Comprehensive Analysis for Laptop Prices & Model Building

Exploring Data Insights, Predictive Modelling and Interative Applications

Explanatory Data Analysis
Predictive Modelling
Interactive Application
Author

Data Analyst - Pythias C

Published

March 3, 2024

1 OVERVIEW

In this mini project focused on laptop prices, a structured approach was employed, encompassing data cleaning, exploratory data analysis (EDA), feature engineering, regression model building, and the development of an interactive Shiny application. The data cleaning phase ensured the dataset’s integrity and usability, while EDA provided valuable insights into the characteristics and relationships within the dataset. Feature engineering enhanced the predictive capabilities of the model, leading to the development of a regression model to understand the determinants of laptop prices. The culmination of the project was the creation of an interactive Shiny application, allowing for dynamic exploration and visualization of the model’s predictions. This project serves as a comprehensive example of the end-to-end process of data analysis and model deployment, highlighting the multifaceted nature of predictive analytics in the realm of pricing.

2 INTRODUCTION

The mini project on laptop prices represents a comprehensive endeavor encompassing various stages of data analysis and model development. Beginning with data cleaning, the project involved meticulous preparation of the dataset to ensure its reliability and suitability for subsequent analysis. Following this, the exploratory data analysis (EDA) phase provided valuable insights into the characteristics and relationships within the dataset, laying the foundation for further exploration.Subsequently, feature engineering was carried out to enhance the predictive capabilities of the model, focusing on creating new features and transforming existing ones to improve its performance. The regression model building phase aimed to establish a predictive relationship between the features and the target variable, which in this case is laptop prices.The culmination of the project involved the development of an interactive Shiny application, enabling dynamic exploration and visualization of the model’s predictions. This interactive tool facilitated user engagement and provided a platform for gaining insights into the factors influencing laptop prices.Overall, this mini project serves as a demonstration of the end-to-end process of data analysis and model deployment, highlighting the multifaceted nature of predictive analytics in the context of pricing.

3 ABOUT DATASET

The dataset that is going to be used for various tasks of EDA is from Kaggle. The link to the dataset is attached below:

https://www.kaggle.com/datasets/muhammetvarl/laptop-price

4 READING DATASET

Code 
laptop=read.csv(file.choose()) #reading dataset

library(janitor)

laptop=clean_names(laptop[2:12]) #Cleaning & keeping important variables

5 DATA DESCRIPTION

The first 5 rows of the data

Code 
library(knitr)
head(laptop,5)%>% kable()  #first 5 rows
company type_name inches screen_resolution cpu ram memory gpu op_sys weight price
Apple Ultrabook 13.3 IPS Panel Retina Display 2560x1600 Intel Core i5 2.3GHz 8GB 128GB SSD Intel Iris Plus Graphics 640 macOS 1.37kg 71378.68
Apple Ultrabook 13.3 1440x900 Intel Core i5 1.8GHz 8GB 128GB Flash Storage Intel HD Graphics 6000 macOS 1.34kg 47895.52
HP Notebook 15.6 Full HD 1920x1080 Intel Core i5 7200U 2.5GHz 8GB 256GB SSD Intel HD Graphics 620 No OS 1.86kg 30636.00
Apple Ultrabook 15.4 IPS Panel Retina Display 2880x1800 Intel Core i7 2.7GHz 16GB 512GB SSD AMD Radeon Pro 455 macOS 1.83kg 135195.34
Apple Ultrabook 13.3 IPS Panel Retina Display 2560x1600 Intel Core i5 3.1GHz 8GB 256GB SSD Intel Iris Plus Graphics 650 macOS 1.37kg 96095.81

Column names of the dataset

Code 
colnames(laptop) %>% kable() #column names
x
company
type_name
inches
screen_resolution
cpu
ram
memory
gpu
op_sys
weight
price

Classes of dataset

Code 
str(laptop) %>% kable()#dataset classes
'data.frame':   1303 obs. of  11 variables:
 $ company          : chr  "Apple" "Apple" "HP" "Apple" ...
 $ type_name        : chr  "Ultrabook" "Ultrabook" "Notebook" "Ultrabook" ...
 $ inches           : num  13.3 13.3 15.6 15.4 13.3 15.6 15.4 13.3 14 14 ...
 $ screen_resolution: chr  "IPS Panel Retina Display 2560x1600" "1440x900" "Full HD 1920x1080" "IPS Panel Retina Display 2880x1800" ...
 $ cpu              : chr  "Intel Core i5 2.3GHz" "Intel Core i5 1.8GHz" "Intel Core i5 7200U 2.5GHz" "Intel Core i7 2.7GHz" ...
 $ ram              : chr  "8GB" "8GB" "8GB" "16GB" ...
 $ memory           : chr  "128GB SSD" "128GB Flash Storage" "256GB SSD" "512GB SSD" ...
 $ gpu              : chr  "Intel Iris Plus Graphics 640" "Intel HD Graphics 6000" "Intel HD Graphics 620" "AMD Radeon Pro 455" ...
 $ op_sys           : chr  "macOS" "macOS" "No OS" "macOS" ...
 $ weight           : chr  "1.37kg" "1.34kg" "1.86kg" "1.83kg" ...
 $ price            : num  71379 47896 30636 135195 96096 ...
  • 9 character variables and 2 numeric variables

Variable Conversion

Converting variable names memory, weight & ram to be in numerical

Code 
#variable conversion
library(dplyr)

laptop$ram=as.numeric(sub("GB","",laptop$ram))
laptop$weight=as.numeric(sub("kg","",laptop$weight))

laptop$memory=gsub("\\D","",laptop$memory) #removing words
laptop$memory=as.numeric(laptop$memory)

laptop$memory=ifelse(laptop$memory=="11",2000,laptop$memory)
laptop$memory=ifelse(laptop$memory=="2",2000,laptop$memory)
laptop$memory=ifelse(laptop$memory=="2561",1256,laptop$memory)
laptop$memory=ifelse(laptop$memory=="1281",1128,laptop$memory)
laptop$memory=ifelse(laptop$memory=="5121",1512,laptop$memory)
laptop$memory=ifelse(laptop$memory=="10",1000,laptop$memory)
laptop$memory=ifelse(laptop$memory=="2562",2256,laptop$memory)
laptop$memory=ifelse(laptop$memory=="5122",2512,laptop$memory)
laptop$memory=ifelse(laptop$memory=="1282",2128,laptop$memory)
laptop$memory=ifelse(laptop$memory=="256256",512,laptop$memory)
laptop$memory=ifelse(laptop$memory=="256500",756,laptop$memory)
laptop$memory=ifelse(laptop$memory=="25610",1256,laptop$memory)
laptop$memory=ifelse(laptop$memory=="51210",1512,laptop$memory)
laptop$memory=ifelse(laptop$memory=="512256",768,laptop$memory)
laptop$memory=ifelse(laptop$memory=="512512",1024,laptop$memory)
laptop$memory=ifelse(laptop$memory=="641",1064,laptop$memory)
laptop$memory=ifelse(laptop$memory=="1",1000,laptop$memory)

laptop %>% 
  dplyr::select(ram,weight,memory) %>% str()
'data.frame':   1303 obs. of  3 variables:
 $ ram   : num  8 8 8 16 8 4 16 8 16 8 ...
 $ weight: num  1.37 1.34 1.86 1.83 1.37 2.1 2.04 1.34 1.3 1.6 ...
 $ memory: num  128 128 256 512 256 500 256 256 512 256 ...

6 DATA CLEANING

Missing values

checking for any missing values in the dataset

Code 
colSums(is.na.data.frame(laptop)) %>% kable() #missing values
x
company 0
type_name 0
inches 0
screen_resolution 0
cpu 0
ram 0
memory 0
gpu 0
op_sys 0
weight 0
price 0
  • no missing values

Duplicate entries

Code 
anyDuplicated.default(laptop)
[1] 0
  • no duplicated entries

7 VISUALIZATIONS & ANALYSIS

Code 
library(ggplot2)
library(plotly)
library(tvthemes)
library(extrafont)

dt=laptop %>%
  ggplot(aes(company,fill=type_name)) +
  geom_bar(position = "dodge",width = 0.5) + theme_bw()+
  theme(axis.text.x = element_text(size = 10, hjust=1,angle = 45))+
  labs(title = "Distribution of Company vs Type of laptop ",
       fill="Type of laptop",y="frequency")+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_avatar(text.font = "trebus ms") #distribution

ggplotly(dt)
AcerAppleAsusChuwiDellFujitsuGoogleHPHuaweiLenovoLGMediacomMicrosoftMSIRazerSamsungToshibaVeroXiaomi050100150
Type of laptop2 in 1 ConvertibleGamingNetbookNotebookUltrabookWorkstationDistribution of Company vs Type of laptopcompanyfrequency
  • Observation HP, Lenovo, Acer, Asus, Toshiba, Mediacom, Vero mostly produce Notebook laptops. Apple, Google, Microsoft mostly produces Ultra book laptops MSI and Razor mostly produces Gaming laptops

Skewness (Histogram/density plot)

Code 
hist1=ggplot(laptop, aes=(x=price))+
  geom_density(aes(x=price), stat = "density", fill="gold2",color="black")+
  theme_bw()+labs(title = "Distribution of Price",
                  caption = "@Data Insights 2024") #density plot

hist2=ggplot(laptop,aes(x=price))+
  geom_histogram(color="black",fill="gold2",stat = "bin")+
  theme_bw()+labs(title = "Distribution of Price",y="frequency",
                  caption = "@Data Insights 2024")#histogram plot

hist1+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_brooklyn99(text.font = "trebuchet ms")

Code 
hist2+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_brooklyn99(text.font = "trebuchet ms")

  • Observation

    This shows that majority of laptops are concentrated on the lower end meaning that there are a very few laptops with high prices and a larger number of laptops with lower prices.

Brand name impact

Code 
brand_name=as.data.frame(table(laptop$company))
colnames(brand_name)=c("Brand Name","Frequency")
brand_name %>%
arrange(desc(Frequency)) %>% kable()#brand name impact
Brand Name Frequency
Dell 297
Lenovo 297
HP 274
Asus 158
Acer 103
MSI 54
Toshiba 48
Apple 21
Samsung 9
Mediacom 7
Razer 7
Microsoft 6
Vero 4
Xiaomi 4
Chuwi 3
Fujitsu 3
Google 3
LG 3
Huawei 2

  • Observation

    Major brand names in the market are dell, hp, acer, asus and lenovo

Expensive brand name in the market

Code 
bn=ggplot(laptop, aes(x=company, y=price, fill=company))+
  geom_boxplot(stat = "boxplot",outlier.color = "blue")+
  theme(axis.text.x = element_text(size = 10, hjust=1,angle = 45))+
  stat_summary(fun.y = median, geom = "point", shape=20, size=3, color="red")+
  labs(y="average price",x="brand name",caption = "@Data Insights 2024")+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_avatar(text.font = "trebuchet ms",
                                   text.size = 5)


ggplotly(bn)
AcerAppleAsusChuwiDellFujitsuGoogleHPHuaweiLenovoLGMediacomMicrosoftMSIRazerSamsungToshibaVeroXiaomi0e+001e+052e+053e+05
companyAcerAppleAsusChuwiDellFujitsuGoogleHPHuaweiLenovoLGMediacomMicrosoftMSIRazerSamsungToshibaVeroXiaomibrand nameaverage price

  • Observation

    Razor is the most expensive as it has the highest average price

Most expensive type of laptops

Code 
me=ggplot(laptop, aes(x=type_name, y=price, fill=type_name))+
  geom_boxplot(stat = "boxplot")+
  theme(legend.position ="right")+ theme_bw()+
  theme(axis.text.x = element_text(size = 10, hjust=1,angle = 45))+
  labs(fill="Laptop Type",y="average price",x="Laptop type",
       caption = "@Data Insights 2024")

ggplotly(me)
2 in 1 ConvertibleGamingNetbookNotebookUltrabookWorkstation0e+001e+052e+053e+05
Laptop Type2 in 1 ConvertibleGamingNetbookNotebookUltrabookWorkstationLaptop typeaverage price

  • Observation

    Workstations are more expensive.

Relationships (Scatter plots)

Relationship between the inches,memory,weight, ram and prices of laptops

Code 
sp1=ggplot(laptop, aes(x=inches, y=price))+
  geom_point(stat="identity",colour="orange",shape="circle")+
  geom_smooth(method = "loess")+theme_linedraw()+
  labs(caption = "@Data Insights 2024")+
  tvthemes::theme_spongeBob(text.font = "trebuchet ms")

sp2=ggplot(laptop, aes(y=price,x=ram))+
  geom_point(stat="identity",colour="red2",shape="triangle")+
  geom_smooth(method = "loess")+theme_linedraw()+
  labs(caption = "@Data Insights 2024")+
  tvthemes::theme_hildaDusk(text.font = "trebuchet ms")

sp3=ggplot(laptop, aes(y=price,x=weight))+
  geom_point(stat="identity",colour="green",shape="square")+
  geom_smooth(method = "loess")+theme_linedraw()+
  labs(caption = "@Data Insights 2024")+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_hildaNight(text.font = "trebuchet ms")

sp4=ggplot(laptop, aes(y=price,x=memory))+
  geom_point(stat="identity",colour="blue4",shape="k")+
  geom_smooth(method = "loess")+theme_linedraw()+
  labs(caption = "@Data Insights 2024")+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_avatar(text.font = "trebuchet ms")

library(gridExtra)

grid.arrange(sp1,sp2,sp3,sp4,ncol=2,nrow=2)

  • Observation

    There is a relationship between the price of a laptop and it’s ram, memory,weight, inches.

    As inches, ram, weight,memory increases, prices also increases

8 FEATURE ENGINEERING

1. SCREEN RESOLUTION

Adding New Columns (touchscreen,ips display,x & y dimensions,hd display)

Code 
fe=as.data.frame(table(laptop$screen_resolution))
fe %>% arrange(desc(Freq)) %>% head(10) %>% kable()
Var1 Freq
Full HD 1920x1080 507
1366x768 281
IPS Panel Full HD 1920x1080 230
IPS Panel Full HD / Touchscreen 1920x1080 53
Full HD / Touchscreen 1920x1080 47
1600x900 23
Touchscreen 1366x768 16
Quad HD+ / Touchscreen 3200x1800 15
IPS Panel 4K Ultra HD 3840x2160 12
IPS Panel 4K Ultra HD / Touchscreen 3840x2160 11
  • Top 10 rows of the screen resolution column
  • The column is very noisy
Code 
library(stringr)

result=as.data.frame(str_match(laptop$screen_resolution,"(\\d+)x(\\d+)"))

laptop=laptop %>%
  mutate(x_dim=as.numeric(result$V2),
         y_dim=as.numeric(result$V3))

laptop=laptop %>%
  mutate(touchscreen=ifelse(grepl("Touchscreen",laptop$screen_resolution),1,0),
         ips_display=ifelse(grepl("IPS Panel",screen_resolution),1,0),
         hd_display=ifelse(grepl("Full HD",screen_resolution),1,0))

laptop %>%
  dplyr::select(x_dim,y_dim,touchscreen,ips_display,hd_display) %>%
  str()
'data.frame':   1303 obs. of  5 variables:
 $ x_dim      : num  2560 1440 1920 2880 2560 ...
 $ y_dim      : num  1600 900 1080 1800 1600 768 1800 900 1080 1080 ...
 $ touchscreen: num  0 0 0 0 0 0 0 0 0 0 ...
 $ ips_display: num  1 0 0 1 1 0 1 0 0 1 ...
 $ hd_display : num  0 0 1 0 0 0 0 0 1 1 ...
  • New dummy variables created

Touchscreen Feature

Code 
tsf=as.data.frame(table(laptop$touchscreen))
colnames(tsf)=c("touchscreen feature","freq")
tsf %>% kable()
touchscreen feature freq
0 1111
1 192
Code 
ggplot(laptop,aes(x=touchscreen,y=price,fill=factor(touchscreen)))+
  geom_boxplot(stat = "boxplot")+theme_bw()+
  labs(fill="Touchscreen Feature",
       caption = "@Data Insights 2024")

  • Observation touchscreen=1 , non touchscreen=0 A few laptops have the touch screen feature Laptops with touchscreen features are more expensive

Ips Display Feature

Code 
ips=as.data.frame(table(laptop$ips_display))
colnames(ips)=c("ips display feature","freq")
ips %>% kable()
ips display feature freq
0 938
1 365
Code 
ggplot(laptop,aes(x=ips_display,y=price,fill=factor(ips_display)))+
  geom_boxplot(stat = "boxplot")+theme_dark()+theme_bw()+
  labs(fill="Ips Display Feature",
       caption = "@Data Insights 2024")

  • Observation laptops with ips display=1 , laptops with non ips displays=0 365 laptops have ips display and 938 do not have ips display Laptops with ips display are more costly

HD Display Feature

Code 
hd=as.data.frame(table(laptop$hd_display))
colnames(hd)=c("hd display feature","freq")
hd %>% kable()
hd display feature freq
0 460
1 843
Code 
ggplot(laptop,aes(x=hd_display,y=price,fill=factor(hd_display)))+
  geom_boxplot(stat = "boxplot")+theme_dark()+theme_bw()+
  labs(fill="HD Display Feature",
       caption = "@Data Insights 2024")

  • Observation

    Hd display=1, non hd display=0

    A lot of laptops have hd display

    Laptops with hd display are more costly

Correlation

The new variables have to be in numeric format for correlation analysis

Code 
co_lp= laptop %>%
  dplyr::select(price,ips_display,hd_display,x_dim,y_dim,touchscreen,inches,
                weight,ram,memory)
co_lp=cor(co_lp)
co_lp %>% kable()
price ips_display hd_display x_dim y_dim touchscreen inches weight ram memory
price 1.0000000 0.2522076 0.1986116 0.5565293 0.5528092 0.1912265 0.0681967 0.2103698 0.7430071 0.1608189
ips_display 0.2522076 1.0000000 0.1854415 0.2814567 0.2890295 0.1505123 -0.1148042 0.0169671 0.2066225 -0.0146866
hd_display 0.1986116 0.1854415 1.0000000 0.0708752 0.0486595 -0.1051885 0.1635506 0.1480029 0.2103593 0.0903041
x_dim 0.5565293 0.2814567 0.0708752 1.0000000 0.9942190 0.3510657 -0.0712453 -0.0328798 0.4331205 0.0715309
y_dim 0.5528092 0.2890295 0.0486595 0.9942190 1.0000000 0.3579300 -0.0954039 -0.0538457 0.4244366 0.0569593
touchscreen 0.1912265 0.1505123 -0.1051885 0.3510657 0.3579300 1.0000000 -0.3617345 -0.2946198 0.1169841 -0.1384806
inches 0.0681967 -0.1148042 0.1635506 -0.0712453 -0.0954039 -0.3617345 1.0000000 0.8276311 0.2379928 0.5383581
weight 0.2103698 0.0169671 0.1480029 -0.0328798 -0.0538457 -0.2946198 0.8276311 1.0000000 0.3838741 0.5497539
ram 0.7430071 0.2066225 0.2103593 0.4331205 0.4244366 0.1169841 0.2379928 0.3838741 1.0000000 0.3513626
memory 0.1608189 -0.0146866 0.0903041 0.0715309 0.0569593 -0.1384806 0.5383581 0.5497539 0.3513626 1.0000000

  • Observation

    All the new variables created have a positive relationship with price x and y dimension have a strong positive relationship

Creating new variable called Pixel Per Inches (PPI) (getting rid of variables with low correlation)

Code 
laptop$ppi= (((laptop$y_dim**2)+(laptop$x_dim**2))**0.5/laptop$inches)

cor(laptop$ppi,laptop$price)
[1] 0.4734873
  • Improved correlation

2. CPU

Code 
cp=laptop$cpu %>%
  table() %>% as.data.frame %>%
  arrange(desc(Freq)) 
cp %>% head(10) %>% kable()
. Freq
Intel Core i5 7200U 2.5GHz 190
Intel Core i7 7700HQ 2.8GHz 146
Intel Core i7 7500U 2.7GHz 134
Intel Core i7 8550U 1.8GHz 73
Intel Core i5 8250U 1.6GHz 72
Intel Core i5 6200U 2.3GHz 68
Intel Core i3 6006U 2GHz 64
Intel Core i7 6500U 2.5GHz 49
Intel Core i7 6700HQ 2.6GHz 43
Intel Core i3 7100U 2.4GHz 37
  • Top 10 rows of the CPU column
  • The column is noisy
Code 
laptop=laptop %>%
  mutate(intel_core_i3=ifelse(grepl("Intel Core i3",cpu),1,0),
         intel_core_i5=ifelse(grepl("Intel Core i5",cpu),1,0),
         intel_core_i7=ifelse(grepl("Intel Core i7",cpu),1,0),
         dual_core=ifelse(grepl("Dual Core",cpu),1,0),
         amd_processor=ifelse(grepl("AMD ",cpu),1,0),
         other_processor=ifelse(grepl("Intel Xeon",cpu),1,0))

laptop %>% dplyr::select(intel_core_i3,intel_core_i5,intel_core_i7,dual_core,
                  amd_processor,other_processor) %>% str()
'data.frame':   1303 obs. of  6 variables:
 $ intel_core_i3  : num  0 0 0 0 0 0 0 0 0 0 ...
 $ intel_core_i5  : num  1 1 1 0 1 0 0 1 0 1 ...
 $ intel_core_i7  : num  0 0 0 1 0 0 1 0 1 0 ...
 $ dual_core      : num  0 0 0 0 0 0 0 0 0 0 ...
 $ amd_processor  : num  0 0 0 0 0 1 0 0 0 0 ...
 $ other_processor: num  0 0 0 0 0 0 0 0 0 0 ...
  • New dummy variables created

3. GPU

Code 
gp=laptop$gpu %>%
  table() %>% as.data.frame %>%arrange(desc(Freq))
gp %>% head(10) %>% kable()
. Freq
Intel HD Graphics 620 281
Intel HD Graphics 520 185
Intel UHD Graphics 620 68
Nvidia GeForce GTX 1050 66
Nvidia GeForce GTX 1060 48
Nvidia GeForce 940MX 43
AMD Radeon 530 41
Intel HD Graphics 500 39
Intel HD Graphics 400 37
Nvidia GeForce GTX 1070 30
  • Top 10 rows of the Gpu column
  • The column is very noisy
Code 
laptop=laptop %>%
  mutate(nvidia_graphics=ifelse(grepl("Nvidia",gpu),1,0),
         amd_graphics=ifelse(grepl("AMD",gpu),1,0),
         intel_graphics=ifelse(grepl("Intel",gpu),1,0))

laptop %>% 
  dplyr::select(nvidia_graphics,amd_graphics,intel_graphics) %>%
  str()
'data.frame':   1303 obs. of  3 variables:
 $ nvidia_graphics: num  0 0 0 0 0 0 0 0 1 0 ...
 $ amd_graphics   : num  0 0 0 1 0 1 0 0 0 0 ...
 $ intel_graphics : num  1 1 1 0 1 0 1 1 0 1 ...
  • New dummy variables created

4. OP_SYS

Code 
op=laptop$op_sys %>%
  table() %>% as.data.frame %>%
  arrange(desc(Freq))
op %>% head(10) %>% kable()
. Freq
Windows 10 1072
No OS 66
Linux 62
Windows 7 45
Chrome OS 27
macOS 13
Mac OS X 8
Windows 10 S 8
Android 2
  • Top 10 rows of the Op_sys column
  • The column is very noisy
Code 
laptop=laptop %>%
  mutate(windows_10=ifelse(grepl("Windows 10",op_sys),1,0),
         no_operating_system=ifelse(grepl("No OS",op_sys),1,0),
         linux=ifelse(grepl("Linux",op_sys),1,0),
         windows_7=ifelse(grepl("Windows 7",op_sys),1,0),
         chrome_os=ifelse(grepl("Chrome OS ",op_sys),1,0),
         mac_os=ifelse(grepl("macOS",op_sys),1,0),
         mac_os_x=ifelse(grepl("Mac OS X",op_sys),1,0),
         windows_10_s=ifelse(grepl("Windows 10 S",op_sys),1,0),
         android=ifelse(grepl("Android",op_sys),1,0))

laptop %>% 
  dplyr::select(windows_10,no_operating_system,linux,windows_7,
                  chrome_os,mac_os,mac_os_x,windows_10,android)%>%str()
'data.frame':   1303 obs. of  8 variables:
 $ windows_10         : num  0 0 0 0 0 1 0 0 1 1 ...
 $ no_operating_system: num  0 0 1 0 0 0 0 0 0 0 ...
 $ linux              : num  0 0 0 0 0 0 0 0 0 0 ...
 $ windows_7          : num  0 0 0 0 0 0 0 0 0 0 ...
 $ chrome_os          : num  0 0 0 0 0 0 0 0 0 0 ...
 $ mac_os             : num  1 1 0 1 1 0 0 1 0 0 ...
 $ mac_os_x           : num  0 0 0 0 0 0 1 0 0 0 ...
 $ android            : num  0 0 0 0 0 0 0 0 0 0 ...
  • New dummy variables created

9 MODEL BUILDING (PREDICTING LAPTOP PRICE)

Multiple Linear Regression (Backward Approach)

Code 
laptop_subset=laptop %>%
  dplyr::select(6:7,10:11,14:35)

set.seed(1)

sample=sample(c(TRUE,FALSE),nrow(laptop_subset),replace=TRUE,prob = c(0.7,0.3))

train=laptop_subset[sample,]

test=laptop_subset[!sample,]

library(MASS)

full_model= lm(price ~ .,data = train)#full model including all the variables

output=capture.output(backward_regression<- 
                        stepAIC(full_model,direction="backward",
                                              scope=list(lower= ~1),
                                            data=train)) #keeping significant variables

summary(backward_regression)

Call:
lm(formula = price ~ ram + memory + weight + hd_display + ppi + 
    intel_core_i3 + intel_core_i5 + intel_core_i7 + amd_processor + 
    other_processor + amd_graphics + no_operating_system + linux + 
    windows_7 + mac_os, data = train)

Residuals:
   Min     1Q Median     3Q    Max 
-66745 -10935  -1733   8259 135827 

Coefficients:
                      Estimate Std. Error t value Pr(>|t|)    
(Intercept)         -27256.725   4312.931  -6.320 4.13e-10 ***
ram                   3630.263    180.835  20.075  < 2e-16 ***
memory                  -5.714      1.723  -3.315 0.000953 ***
weight                6531.871   1385.708   4.714 2.82e-06 ***
hd_display           -2834.155   1523.850  -1.860 0.063233 .  
ppi                    211.398     19.225  10.996  < 2e-16 ***
intel_core_i3         7918.765   2824.250   2.804 0.005159 ** 
intel_core_i5        18141.763   2407.517   7.535 1.19e-13 ***
intel_core_i7        29296.716   2697.793  10.860  < 2e-16 ***
amd_processor        11233.685   4243.068   2.648 0.008251 ** 
other_processor      97142.568  14179.695   6.851 1.37e-11 ***
amd_graphics        -11594.426   2336.324  -4.963 8.32e-07 ***
no_operating_system -13535.440   3352.309  -4.038 5.86e-05 ***
linux                -9475.005   3244.385  -2.920 0.003583 ** 
windows_7            27642.342   3370.634   8.201 8.25e-16 ***
mac_os               10963.073   7525.384   1.457 0.145519    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 19370 on 894 degrees of freedom
Multiple R-squared:  0.7348,    Adjusted R-squared:  0.7304 
F-statistic: 165.2 on 15 and 894 DF,  p-value: < 2.2e-16

Adjusted R Squared = 0.7304 means 73 percent of variance in the dependent variable (price) is explained by the independent variables hence it is a better fit of the model to the data.

P Value of 2.2e-16 <0.05 means that the model is statistical significant in predicting the price of a laptop.

Assumptions of Multiple Linear Regression

1. Linearity of the relationship

Code 
plot(backward_regression$fitted.values,backward_regression$residuals,
     xlab = "Fitted Values",ylab = "Residuals")

  • There is no pattern hence assumption not violated

2. Independence of errors

Code 
library(car)
durbinWatsonTest(backward_regression)
 lag Autocorrelation D-W Statistic p-value
   1       0.0472467       1.90477    0.14
 Alternative hypothesis: rho != 0
  • D-W statistic close to 2 indicates no auto correlation (1.90477 is approximately 2) hence assumption not violated

3. Homoscedacity (Constant Variance of Residuals)

Code 
plot(backward_regression, which = 3)

  • No cone shaped pattern hence assumption not violated

4. Normality of Residuals

Code 
qqnorm(backward_regression$residuals)

  • No deviations from normality hence assumption not violated

5. Multicollinearity

Code 
vif(backward_regression)
                ram              memory              weight          hd_display 
           1.981195            1.548162            1.986819            1.285926 
                ppi       intel_core_i3       intel_core_i5       intel_core_i7 
           1.580429            1.808834            3.079975            4.251826 
      amd_processor     other_processor        amd_graphics no_operating_system 
           2.009287            1.069451            1.621121            1.035651 
              linux           windows_7              mac_os 
           1.098448            1.047004            1.048466
  • All the Variance Inflation Factors are less than 10 hence assumption not violated.

The Price of the laptop can be predicted using the final regression model.

Initial Model:

price ~ ram + memory + weight + touchscreen + ips_display + hd_display + ppi + intel_core_i3 + intel_core_i5 + intel_core_i7 + dual_core + amd_processor + other_processor + nvidia_graphics + amd_graphics + intel_graphics + windows_10 + no_operating_system + linux + windows_7 + chrome_os + mac_os + mac_os_x + windows_10_s + android

Final Model:

price ~ ram + memory + weight + hd_display + ppi + intel_core_i3 + intel_core_i5 + intel_core_i7 + amd_processor + other_processor + amd_graphics + no_operating_system + linux + windows_7 + mac_os

Regression model

Price= -27256.725416 + 3630.263099 (ram ) + memory (-5.713723) + weight (6531.871213) + hd_display ( -2834.155057) + ppi ( 211.397877) + intel_core_i3 ( 7918.764798) + intel_core_i5 (18141.763446) + intel_core_i7 (29296.716320) + amd_processor (11233.684745) + other_processor (97142.568150) + amd_graphics (-11594.426323) + no_operating_system ( -13535.440259) + linux (-9475.004938) + windows_7 ( 27642.341792) + mac_os (10963.073052)

10 LAPTOP PRICE DETECTION APPLICATION

The application was build using using Shiny package. Here is the link to the application: https://pythias.shinyapps.io/LPDA/

11 CODE APPENDIX

Code 
knitr::opts_chunk$set(echo = T, message=F, warning = F)
laptop=read.csv(file.choose()) #reading dataset

library(janitor)

laptop=clean_names(laptop[2:12]) #Cleaning & keeping important variables
library(knitr)
head(laptop,5)%>% kable()  #first 5 rows
colnames(laptop) %>% kable() #column names
str(laptop) %>% kable()#dataset classes
#variable conversion
library(dplyr)

laptop$ram=as.numeric(sub("GB","",laptop$ram))
laptop$weight=as.numeric(sub("kg","",laptop$weight))

laptop$memory=gsub("\\D","",laptop$memory) #removing words
laptop$memory=as.numeric(laptop$memory)

laptop$memory=ifelse(laptop$memory=="11",2000,laptop$memory)
laptop$memory=ifelse(laptop$memory=="2",2000,laptop$memory)
laptop$memory=ifelse(laptop$memory=="2561",1256,laptop$memory)
laptop$memory=ifelse(laptop$memory=="1281",1128,laptop$memory)
laptop$memory=ifelse(laptop$memory=="5121",1512,laptop$memory)
laptop$memory=ifelse(laptop$memory=="10",1000,laptop$memory)
laptop$memory=ifelse(laptop$memory=="2562",2256,laptop$memory)
laptop$memory=ifelse(laptop$memory=="5122",2512,laptop$memory)
laptop$memory=ifelse(laptop$memory=="1282",2128,laptop$memory)
laptop$memory=ifelse(laptop$memory=="256256",512,laptop$memory)
laptop$memory=ifelse(laptop$memory=="256500",756,laptop$memory)
laptop$memory=ifelse(laptop$memory=="25610",1256,laptop$memory)
laptop$memory=ifelse(laptop$memory=="51210",1512,laptop$memory)
laptop$memory=ifelse(laptop$memory=="512256",768,laptop$memory)
laptop$memory=ifelse(laptop$memory=="512512",1024,laptop$memory)
laptop$memory=ifelse(laptop$memory=="641",1064,laptop$memory)
laptop$memory=ifelse(laptop$memory=="1",1000,laptop$memory)

laptop %>% 
  dplyr::select(ram,weight,memory) %>% str()
colSums(is.na.data.frame(laptop)) %>% kable() #missing values

anyDuplicated.default(laptop)
library(ggplot2)
library(plotly)
library(tvthemes)
library(extrafont)

dt=laptop %>%
  ggplot(aes(company,fill=type_name)) +
  geom_bar(position = "dodge",width = 0.5) + theme_bw()+
  theme(axis.text.x = element_text(size = 10, hjust=1,angle = 45))+
  labs(title = "Distribution of Company vs Type of laptop ",
       fill="Type of laptop",y="frequency")+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_avatar(text.font = "trebus ms") #distribution

ggplotly(dt)
hist1=ggplot(laptop, aes=(x=price))+
  geom_density(aes(x=price), stat = "density", fill="gold2",color="black")+
  theme_bw()+labs(title = "Distribution of Price",
                  caption = "@Data Insights 2024") #density plot

hist2=ggplot(laptop,aes(x=price))+
  geom_histogram(color="black",fill="gold2",stat = "bin")+
  theme_bw()+labs(title = "Distribution of Price",y="frequency",
                  caption = "@Data Insights 2024")#histogram plot

hist1+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_brooklyn99(text.font = "trebuchet ms")

hist2+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_brooklyn99(text.font = "trebuchet ms")

brand_name=as.data.frame(table(laptop$company))
colnames(brand_name)=c("Brand Name","Frequency")
brand_name %>%
arrange(desc(Frequency)) %>% kable()#brand name impact


bn=ggplot(laptop, aes(x=company, y=price, fill=company))+
  geom_boxplot(stat = "boxplot",outlier.color = "blue")+
  theme(axis.text.x = element_text(size = 10, hjust=1,angle = 45))+
  stat_summary(fun.y = median, geom = "point", shape=20, size=3, color="red")+
  labs(y="average price",x="brand name",caption = "@Data Insights 2024")+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_avatar(text.font = "trebuchet ms",
                                   text.size = 5)


ggplotly(bn)

mb=as.data.frame(table(laptop$type_name))
colnames(mb)=c("type of laptop","frequency")
mb %>% arrange(desc(frequency)) %>% kable()

me=ggplot(laptop, aes(x=type_name, y=price, fill=type_name))+
  geom_boxplot(stat = "boxplot")+
  theme(legend.position ="right")+ theme_bw()+
  theme(axis.text.x = element_text(size = 10, hjust=1,angle = 45))+
  labs(fill="Laptop Type",y="average price",x="Laptop type",
       caption = "@Data Insights 2024")

ggplotly(me)


sp1=ggplot(laptop, aes(x=inches, y=price))+
  geom_point(stat="identity",colour="orange",shape="circle")+
  geom_smooth(method = "loess")+theme_linedraw()+
  labs(caption = "@Data Insights 2024")+
  tvthemes::theme_spongeBob(text.font = "trebuchet ms")

sp2=ggplot(laptop, aes(y=price,x=ram))+
  geom_point(stat="identity",colour="red2",shape="triangle")+
  geom_smooth(method = "loess")+theme_linedraw()+
  labs(caption = "@Data Insights 2024")+
  tvthemes::theme_hildaDusk(text.font = "trebuchet ms")

sp3=ggplot(laptop, aes(y=price,x=weight))+
  geom_point(stat="identity",colour="green",shape="square")+
  geom_smooth(method = "loess")+theme_linedraw()+
  labs(caption = "@Data Insights 2024")+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_hildaNight(text.font = "trebuchet ms")

sp4=ggplot(laptop, aes(y=price,x=memory))+
  geom_point(stat="identity",colour="blue4",shape="k")+
  geom_smooth(method = "loess")+theme_linedraw()+
  labs(caption = "@Data Insights 2024")+
  ggthemes::scale_fill_tableau()+
  tvthemes::theme_avatar(text.font = "trebuchet ms")

library(gridExtra)

grid.arrange(sp1,sp2,sp3,sp4,ncol=2,nrow=2)

fe=as.data.frame(table(laptop$screen_resolution))
fe %>% arrange(desc(Freq)) %>% head(10) %>% kable()

library(stringr)

result=as.data.frame(str_match(laptop$screen_resolution,"(\\d+)x(\\d+)"))

laptop=laptop %>%
  mutate(x_dim=as.numeric(result$V2),
         y_dim=as.numeric(result$V3))

laptop=laptop %>%
  mutate(touchscreen=ifelse(grepl("Touchscreen",laptop$screen_resolution),1,0),
         ips_display=ifelse(grepl("IPS Panel",screen_resolution),1,0),
         hd_display=ifelse(grepl("Full HD",screen_resolution),1,0))

laptop %>%
  dplyr::select(x_dim,y_dim,touchscreen,ips_display,hd_display) %>%
  str()

tsf=as.data.frame(table(laptop$touchscreen))
colnames(tsf)=c("touchscreen feature","freq")
tsf %>% kable()

ggplot(laptop,aes(x=touchscreen,y=price,fill=factor(touchscreen)))+
  geom_boxplot(stat = "boxplot")+theme_bw()+
  labs(fill="Touchscreen Feature",
       caption = "@Data Insights 2024")

ips=as.data.frame(table(laptop$ips_display))
colnames(ips)=c("ips display feature","freq")
ips %>% kable()


ggplot(laptop,aes(x=ips_display,y=price,fill=factor(ips_display)))+
  geom_boxplot(stat = "boxplot")+theme_dark()+theme_bw()+
  labs(fill="Ips Display Feature",
       caption = "@Data Insights 2024")

hd=as.data.frame(table(laptop$hd_display))
colnames(hd)=c("hd display feature","freq")
hd %>% kable()

ggplot(laptop,aes(x=hd_display,y=price,fill=factor(hd_display)))+
  geom_boxplot(stat = "boxplot")+theme_dark()+theme_bw()+
  labs(fill="HD Display Feature",
       caption = "@Data Insights 2024")

co_lp= laptop %>%
  dplyr::select(price,ips_display,hd_display,x_dim,y_dim,touchscreen,inches,
                weight,ram,memory)
co_lp=cor(co_lp)
co_lp %>% kable()

laptop$ppi= (((laptop$y_dim**2)+(laptop$x_dim**2))**0.5/laptop$inches)

cor(laptop$ppi,laptop$price)


cp=laptop$cpu %>%
  table() %>% as.data.frame %>%
  arrange(desc(Freq)) 
cp %>% head(10) %>% kable()
laptop=laptop %>%
  mutate(intel_core_i3=ifelse(grepl("Intel Core i3",cpu),1,0),
         intel_core_i5=ifelse(grepl("Intel Core i5",cpu),1,0),
         intel_core_i7=ifelse(grepl("Intel Core i7",cpu),1,0),
         dual_core=ifelse(grepl("Dual Core",cpu),1,0),
         amd_processor=ifelse(grepl("AMD ",cpu),1,0),
         other_processor=ifelse(grepl("Intel Xeon",cpu),1,0))

laptop %>% dplyr::select(intel_core_i3,intel_core_i5,intel_core_i7,dual_core,
                  amd_processor,other_processor) %>% str()

gp=laptop$gpu %>%
  table() %>% as.data.frame %>%arrange(desc(Freq))
gp %>% head(10) %>% kable()

laptop=laptop %>%
  mutate(nvidia_graphics=ifelse(grepl("Nvidia",gpu),1,0),
         amd_graphics=ifelse(grepl("AMD",gpu),1,0),
         intel_graphics=ifelse(grepl("Intel",gpu),1,0))

laptop %>% 
  dplyr::select(nvidia_graphics,amd_graphics,intel_graphics) %>%
  str()
op=laptop$op_sys %>%
  table() %>% as.data.frame %>%
  arrange(desc(Freq))
op %>% head(10) %>% kable()
laptop=laptop %>%
  mutate(windows_10=ifelse(grepl("Windows 10",op_sys),1,0),
         no_operating_system=ifelse(grepl("No OS",op_sys),1,0),
         linux=ifelse(grepl("Linux",op_sys),1,0),
         windows_7=ifelse(grepl("Windows 7",op_sys),1,0),
         chrome_os=ifelse(grepl("Chrome OS ",op_sys),1,0),
         mac_os=ifelse(grepl("macOS",op_sys),1,0),
         mac_os_x=ifelse(grepl("Mac OS X",op_sys),1,0),
         windows_10_s=ifelse(grepl("Windows 10 S",op_sys),1,0),
         android=ifelse(grepl("Android",op_sys),1,0))

laptop %>% 
  dplyr::select(windows_10,no_operating_system,linux,windows_7,
                  chrome_os,mac_os,mac_os_x,windows_10,android)%>%str()

laptop_subset=laptop %>%
  dplyr::select(6:7,10:11,14:35)

set.seed(1)

sample=sample(c(TRUE,FALSE),nrow(laptop_subset),replace=TRUE,prob = c(0.7,0.3))

train=laptop_subset[sample,]

test=laptop_subset[!sample,]

library(MASS)

full_model= lm(price ~ .,data = train)#full model including all the variables

output=capture.output(backward_regression<- 
                        stepAIC(full_model,direction="backward",
                                              scope=list(lower= ~1),
                                            data=train)) #keeping significant variables

summary(backward_regression)

plot(backward_regression$fitted.values,backward_regression$residuals,
     xlab = "Fitted Values",ylab = "Residuals")

library(car)
durbinWatsonTest(backward_regression)
plot(backward_regression, which = 3)
qqnorm(backward_regression$residuals)
vif(backward_regression)
Back to top