To capture the high sales attribute information :
install.packages("rmarkdown",repos = "http://cran.us.r-project.org")
## Installing package into 'C:/Users/tswaminathan/Documents/R/win-library/3.5'
## (as 'lib' is unspecified)
## package 'rmarkdown' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\tswaminathan\AppData\Local\Temp\Rtmp2xR758\downloaded_packages
install.packages("C50",repos = "http://cran.us.r-project.org")
## Installing package into 'C:/Users/tswaminathan/Documents/R/win-library/3.5'
## (as 'lib' is unspecified)
## package 'C50' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\tswaminathan\AppData\Local\Temp\Rtmp2xR758\downloaded_packages
install.packages("tree",repos = "http://cran.us.r-project.org")
## Installing package into 'C:/Users/tswaminathan/Documents/R/win-library/3.5'
## (as 'lib' is unspecified)
## package 'tree' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\tswaminathan\AppData\Local\Temp\Rtmp2xR758\downloaded_packages
install.packages("caret",repos = "http://cran.us.r-project.org")
## Installing package into 'C:/Users/tswaminathan/Documents/R/win-library/3.5'
## (as 'lib' is unspecified)
## package 'caret' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\tswaminathan\AppData\Local\Temp\Rtmp2xR758\downloaded_packages
install.packages("gmodels",repos = "http://cran.us.r-project.org")
## Installing package into 'C:/Users/tswaminathan/Documents/R/win-library/3.5'
## (as 'lib' is unspecified)
## package 'gmodels' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\tswaminathan\AppData\Local\Temp\Rtmp2xR758\downloaded_packages
install.packages("party",repos = "http://cran.us.r-project.org")
## Installing package into 'C:/Users/tswaminathan/Documents/R/win-library/3.5'
## (as 'lib' is unspecified)
## package 'party' successfully unpacked and MD5 sums checked
##
## The downloaded binary packages are in
## C:\Users\tswaminathan\AppData\Local\Temp\Rtmp2xR758\downloaded_packages
library(party)
## Loading required package: grid
## Loading required package: mvtnorm
## Loading required package: modeltools
## Loading required package: stats4
## Loading required package: strucchange
## Loading required package: zoo
##
## Attaching package: 'zoo'
## The following objects are masked from 'package:base':
##
## as.Date, as.Date.numeric
## Loading required package: sandwich
library(caret)
## Loading required package: lattice
## Loading required package: ggplot2
library(C50)
library(tree)
library(gmodels)
CompanyData <- read.csv(file.choose())
# Splitting data into training and testing.
# splitting the data based on Sales
hist(CompanyData$Sales)
High = ifelse(CompanyData$Sales<10, "No", "Yes")
CD = data.frame(CompanyData, High)
#CD <- CompanyData[,2:12]
# View(CD)
CD_train <- CD[1:200,]
# View(CD_train)
CD_test <- CD[201:400,]
# View(CD_test)
#Using Party Function
op_tree = ctree(High ~ CompPrice + Income + Advertising + Population + Price + ShelveLoc
+ Age + Education + Urban + US, data = CD_train)
summary(op_tree)
## Length Class Mode
## 1 BinaryTree S4
plot(op_tree)
# On looking into the Above tree, i see that if the Location of the Shelv is good,
# then there is a probability of 60% chance that the customer will buy.
# With ShelveLoc having a Bad or Medium and Price <= 87, the probability of High sales
# could be 60%.
# If ShelveLoc is Bad or Medium, With Price >= 87 and Advertising less then <= 7 then there
# is a zero percent chance of high sales.
# If ShelveLoc is Bad or Medium, With Price >= 87 and Advertising less then > 7 then there
# is a 20 % percent chance of high sales.
pred_tree <- as.data.frame(predict(op_tree,newdata=CD_test))
pred_tree["final"] <- NULL
pred_test_df <- predict(op_tree,newdata=CD_test)
mean(pred_test_df==CD$High) # Accuracy = 68.75%
## [1] 0.6875
CrossTable(CD_test$High,pred_test_df)
##
##
## Cell Contents
## |-------------------------|
## | N |
## | Chi-square contribution |
## | N / Row Total |
## | N / Col Total |
## | N / Table Total |
## |-------------------------|
##
##
## Total Observations in Table: 200
##
##
## | pred_test_df
## CD_test$High | No | Yes | Row Total |
## -------------|-----------|-----------|-----------|
## No | 131 | 31 | 162 |
## | 2.468 | 5.899 | |
## | 0.809 | 0.191 | 0.810 |
## | 0.929 | 0.525 | |
## | 0.655 | 0.155 | |
## -------------|-----------|-----------|-----------|
## Yes | 10 | 28 | 38 |
## | 10.523 | 25.148 | |
## | 0.263 | 0.737 | 0.190 |
## | 0.071 | 0.475 | |
## | 0.050 | 0.140 | |
## -------------|-----------|-----------|-----------|
## Column Total | 141 | 59 | 200 |
## | 0.705 | 0.295 | |
## -------------|-----------|-----------|-----------|
##
##
confusionMatrix(CD_test$High,pred_test_df)
## Confusion Matrix and Statistics
##
## Reference
## Prediction No Yes
## No 131 31
## Yes 10 28
##
## Accuracy : 0.795
## 95% CI : (0.7323, 0.8487)
## No Information Rate : 0.705
## P-Value [Acc > NIR] : 0.002590
##
## Kappa : 0.4503
## Mcnemar's Test P-Value : 0.001787
##
## Sensitivity : 0.9291
## Specificity : 0.4746
## Pos Pred Value : 0.8086
## Neg Pred Value : 0.7368
## Prevalence : 0.7050
## Detection Rate : 0.6550
## Detection Prevalence : 0.8100
## Balanced Accuracy : 0.7018
##
## 'Positive' Class : No
##
##### Using tree function
cd_tree_org <- tree(High~.-Sales,data=CD)
summary(cd_tree_org)
##
## Classification tree:
## tree(formula = High ~ . - Sales, data = CD)
## Variables actually used in tree construction:
## [1] "ShelveLoc" "Price" "Advertising" "Age" "CompPrice"
## [6] "Population" "Income"
## Number of terminal nodes: 21
## Residual mean deviance: 0.297 = 112.6 / 379
## Misclassification error rate: 0.0725 = 29 / 400
plot(cd_tree_org)
text(cd_tree_org,pretty = 0)
# Using the training data
##### Using tree function
cd_tree <- tree(High~.-Sales,data=CD_train)
summary(cd_tree)
##
## Classification tree:
## tree(formula = High ~ . - Sales, data = CD_train)
## Variables actually used in tree construction:
## [1] "ShelveLoc" "Price" "Advertising" "Age" "CompPrice"
## [6] "Income"
## Number of terminal nodes: 12
## Residual mean deviance: 0.2927 = 55.02 / 188
## Misclassification error rate: 0.08 = 16 / 200
plot(cd_tree)
text(cd_tree,pretty = 0)
### Evaluate the Model
# Predicting the test data using the model
pred_tree <- as.data.frame(predict(cd_tree,newdata=CD_test))
pred_tree["final"] <- NULL
pred_test_df <- predict(cd_tree,newdata=CD_test)
pred_tree$final <- colnames(pred_test_df)[apply(pred_test_df,1,which.max)]
pred_tree$final <- as.factor(pred_tree$final)
summary(pred_tree$final)
## No Yes
## 172 28
summary(CD_test$High)
## No Yes
## 162 38
mean(pred_tree$final==CD$High) # Accuracy = 77.25
## [1] 0.7725
CrossTable(CD_test$High,pred_tree$final)
##
##
## Cell Contents
## |-------------------------|
## | N |
## | Chi-square contribution |
## | N / Row Total |
## | N / Col Total |
## | N / Table Total |
## |-------------------------|
##
##
## Total Observations in Table: 200
##
##
## | pred_tree$final
## CD_test$High | No | Yes | Row Total |
## -------------|-----------|-----------|-----------|
## No | 153 | 9 | 162 |
## | 1.343 | 8.251 | |
## | 0.944 | 0.056 | 0.810 |
## | 0.890 | 0.321 | |
## | 0.765 | 0.045 | |
## -------------|-----------|-----------|-----------|
## Yes | 19 | 19 | 38 |
## | 5.727 | 35.177 | |
## | 0.500 | 0.500 | 0.190 |
## | 0.110 | 0.679 | |
## | 0.095 | 0.095 | |
## -------------|-----------|-----------|-----------|
## Column Total | 172 | 28 | 200 |
## | 0.860 | 0.140 | |
## -------------|-----------|-----------|-----------|
##
##
confusionMatrix(CD_test$High,pred_tree$final)
## Confusion Matrix and Statistics
##
## Reference
## Prediction No Yes
## No 153 9
## Yes 19 19
##
## Accuracy : 0.86
## 95% CI : (0.8041, 0.9049)
## No Information Rate : 0.86
## P-Value [Acc > NIR] : 0.55018
##
## Kappa : 0.4942
## Mcnemar's Test P-Value : 0.08897
##
## Sensitivity : 0.8895
## Specificity : 0.6786
## Pos Pred Value : 0.9444
## Neg Pred Value : 0.5000
## Prevalence : 0.8600
## Detection Rate : 0.7650
## Detection Prevalence : 0.8100
## Balanced Accuracy : 0.7841
##
## 'Positive' Class : No
##