Introduction

About Author

I am a Mechanical Engineering graduate fascinated with Data Science. #Rstats is my favorite data analysis tool.

You can find me at: Facebook

About Dataset

This is a dataset containing of many observations of wine scraped from the WineEnthusiast. I downloaded the dataset from Kaggle

Scope of the Project

The scope of the project shall involve solely on Italian wines.

Codes i will also partly reference from:

My previous rpubs project on Australian wines with the same dataset
Ariel, Galvis Moreno, Orsini-Rosenberg, Pi, Flores, Saggese, Savic - I am imitating parts of their code for analysis, albeit with the sole purpose of analyzing Italian-made wines.

Loading the required libraries and data

Libraries

I uploaded the necessary libraries

library(tidyverse)

## -- Attaching packages ------------------- tidyverse 1.2.1 --

## v ggplot2 3.1.0     v purrr   0.2.5
## v tibble  1.4.2     v dplyr   0.7.8
## v tidyr   0.8.1     v stringr 1.3.1
## v readr   1.1.1     v forcats 0.3.0

## Warning: package 'dplyr' was built under R version 3.5.2

## -- Conflicts ---------------------- tidyverse_conflicts() --
## x dplyr::filter() masks stats::filter()
## x dplyr::lag()    masks stats::lag()

library(gridExtra)

## 
## Attaching package: 'gridExtra'

## The following object is masked from 'package:dplyr':
## 
##     combine

library(stringr)
library(lubridate)

## 
## Attaching package: 'lubridate'

## The following object is masked from 'package:base':
## 
##     date

library(effects)

## Warning: package 'effects' was built under R version 3.5.3

## Loading required package: carData

## Warning: package 'carData' was built under R version 3.5.2

## lattice theme set by effectsTheme()
## See ?effectsTheme for details.

library(httr)

## Warning: package 'httr' was built under R version 3.5.3

library(tm)

## Warning: package 'tm' was built under R version 3.5.3

## Loading required package: NLP

## Warning: package 'NLP' was built under R version 3.5.2

## 
## Attaching package: 'NLP'

## The following object is masked from 'package:httr':
## 
##     content

## The following object is masked from 'package:ggplot2':
## 
##     annotate

library(wordcloud)

## Warning: package 'wordcloud' was built under R version 3.5.3

## Loading required package: RColorBrewer

library(SnowballC)

## Warning: package 'SnowballC' was built under R version 3.5.2

library(car)

## Warning: package 'car' was built under R version 3.5.3

## 
## Attaching package: 'car'

## The following object is masked from 'package:dplyr':
## 
##     recode

## The following object is masked from 'package:purrr':
## 
##     some

library(ggthemes)

## Warning: package 'ggthemes' was built under R version 3.5.2

Data

Uploading the data

library(readxl)
wine <- read_excel("D:/Working Directory/winemag_data_first150k.xlsx")

Understanding the Data

glimpse(wine)

## Observations: 150,930
## Variables: 11
## $ X__1        <dbl> 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, ...
## $ country     <chr> "US", "Spain", "US", "US", "France", "Spain", "Spa...
## $ description <chr> "This tremendous 100% varietal wine hails from Oak...
## $ designation <chr> "Martha's Vineyard", "Carodorum SelecciÃ³n Especia...
## $ points      <dbl> 96, 96, 96, 96, 95, 95, 95, 95, 95, 95, 95, 95, 95...
## $ price       <dbl> 235, 110, 90, 65, 66, 73, 65, 110, 65, 60, 80, 48,...
## $ province    <chr> "California", "Northern Spain", "California", "Ore...
## $ region_1    <chr> "Napa Valley", "Toro", "Knights Valley", "Willamet...
## $ region_2    <chr> "Napa", NA, "Sonoma", "Willamette Valley", NA, NA,...
## $ variety     <chr> "Cabernet Sauvignon", "Tinta de Toro", "Sauvignon ...
## $ winery      <chr> "Heitz", "Bodega Carmen RodrÃguez", "Macauley", "...

Description of Variables.

X - Indicates the row number
country - Indicates the origin of the wine
Province - partuclar wine state the wine is made from
Region_1 - wine growing area
Region_2 - specific region
description - review for the particular bottle
designation - denotes which particular vineyard the grapes the wine is made from
points - wine rating from 0 to 100
variety - type of grapes the wine is from
price - the price of the wine. While the dataset lists this in USD, i will convert this into Euros. As of today (March 9, 2019), 1 USD equals 0.89 EUR

Mutating (converting the price to euros)

wine %>%
  mutate(price = price * 0.89)

Splitting only for Italian wines

Given i indicated the scope of this project to detail only wines from Italy, i will filter the wines to include only those from Italy.

wines_Italia <- wine %>%
  filter(country == "Italy") %>%
  filter(!is.na(price))

summary(wines_Italia)

##       X__1          country          description        designation       
##  Min.   :    10   Length:18784       Length:18784       Length:18784      
##  1st Qu.: 31646   Class :character   Class :character   Class :character  
##  Median : 71596   Mode  :character   Mode  :character   Mode  :character  
##  Mean   : 72285                                                           
##  3rd Qu.:108981                                                           
##  Max.   :150929                                                           
##      points           price          province           region_1        
##  Min.   : 80.00   Min.   :  5.00   Length:18784       Length:18784      
##  1st Qu.: 87.00   1st Qu.: 17.00   Class :character   Class :character  
##  Median : 88.00   Median : 25.00   Mode  :character   Mode  :character  
##  Mean   : 88.45   Mean   : 37.55                                        
##  3rd Qu.: 90.00   3rd Qu.: 48.00                                        
##  Max.   :100.00   Max.   :900.00                                        
##    region_2           variety             winery         
##  Length:18784       Length:18784       Length:18784      
##  Class :character   Class :character   Class :character  
##  Mode  :character   Mode  :character   Mode  :character  
##                                                          
##                                                          
##

Cleaning the Data

I would like to ensure that i am working with clean data with no duplicate values. Given that “Description” consists of free text in which each taster who rates and reviews the wine writes a description using his/her own language, it would be a great place to start.

wines_Italia$description %>%
  glimpse()

##  chr [1:18784] "Elegance, complexity and structure come together in this drop-dead gorgeous winethat ranks among Italy's greate"| __truncated__ ...

The next step is to ensure that we will not be deleting empty descriptions that might be mistaken for duplicates

## Empty spaces: 0

## NAs: 0

Thankfully, no empty spaces and NAs exist, hence i would like to get rid of duplicates

wines_Italia %>%
  mutate(duplicate = duplicated(description)) %>%
  filter(duplicate == FALSE) %>%
  select(-duplicate)

glimpse(wines_Italia)

## Observations: 18,784
## Variables: 11
## $ X__1        <dbl> 10, 35, 37, 38, 39, 43, 45, 46, 48, 50, 59, 79, 83...
## $ country     <chr> "Italy", "Italy", "Italy", "Italy", "Italy", "Ital...
## $ description <chr> "Elegance, complexity and structure come together ...
## $ designation <chr> "Ronco della Chiesa", "Riserva", NA, NA, "Riserva"...
## $ points      <dbl> 95, 90, 90, 90, 90, 90, 90, 90, 90, 90, 90, 91, 91...
## $ price       <dbl> 80, 135, 29, 23, 29, 39, 30, 90, 50, 100, 60, 75, ...
## $ province    <chr> "Northeastern Italy", "Tuscany", "Tuscany", "Tusca...
## $ region_1    <chr> "Collio", "Brunello di Montalcino", "Vino Nobile d...
## $ region_2    <chr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA...
## $ variety     <chr> "Friulano", "Sangiovese", "Sangiovese", "Sangioves...
## $ winery      <chr> "Borgo del Tiglio", "Carillon", "Avignonesi", "Cas...

Data Exploration and Visualization

Province

Let’s create a table and a graph to analyze the distribution of Italian wines per province. It would be wise to keep it to the top 5 provinces

wines_Italia_province <- wines_Italia %>%
  group_by(province) %>%
  summarise(total = n()) %>%
  arrange(desc(total)) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt))

print(wines_Italia_province)

## # A tibble: 10 x 4
##    province           total    pcnt accum
##    <chr>              <int>   <dbl> <dbl>
##  1 Tuscany             5961 31.7     31.7
##  2 Veneto              3103 16.5     48.3
##  3 Piedmont            2909 15.5     63.7
##  4 Sicily & Sardinia   2032 10.8     74.6
##  5 Northeastern Italy  1852  9.86    84.4
##  6 Central Italy       1257  6.69    91.1
##  7 Southern Italy      1167  6.21    97.3
##  8 Lombardy             444  2.36    99.7
##  9 Italy Other           52  0.277  100.0
## 10 Northwestern Italy     7  0.0373 100.0

wines_Italia_province %>%
  head(5) %>%
  ggplot(aes(x = reorder(province, -pcnt), y = total, FILL = province)) + 
  geom_bar(aes(fill = province), stat = "identity") + 
  labs(x = "\n Vini Italiani (per provinzia), Top 5 \n") + labs(y = "\n Total \n") + 
  labs(title = "\n Rassegna di Vini Italiani") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized()

Tuscany has around most of the wines on the list with 31%. They have always been a historically wine-producing region. Piedmont second, albeit far behind from Tuscany and so on.

Rating

Wine reviews are generally about quality over quantity. Hence it’s a test to see which Italian province has the best-rated wines.

Rating <- wines_Italia %>%
  group_by(province) %>%
  summarise_at(vars(points), funs(points = mean(., na.rm = T))) %>%
  arrange(desc(points)) %>%
  head(5)

Rating %>%
  head(5) %>%
  ggplot(aes(x = reorder(province, - points), y = points)) +
  geom_bar(fill = "yellowgreen", stat = "identity") + 
  coord_cartesian(ylim = c(50, 99)) + 
  labs(x = "Provinzia Italiani") + labs(y = "Valutazione") +
  geom_text(aes(label = sprintf("%.1f %%", points))) +
  theme_solarized()

It would, however, be an unfair comparison to compare Tuscan with Lombardy wines given the huge disparity of wines reviewed (7281 vs 580)

Filtering wines via their rating category

I have previously made an rpubs project albeit on Australian wines. On that project i filtered wines in three categories:

Excellent (90 and above)
Okay (Less than 89)

For Excellent Wines (rated 90 and above)

Excellent <- wines_Italia %>%
  filter(points >= 90) %>%
  group_by(province) %>%
  summarise(total = n()) %>%
  arrange(desc(total)) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 3))

Excellent

Excellent %>%
  head(5) %>%
  ggplot(aes(x = reorder(province, - total), y = total)) +
  geom_bar(fill = "yellowgreen", stat = "identity") + 
  coord_cartesian(ylim = c(400, 4000)) + 
  labs(x = "Provinzia Italiani") + labs(y = "Total") +
  labs(title = "\n Il Vino Italiano classificati da 90 in superiore (top 5)") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) +
  theme_solarized()

So, almost half of the wines rated excellent come from Tuscany. Piedmont comes at 22.5%.

For wines rated okay

Okay <- wines_Italia %>%
  filter(points <= 89) %>%
  group_by(province) %>%
  summarise(total = n()) %>%
  arrange(desc(total)) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 3))

Okay

Okay %>%
  head(5) %>%
  ggplot(aes(x = reorder(province, - total), y = total)) +
  geom_bar(fill = "yellowgreen", stat = "identity") + 
  coord_cartesian(ylim = c(400, 4000)) + 
  labs(x = "Provinzia Italiani") + labs(y = "Total") +
  labs(title = "\n Il Vino Italiano classificati da 89 e sotto (top 5)") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) +
  theme_solarized()

Conducting a t.test on the provinces with the highest ratings

Ariel, Moreno, Orsini-Rosberg, Saggese et al on a similar project whose code on this same dataset i have been inspired to replicate to some extent, employed a Welch Two Sample t-test on the dataset to see if even though a max price of wine on a certain country is higher than that of another, the average price would still be the same.

However, i will try to determine the max rating of a province’s wine and see if the average rating of the two highest-rated wines are the same.

max_rating <- wines_Italia %>%
  select(province, points) %>%
  group_by(province) %>%
  summarise(maxpoints = max(points, na.rm = TRUE)) %>%
  arrange(desc(maxpoints)) %>%
  head(5)

max_rating

t.test(wines_Italia$points[wines_Italia$province == "Tuscany"], wines_Italia$points[wines_Italia$province == "Piedmont"], paired = FALSE)

## 
##  Welch Two Sample t-test
## 
## data:  wines_Italia$points[wines_Italia$province == "Tuscany"] and wines_Italia$points[wines_Italia$province == "Piedmont"]
## t = 8.5037, df = 6186.7, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  0.4215666 0.6741644
## sample estimates:
## mean of x mean of y 
##  89.39524  88.84737

Given the p-value is less than 0.05, then we can say (and we proved earlier) that the average rating difference is significant.

Price

Most expensive wine?.

So how much are the most expensive wines per province?.

wines_Italia %>%
  select(province, price) %>%
  group_by(province) %>%
  summarise(maxprice = max(price, na.rm = TRUE)) %>%
  arrange(desc(maxprice)) %>%
  head(10)

T.test on the wine prices.

Given Tuscany and Veneto are 1st and second amongst the priciest wine list, we want to know if the disparity has an effect on the average.

t.test(wines_Italia$price[wines_Italia$province == "Tuscany"], wines_Italia$price[wines_Italia$province == "Veneto"], paired = FALSE)

## 
##  Welch Two Sample t-test
## 
## data:  wines_Italia$price[wines_Italia$province == "Tuscany"] and wines_Italia$price[wines_Italia$province == "Veneto"]
## t = 18.52, df = 8388.6, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  13.86257 17.14454
## sample estimates:
## mean of x mean of y 
##  48.12263  32.61908

Given p-value is less than 0.05, then we can conclude with 95% confidence that there is a significant difference between prices of wines at Tuscany and at Veneto.

Variety.

Let’s create a graph detailing the distribution of varieties in this dataset. For legibility sake, we will restrict ourselves to the top 5.

wine_varieties <- wines_Italia %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  arrange(desc(total)) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt))

print(wine_varieties)

## # A tibble: 175 x 4
##    variety                       total  pcnt accum
##    <chr>                         <int> <dbl> <dbl>
##  1 Red Blend                      3143 16.7   16.7
##  2 Sangiovese                     2258 12.0   28.8
##  3 Nebbiolo                       1479  7.87  36.6
##  4 Corvina, Rondinella, Molinara  1292  6.88  43.5
##  5 Sangiovese Grosso               937  4.99  48.5
##  6 White Blend                     797  4.24  52.7
##  7 Barbera                         706  3.76  56.5
##  8 Pinot Grigio                    682  3.63  60.1
##  9 Prosecco                        515  2.74  62.9
## 10 Glera                           479  2.55  65.4
## # ... with 165 more rows

wine_varieties %>%
  head(5) %>%
  ggplot(aes(x = reorder(variety, -pcnt), y = total)) + 
  geom_bar(fill = "orchid", stat = "identity") + 
  labs(x = "\n Varieta di Vino, Top 5 \n") + labs(y = "\n Total \n") + 
  labs(title = "\n Rassegna di Vini Italiani") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

The red blend is the most prominent wine in the dataset whereas the Sangiovese variety (wine common in most of Central Italy frorm the Emilia-Romagna region down to Lazio, Campania and Sicily that tastes oaky with a strawberry fruitiness and some mild spiciness) comes second.

Rating of variety

We want to know the average rating per wine variety present. As always, we limit ourselves to the top 5 for legibility sake. Like before i will filter based on the category. We will use the same category

General Rating

We want to know, however, the average rating per variety.

Rating2 <- wines_Italia %>%
  group_by(variety) %>%
  summarise_at(vars(points), funs(points = mean(., na.rm = T))) %>%
  arrange(desc(points)) %>%
  head(10)

Rating2 %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -points), y = points)) + 
  geom_bar(fill = "orchid", stat = "identity") + 
  coord_cartesian(ylim = c(50, 99)) +
  labs(x = "Varieta di Vino (top 10)") + labs(y = "Per cento") +
  geom_text(aes(label = sprintf("%.1f %%", points))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

For Excellent Wines (rated 90 and above)

As before we will sort out wines rated 90 and higher and everything else.

Excellent_variety <- wines_Italia %>%
  filter(points >= 90) %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  arrange(desc(total)) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 3))

Excellent_variety

Excellent_variety %>%
  head(5) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "orchid", stat = "identity") + 
  coord_cartesian(ylim = c(400, 2000)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Italiano classificati da 90 in superiore (top 10)") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

So, it’s kind of apparent that the Red Blend has more wines rated 90 and above in spite of not being in the top 10 in terms of general rating.

For wines rated okay (89 and below)

Okay_variety <- wines_Italia %>%
  filter(points <= 89) %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  arrange(desc(total)) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 3))

Okay_variety

Okay_variety %>%
  head(5) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "orchid", stat = "identity") + 
  coord_cartesian(ylim = c(400, 2500)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Italiano classificati da 89 e sotto (top 5)") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.5, hjust=1))

The Red Blend also dominates here, having 13.6% of all wines reviewed here.

Wine varieties per province

There are 10 wine-producing provinces indicated in this dataset. It is my interest to determine which wine varieties are widely produced per province and that i believe it is manageable.

Tuscany

Filtering the wine varieties for the Tuscany province.

wines_Toscania <- wines_Italia %>%
  filter(province == "Tuscany") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_Toscania

wines_Toscania %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(50, 2700)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Toscania") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

Through our exploratory data analysis, the Sangiovese is the wine most prominently produced here, followed by the Red Blend and the Sangiovese Grosso.

Piedmont

Filtering the wine varieties for the Piedmont province

wines_Piemonte <- wines_Italia %>%
  filter(province == "Piedmont") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_Piemonte

wines_Piemonte %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(50, 2500)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Piemonte") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

The Nebbiolo variety of wine is most prominently produced here, followed by the Barbera and Dolcetto.

Veneto

Filtering the wine varieties for the Veneto province

wines_Venetto <- wines_Italia %>%
  filter(province == "Veneto") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_Venetto

wines_Venetto %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(50, 2500)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Venetto") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

The Corvina (a wine known for being light in variety), Rondinella (wine from blackish berries that appears in wines blended with Corvina) and Molinara (variety that is commonly found in the corvina and Rondinella that gives those wines acidity) varieties are the most prominent here.

Sicily and Sardinia

wines_SS <- wines_Italia %>%
  filter(province == "Sicily & Sardinia") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_SS

wines_SS %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(50, 500)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Sicilia e Sardinia") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

The Red Blend is prominent here, followed by the Nero d’Aviola (a wine native to the Avola in Southern Sicily known to have sweet tannins and plum or peppery flavors) and the White Blend.

Lombardy

wines_Lombardia <- wines_Italia %>%
  filter(province == "Sicily & Sardinia") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_Lombardia

wines_Lombardia %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(50, 500)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Lombardia") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

Northeastern Italy

wines_NE <- wines_Italia %>%
  filter(province == "Northeastern Italy") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_NE

wines_NE %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(50, 700)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Italia di Nord-Orientale") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

Central Italy

wines_CI <- wines_Italia %>%
  filter(province == "Central Italy") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_CI

wines_CI %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(50, 400)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Italia Centrale") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

The piniot grigio variety (a white wine variety) is common here.

Southern Italy

wines_S <- wines_Italia %>%
  filter(province == "Southern Italy") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_S

wines_S %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(50, 300)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Italia Sud") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

The Aglianico (a wine variety known for being full-bodied with firm tannins and high acidity with a rich lavor) is most prominent here.

Northwestern Italy

wines_NI <- wines_Italia %>%
  filter(province == "Northwestern Italy") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_NI

wines_NI %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(0, 5)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino Italia Verso Nord") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

Rest of Italy

wines_Rest <- wines_Italia %>%
  filter(province == "Italy Other") %>%
  group_by(variety) %>%
  summarise(total = n()) %>%
  mutate(pcnt = round(100*(total/sum(total)), digits = 5), accum = cumsum(pcnt)) %>%
  arrange(desc(total))

wines_Rest

wines_Rest %>%
  head(10) %>%
  ggplot(aes(x = reorder(variety, -total), y = total)) + 
  geom_bar(fill = "wheat", stat = "identity") + 
  coord_cartesian(ylim = c(0, 20)) + 
  labs (x = "\n Varieta di Vino \n") + labs(y = "Total") +
  labs(title = "\n Il varieta di vino per il resto d'Italia") + 
  geom_text(aes(label = sprintf("%.1f %%", pcnt))) + 
  theme_solarized() + 
  theme(axis.text.x = element_text(angle = 90, vjust = 0.6, hjust=1.025))

Text Mining and Word Clouding

For reference, i will use course notes i got from William Surles’course notes from Datacamp course “Text Mining: Bag of Words”. While Ariel, Moreno, Orsini-Rosberg, Pi and et al focused more on the reviews, my focus here will be more on the description.

# Creating vector from descriptions
description <- wines_Italia$description

#Interpreting review vector with tm
review <- VectorSource(description)

# Creating VCorpus object
rev <- VCorpus(review)

# Removing punctuation from reviews
rev <- tm_map(rev, removePunctuation)

# Converting to lowercase to make future cleaning easier
rev <- tm_map(rev, content_transformer(tolower))

# Removing Numbers
rev <- tm_map(rev, removeNumbers)

Let’s remove all excess whitespaces and stop words, making use of tm’s package list of stopwords

#Removing stop words
rev <- tm_map(rev, removeWords, stopwords("en"))

#Remove excess white spaces
rev <- tm_map(rev, stripWhitespace)

We need to do word stemming in order to unify the natural language of our descriptions

rev <- tm_map(rev, stemDocument)

At the same time we also want to figure out what words are commonly used in the descriptions more often and generate a list of stems that appear at least 1000 times

# Creating a DTM matrix
rev_dtm <- DocumentTermMatrix(rev)

# Finding the most popular stems
rev_freq <- findFreqTerms(rev_dtm, lowfreq = 1000)
rev_freq

##  [1] "acid"       "age"        "almond"     "alongsid"   "appl"      
##  [6] "aroma"      "back"       "beauti"     "berri"      "black"     
## [11] "blackberri" "blend"      "bright"     "cabernet"   "cherri"    
## [16] "chocol"     "citrus"     "clean"      "close"      "cola"      
## [21] "color"      "concentr"   "creami"     "crisp"      "dark"      
## [26] "deliv"      "dri"        "drink"      "easi"       "eleg"      
## [31] "exot"       "express"    "finish"     "firm"       "flavor"    
## [36] "flower"     "follow"     "fresh"      "fruit"      "good"      
## [41] "grape"      "herb"       "here"       "honey"      "import"    
## [46] "intens"     "leather"    "licoric"    "light"      "long"      
## [51] "made"       "make"       "matur"      "meat"       "merlot"    
## [56] "miner"      "mouth"      "mouthfeel"  "nose"       "note"      
## [61] "oak"        "offer"      "open"       "pair"       "palat"     
## [66] "peach"      "pepper"     "plum"       "raspberri"  "red"       
## [71] "rich"       "ripe"       "sangioves"  "sauvignon"  "show"      
## [76] "slight"     "smooth"     "soft"       "spice"      "stone"     
## [81] "structur"   "sweet"      "tannin"     "textur"     "there"     
## [86] "thick"      "tight"      "toast"      "tobacco"    "tone"      
## [91] "touch"      "vanilla"    "well"       "white"      "wild"      
## [96] "wine"       "year"

Removing irrelevant words.

I want to retain words that are relevant to wines, hence i will remove words that are not relevant here.

new_stopwords <- c("age", "aroma", "alongsid", "drink", "easi", "express", "flavor","follow", "good", "grape", "here", "import", "intens", "long", "made", "make", "matur", "meat", "mouth", "mouthfeel", "nose", "note", "offer", "open", "palat", "pair", "show", "textur", "there", "tone", "touch","wine", "year")

rev_new <- tm_map(rev, removeWords, new_stopwords)

Let us create a second matrix to find the top word stems in wine reviews by only choosing the most relevant ones.

I will try to set the benchmark to a minimum of 2500 times

# Creating second DTM
rev_dtm2 <- DocumentTermMatrix(rev_new)

# Finding top stems in reviews
top_stems <- findFreqTerms(rev_dtm2, lowfreq = 2500)
top_stems

##  [1] "acid"   "berri"  "black"  "blend"  "bright" "cherri" "dri"   
##  [8] "finish" "fresh"  "fruit"  "ripe"   "spice"  "sweet"  "tannin"
## [15] "white"

I will try to create the corpus function for this:

clean_corpus <- function(corpus){
  corpus <- tm_map(corpus, removePunctuation)
  corpus <- tm_map(corpus, removeNumbers)
  corpus <- tm_map(corpus, content_transformer(tolower))
  corpus <- tm_map(corpus, removeWords, stopwords("en"))
  corpus <- tm_map(corpus, stemDocument)
  corpus <- tm_map(corpus, removeWords, new_stopwords)
  corpus <-tm_map(corpus, stripWhitespace)
  return(corpus)
}
italian_wines_corpus <- clean_corpus(rev)
italian_wines_corpus_dm <- TermDocumentMatrix(italian_wines_corpus)

Building the wordCloud and creating a plot for the top stems

# Creating matrix from TDM

italian_wines_corpus_cdm <- as.matrix(italian_wines_corpus_dm)

italianf <- rowSums(italian_wines_corpus_cdm)
italianf <- sort(italianf, decreasing = T)

# Creating the data frame 
iwf <- data.frame(term=names(italianf), num = italianf)

# Creating a plot of 10 top stems
ggplot(data = head(iwf, 10), aes(x = factor(term, levels = iwf$term[order(-iwf$num)]), y = num)) + 
  geom_col(fill = "orchid") + 
  labs(x = "Word Stems", y = "Count", title = "Most common stems in Italian Wine Reviews") + 
    scale_y_continuous(expand = c(0, 0), breaks=seq(0,10000,500))

# Creating a wordcloud

wordcloud(iwf$term, iwf$num, max.words = 25, color = "orchid")

We can see here that the most common words described here are related to berries or those related to the texture or spice of the wine.

Italian Wines

Joel Jr Rudinas

May 9, 2019

Introduction

About Author

About Dataset

Scope of the Project

Codes i will also partly reference from:

Loading the required libraries and data

Libraries

Data

Understanding the Data

Description of Variables.

Mutating (converting the price to euros)

Splitting only for Italian wines

Cleaning the Data

Data Exploration and Visualization

Province

Rating

Filtering wines via their rating category

For Excellent Wines (rated 90 and above)

For wines rated okay

Conducting a t.test on the provinces with the highest ratings

Price

Most expensive wine?.

T.test on the wine prices.

Variety.

Rating of variety

General Rating

For Excellent Wines (rated 90 and above)

For wines rated okay (89 and below)

Wine varieties per province

Tuscany

Piedmont

Veneto

Sicily and Sardinia

Lombardy

Northeastern Italy

Central Italy

Southern Italy

Northwestern Italy

Rest of Italy

Text Mining and Word Clouding

Removing irrelevant words.

Building the wordCloud and creating a plot for the top stems