1.1 데이터 사전

캐글 Women’s e-commerce cloting reviews 데이터는 총 11개 변수로 구성되어 있고 관측점이 23,486개로 구성되어 있다. Recommended IND를 라벨 목표변수로 두고 예측모형을 구축해보자.

• Clothing ID: Integer Categorical variable that refers to the specific piece being reviewed.
• Age: Positive Integer variable of the reviewers age.
• Title: String variable for the title of the review.
• Review Text: String variable for the review body.
• Rating: Positive Ordinal Integer variable for the product score granted by the customer from 1 Worst, to 5 Best.
• Recommended IND: Binary variable stating where the customer recommends the product where 1 is recommended, 0 is not - recommended.
• Positive Feedback Count: Positive Integer documenting the number of other customers who found this review positive.
• Division Name: Categorical name of the product high level division.
• Department Name: Categorical name of the product department name.
• Class Name: Categorical name of the product class name.

library(tidyverse)
library(janitor)

cloth_dat <- read_csv("data/Womens Clothing E-Commerce Reviews.csv")

cloth_dat %>% 
  clean_names() %>% 
  sample_n(100) %>% 
  DT::datatable()

1.2 데이터 전처리

cloth_dat <- cloth_dat %>% 
  clean_names() %>% 
  filter(complete.cases(.)) %>% 
  rename(y = recommended_ind)

cloth_df <- cloth_dat %>% 
  mutate(y = factor(y, levels=c(1,0), labels=c("yes", "no"))) %>% 
  mutate_if(is.character, as.factor) %>% 
  select(y, age, title, review_text, division_name, department_name, class_name) %>% 
  mutate(class = fct_lump(class_name, 9)) %>% 
  select(y, age, division = division_name, department = department_name, class)

library(tm)
library(text2vec)
library(SnowballC)
library(tidytext)

cloth_dat %>% 
  # unnest_tokens(output=word, input = title) %>% 
  # anti_join(get_stopwords()) %>% 
  cast_dtm(review_text, term, count)

2.1 텍스트 제외 예측모형

# 2. 예측모형 -----
## 2.1. 훈련/시험 데이터 분할 ------
library(caret)

xy_index <- createDataPartition(cloth_df$y, times =1, p=0.5, list=FALSE)

train_df <- cloth_df[xy_index, ]
test_df  <- cloth_df[-xy_index, ]

## 2.2. 모형 개발/검증 데이터셋 준비 ------
cv_folds <- createMultiFolds(train_df$y, k = 5, times = 1)

cv_cntrl <- trainControl(method = "repeatedcv", number = 5,
                         sampling = "down",
                         summaryFunction = twoClassSummary,
                         classProbs = TRUE,
                         repeats = 1, index = cv_folds)


## 2.2. 모형 개발/검증 데이터셋 준비 ------
library(doSNOW)
# 실행시간
start.time <- Sys.time()

cl <- makeCluster(4, type = "SOCK")
registerDoSNOW(cl)

cloth_glm   <- train(y ~ ., data = train_df, 
                    method = "glm",
                    family = "binomial",
                    metric='Sens',
                    trControl = cv_cntrl, 
                    tuneLength = 7)

cloth_rf   <- train(y ~ ., data = train_df, 
                    method = "ranger",
                    metric='Sens',
                    num.tree = 100,
                    importance = "permutation",
                    trControl = cv_cntrl, 
                    tuneLength = 7)


stopCluster(cl)

total.time <- Sys.time() - start.time
total.time

Time difference of 31.08911 secs

# 3. 예측모형 성능 -----
## GLM
glm_pred_df <- predict(cloth_glm, newdata=test_df, type="prob") %>%
  tbl_df %>% 
  mutate(class = factor(ifelse(yes > no, "yes", "no"), levels = c("yes", "no")),
         prob  = yes)

confusionMatrix(glm_pred_df$class, test_df$y)

Confusion Matrix and Statistics

          Reference
Prediction  yes   no
       yes 3881  756
       no  4162 1031
                                          
               Accuracy : 0.4997          
                 95% CI : (0.4898, 0.5096)
    No Information Rate : 0.8182          
    P-Value [Acc > NIR] : 1               
                                          
                  Kappa : 0.0342          
 Mcnemar's Test P-Value : <2e-16          
                                          
            Sensitivity : 0.4825          
            Specificity : 0.5769          
         Pos Pred Value : 0.8370          
         Neg Pred Value : 0.1985          
             Prevalence : 0.8182          
         Detection Rate : 0.3948          
   Detection Prevalence : 0.4717          
      Balanced Accuracy : 0.5297          
                                          
       'Positive' Class : yes             
                                          

## randomForest
rf_pred_df <- predict(cloth_rf, newdata=test_df) %>% 
  tbl_df %>% 
  rename(class = value)

confusionMatrix(rf_pred_df$class, test_df$y)

Confusion Matrix and Statistics

          Reference
Prediction  yes   no
       yes 3917  794
       no  4126  993
                                          
               Accuracy : 0.4995          
                 95% CI : (0.4896, 0.5094)
    No Information Rate : 0.8182          
    P-Value [Acc > NIR] : 1               
                                          
                  Kappa : 0.0247          
 Mcnemar's Test P-Value : <2e-16          
                                          
            Sensitivity : 0.4870          
            Specificity : 0.5557          
         Pos Pred Value : 0.8315          
         Neg Pred Value : 0.1940          
             Prevalence : 0.8182          
         Detection Rate : 0.3985          
   Detection Prevalence : 0.4792          
      Balanced Accuracy : 0.5213          
                                          
       'Positive' Class : yes