lbb inclass

library

library(dplyr)
library(GGally)
library(MLmetrics)

read data copiers

copiers <- read.csv("RM/copiers.csv")
str(copiers)

## 'data.frame':    62 obs. of  15 variables:
##  $ Row.ID      : int  336 393 407 516 596 754 1151 1234 1550 1645 ...
##  $ Order.ID    : Factor w/ 62 levels "CA-2014-116666",..: 17 49 37 39 4 26 15 32 5 14 ...
##  $ Order.Date  : Factor w/ 61 levels "1/22/17","1/4/16",..: 53 58 21 1 52 34 33 37 47 25 ...
##  $ Ship.Date   : Factor w/ 59 levels "1/27/17","1/9/16",..: 57 56 15 1 48 32 31 37 45 25 ...
##  $ Ship.Mode   : Factor w/ 4 levels "First Class",..: 3 3 4 4 4 1 2 1 1 1 ...
##  $ Customer.ID : Factor w/ 59 levels "AB-10255","AD-10180",..: 14 36 41 2 43 59 46 6 25 27 ...
##  $ Segment     : Factor w/ 3 levels "Consumer","Corporate",..: 1 1 1 3 1 2 1 2 1 2 ...
##  $ Product.ID  : Factor w/ 12 levels "TEC-CO-10000971",..: 3 8 11 9 11 7 12 2 2 5 ...
##  $ Category    : Factor w/ 1 level "Technology": 1 1 1 1 1 1 1 1 1 1 ...
##  $ Sub.Category: Factor w/ 1 level "Copiers": 1 1 1 1 1 1 1 1 1 1 ...
##  $ Product.Name: Factor w/ 12 levels "Brother DCP1000 Digital 3 in 1 Multifunction Machine",..: 10 6 12 2 12 9 1 3 3 7 ...
##  $ Sales       : num  960 1800 1200 3000 1200 ...
##  $ Quantity    : int  2 3 3 5 3 3 2 2 1 7 ...
##  $ Discount    : num  0.2 0 0.2 0 0.2 0.2 0 0.4 0.2 0 ...
##  $ Profit      : num  336 702 435 1380 435 ...

copiers %>% 
  group_by(Ship.Mode) %>% 
  summarise(avg_Profit = mean(Profit))

## # A tibble: 4 x 2
##   Ship.Mode      avg_Profit
##   <fct>               <dbl>
## 1 First Class          416.
## 2 Same Day             516.
## 3 Second Class         459.
## 4 Standard Class       444.

ambil beberapa variabel yang memungkinkan mempunyai pengaruh terhadap model

copiers_df <- copiers %>% 
  select(Ship.Mode, Segment, Sales, Quantity, Discount, Profit)

cek korelasi antar variabel

copiers_df %>% 
  select_if(is.numeric) %>% 
  ggcorr(label = T)

cek outlier setiap predictor. untuk sementara kita tidak menghilangkan outlier terlebih dahulu, karena ingin dianalisis lebih dalam

boxplot(copiers_df$Sales)

boxplot(copiers_df$Quantity)

boxplot(copiers_df$Discount)

outlier <- boxplot(copiers_df$Sales, plot = F)$out

buat model full

copiers_model <- lm(Profit~., data = copiers_df)
summary(copiers_model)

## 
## Call:
## lm(formula = Profit ~ ., data = copiers_df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -285.64  -56.21    6.49   66.51  246.10 
## 
## Coefficients:
##                           Estimate Std. Error t value             Pr(>|t|)
## (Intercept)               63.45200   55.06367   1.152                0.254
## Ship.ModeSame Day          9.13795   57.48532   0.159                0.874
## Ship.ModeSecond Class     42.97158   43.92567   0.978                0.332
## Ship.ModeStandard Class   13.92121   38.33509   0.363                0.718
## SegmentCorporate           9.30122   31.29427   0.297                0.767
## SegmentHome Office       -30.70992   39.86033  -0.770                0.444
## Sales                      0.42124    0.02522  16.701 < 0.0000000000000002
## Quantity                 -13.40975   13.54512  -0.990                0.327
## Discount                -874.83099  117.02350  -7.476       0.000000000774
##                            
## (Intercept)                
## Ship.ModeSame Day          
## Ship.ModeSecond Class      
## Ship.ModeStandard Class    
## SegmentCorporate           
## SegmentHome Office         
## Sales                   ***
## Quantity                   
## Discount                ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 102.7 on 53 degrees of freedom
## Multiple R-squared:  0.9515, Adjusted R-squared:  0.9442 
## F-statistic: 130.1 on 8 and 53 DF,  p-value: < 0.00000000000000022

coba step- wise

copiers_backward <- step(copiers_model, direction = "backward")

## Start:  AIC=582.65
## Profit ~ Ship.Mode + Segment + Sales + Quantity + Discount
## 
##             Df Sum of Sq     RSS    AIC
## - Ship.Mode  3     12392  571679 578.01
## - Segment    2     10395  569682 579.79
## - Quantity   1     10343  569630 581.79
## <none>                    559287 582.65
## - Discount   1    589741 1149027 625.29
## - Sales      1   2943279 3502566 694.40
## 
## Step:  AIC=578.01
## Profit ~ Segment + Sales + Quantity + Discount
## 
##            Df Sum of Sq     RSS    AIC
## - Segment   2     15938  587617 575.72
## - Quantity  1     15195  586874 577.64
## <none>                   571679 578.01
## - Discount  1    640224 1211903 622.60
## - Sales     1   3074393 3646072 690.89
## 
## Step:  AIC=575.72
## Profit ~ Sales + Quantity + Discount
## 
##            Df Sum of Sq     RSS    AIC
## - Quantity  1     13134  600751 575.09
## <none>                   587617 575.72
## - Discount  1    635820 1223437 619.18
## - Sales     1   3256039 3843656 690.16
## 
## Step:  AIC=575.09
## Profit ~ Sales + Discount
## 
##            Df Sum of Sq     RSS    AIC
## <none>                   600751 575.09
## - Discount  1    751791 1352543 623.40
## - Sales     1   8783232 9383984 743.50

summary(copiers_backward)

## 
## Call:
## lm(formula = Profit ~ Sales + Discount, data = copiers_df)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -277.69  -62.73    6.89   66.79  275.86 
## 
## Coefficients:
##               Estimate Std. Error t value             Pr(>|t|)    
## (Intercept)   66.21587   30.34849   2.182               0.0331 *  
## Sales          0.40019    0.01363  29.370 < 0.0000000000000002 ***
## Discount    -894.82460  104.13831  -8.593     0.00000000000549 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 100.9 on 59 degrees of freedom
## Multiple R-squared:  0.9479, Adjusted R-squared:  0.9462 
## F-statistic: 537.2 on 2 and 59 DF,  p-value: < 0.00000000000000022

RMSE(copiers_backward$fitted.values, copiers_df$Profit)

## [1] 98.43544

MAE(copiers_backward$fitted.values, copiers_df$Profit)

## [1] 75.08537

MAPE(copiers_backward$fitted.values, copiers_df$Profit)

## [1] 0.3009575

range(copiers_df$Profit)

## [1]   59.998 2302.967

pengaruh outlier

plot(copiers_backward)

residual menyebar normal (normality)

shapiro.test(copiers_backward$residuals)

## 
##  Shapiro-Wilk normality test
## 
## data:  copiers_backward$residuals
## W = 0.9834, p-value = 0.5656

homoscedasticity

lmtest::bptest(copiers_backward)

## 
##  studentized Breusch-Pagan test
## 
## data:  copiers_backward
## BP = 16.869, df = 2, p-value = 0.0002172

plot(copiers_backward, which = 1)

multicollinearity

car::vif(copiers_backward)

##    Sales Discount 
## 1.038959 1.038959

cor.test(copiers_df$Sales, copiers_df$Profit)

## 
##  Pearson's product-moment correlation
## 
## data:  copiers_df$Sales and copiers_df$Profit
## t = 21.26, df = 60, p-value < 0.00000000000000022
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.9013320 0.9632858
## sample estimates:
##       cor 
## 0.9395785

cor.test(copiers_df$Discount, copiers_df$Profit)

## 
##  Pearson's product-moment correlation
## 
## data:  copiers_df$Discount and copiers_df$Profit
## t = -3.7139, df = 60, p-value = 0.0004496
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.6156283 -0.2046714
## sample estimates:
##        cor 
## -0.4323383

LBB RM : 1. lakukan eksplorasi data - str() - cor() - group_by() + summarise() utk factor terhadap y

feature selection berdasarkan bisnis
modeling

interpretasi koefisien
bahas adj R-Square
bahas signifikansi predictor

feature selection

stepwise
regsubset

Eror

RMSE / MSE
MAE
MAPE

asumsi

normality
homoscedasticity
Multicollinearity
linearity

Kesimpulan