= 0.54: 1 (12.0/2.0)
## | | | | | | | | | | | | | word_freq_edu > 0.54: 0 (2.0)
## | | | | | | | | | | | word_freq_re > 0.05: 0 (66.0)
## | | | | | | | | | | word_freq_our > 0.71
## | | | | | | | | | | | word_freq_re <= 0.16
## | | | | | | | | | | | | word_freq_will <= 1.25: 1 (12.0)
## | | | | | | | | | | | | word_freq_will > 1.25: 0 (3.0/1.0)
## | | | | | | | | | | | word_freq_re > 0.16: 0 (2.0)
## | | | | | | | | word_freq_business > 0.05
## | | | | | | | | | word_freq_650 <= 0.6
## | | | | | | | | | | char_freq_; <= 0.069
## | | | | | | | | | | | char_freq_! <= 0.067: 0 (14.0)
## | | | | | | | | | | | char_freq_! > 0.067: 1 (5.0/1.0)
## | | | | | | | | | | char_freq_; > 0.069: 1 (4.0)
## | | | | | | | | | word_freq_650 > 0.6: 1 (2.0)
## | | | | | | | word_freq_hp > 0: 0 (402.0/5.0)
## | | | | | | word_freq_george > 0: 0 (465.0)
## | | | | | word_freq_free > 0.1
## | | | | | | word_freq_george <= 0.02
## | | | | | | | word_freq_our <= 1.14
## | | | | | | | | word_freq_will <= 1
## | | | | | | | | | word_freq_receive <= 0.24
## | | | | | | | | | | word_freq_email <= 0.25
## | | | | | | | | | | | word_freq_you <= 0.25: 1 (20.0/4.0)
## | | | | | | | | | | | word_freq_you > 0.25
## | | | | | | | | | | | | word_freq_credit <= 0.6
## | | | | | | | | | | | | | capital_run_length_average <= 3.816: 0 (56.0/8.0)
## | | | | | | | | | | | | | capital_run_length_average > 3.816: 1 (8.0/1.0)
## | | | | | | | | | | | | word_freq_credit > 0.6: 1 (3.0)
## | | | | | | | | | | word_freq_email > 0.25: 0 (20.0/1.0)
## | | | | | | | | | word_freq_receive > 0.24: 1 (8.0/1.0)
## | | | | | | | | word_freq_will > 1: 0 (18.0)
## | | | | | | | word_freq_our > 1.14
## | | | | | | | | char_freq_! <= 0.279: 1 (23.0/1.0)
## | | | | | | | | char_freq_! > 0.279: 0 (2.0)
## | | | | | | word_freq_george > 0.02: 0 (27.0)
## | | | | word_freq_money > 0.06
## | | | | | word_freq_hp <= 0.05
## | | | | | | word_freq_edu <= 0.2
## | | | | | | | capital_run_length_longest <= 9: 0 (9.0/1.0)
## | | | | | | | capital_run_length_longest > 9: 1 (20.0)
## | | | | | | word_freq_edu > 0.2: 0 (6.0)
## | | | | | word_freq_hp > 0.05: 0 (7.0)
## | | | char_freq_! > 0.355
## | | | | capital_run_length_total <= 49
## | | | | | word_freq_free <= 2.32
## | | | | | | word_freq_money <= 0.23
## | | | | | | | char_freq_! <= 1.501
## | | | | | | | | capital_run_length_longest <= 8: 0 (63.0/1.0)
## | | | | | | | | capital_run_length_longest > 8
## | | | | | | | | | capital_run_length_average <= 3.26
## | | | | | | | | | | capital_run_length_longest <= 9
## | | | | | | | | | | | word_freq_85 <= 1.51
## | | | | | | | | | | | | word_freq_will <= 1.61: 1 (3.0)
## | | | | | | | | | | | | word_freq_will > 1.61: 0 (3.0)
## | | | | | | | | | | | word_freq_85 > 1.51: 0 (2.0)
## | | | | | | | | | | capital_run_length_longest > 9: 0 (15.0)
## | | | | | | | | | capital_run_length_average > 3.26: 1 (3.0)
## | | | | | | | char_freq_! > 1.501
## | | | | | | | | word_freq_re <= 0.47
## | | | | | | | | | char_freq_! <= 3.333
## | | | | | | | | | | capital_run_length_total <= 6: 0 (3.0)
## | | | | | | | | | | capital_run_length_total > 6
## | | | | | | | | | | | word_freq_you <= 2.17
## | | | | | | | | | | | | char_freq_! <= 1.765: 1 (2.0)
## | | | | | | | | | | | | char_freq_! > 1.765: 0 (2.0)
## | | | | | | | | | | | word_freq_you > 2.17: 1 (2.0)
## | | | | | | | | | char_freq_! > 3.333: 1 (7.0)
## | | | | | | | | word_freq_re > 0.47: 0 (4.0)
## | | | | | | word_freq_money > 0.23
## | | | | | | | word_freq_all <= 0.23: 1 (4.0)
## | | | | | | | word_freq_all > 0.23: 0 (2.0)
## | | | | | word_freq_free > 2.32: 1 (11.0)
## | | | | capital_run_length_total > 49
## | | | | | word_freq_pm <= 0.19
## | | | | | | char_freq_! <= 0.564
## | | | | | | | word_freq_hp <= 0.22
## | | | | | | | | word_freq_meeting <= 0.09
## | | | | | | | | | word_freq_re <= 0.47: 1 (35.0/6.0)
## | | | | | | | | | word_freq_re > 0.47: 0 (5.0/1.0)
## | | | | | | | | word_freq_meeting > 0.09: 0 (3.0/1.0)
## | | | | | | | word_freq_hp > 0.22: 0 (3.0)
## | | | | | | char_freq_! > 0.564: 1 (96.0/5.0)
## | | | | | word_freq_pm > 0.19: 0 (8.0/1.0)
## | | word_freq_000 > 0.25
## | | | word_freq_re <= 0.3
## | | | | word_freq_technology <= 0.09
## | | | | | word_freq_our <= 0.83: 1 (40.0/1.0)
## | | | | | word_freq_our > 0.83
## | | | | | | word_freq_all <= 0.56: 0 (2.0)
## | | | | | | word_freq_all > 0.56: 1 (2.0)
## | | | | word_freq_technology > 0.09: 0 (3.0/1.0)
## | | | word_freq_re > 0.3: 0 (4.0/1.0)
## | char_freq_$ > 0.058
## | | word_freq_hp <= 0.4
## | | | capital_run_length_average <= 1.987
## | | | | word_freq_meeting <= 0.35
## | | | | | word_freq_1999 <= 0.16
## | | | | | | char_freq_; <= 0.227
## | | | | | | | char_freq_( <= 0.28
## | | | | | | | | word_freq_george <= 0.32
## | | | | | | | | | word_freq_receive <= 0.45: 1 (37.0/4.0)
## | | | | | | | | | word_freq_receive > 0.45: 0 (3.0/1.0)
## | | | | | | | | word_freq_george > 0.32: 0 (2.0)
## | | | | | | | char_freq_( > 0.28: 0 (4.0)
## | | | | | | char_freq_; > 0.227: 0 (3.0)
## | | | | | word_freq_1999 > 0.16: 0 (6.0)
## | | | | word_freq_meeting > 0.35: 0 (7.0)
## | | | capital_run_length_average > 1.987
## | | | | word_freq_1999 <= 0.06: 1 (335.0/7.0)
## | | | | word_freq_1999 > 0.06
## | | | | | word_freq_edu <= 0.26
## | | | | | | char_freq_$ <= 0.097: 0 (2.0)
## | | | | | | char_freq_$ > 0.097: 1 (13.0/1.0)
## | | | | | word_freq_edu > 0.26: 0 (3.0)
## | | word_freq_hp > 0.4: 0 (41.0/1.0)
## word_freq_remove > 0
## | word_freq_85 <= 0.07
## | | word_freq_edu <= 0.07: 1 (522.0/19.0)
## | | word_freq_edu > 0.07
## | | | word_freq_over <= 0.15: 0 (5.0)
## | | | word_freq_over > 0.15: 1 (29.0)
## | word_freq_85 > 0.07
## | | char_freq_! <= 0.079: 0 (9.0)
## | | char_freq_! > 0.079: 1 (17.0)
##
## Number of Leaves : 79
##
## Size of the tree : 157
Validação do modelo usando o próprio conjunto de treinamento:
t <- table(predict(spam_model_j48), trainData$spam)
acuracia_treinamento <- (t[1, 1] + t[2, 2])/sum(t)
t
##
## 0 1
## 0 1886 49
## 1 58 1248
acuracia_treinamento
## [1] 0.967
Validação do modelo usando o conjunto de testes:
testPred <- predict(spam_model_j48, newdata = testData)
t <- table(testPred, testData$spam)
acuracia_teste <- (t[1, 1] + t[2, 2])/sum(t)
t
##
## testPred 0 1
## 0 789 54
## 1 55 462
acuracia_teste
## [1] 0.9199
Este material faz parte do curso sobre Web Data Mining com R