Data Import and Initial Processing

1. Importing the Data
   • The core of our analysis begins with the importation of the ramen reviews dataset. Utilizing the readr package, we efficiently load the ramen-ratings.csv file. This package is chosen for its speed and ability to handle larger datasets more effectively than the base R functions. Additionally, we handle missing values at this stage by specifying empty strings and the term “Unrated” as NA (Not Available) during the import process.
   • Alongside the ramen reviews, we import a second dataset, `Countries

-Continents.csv`, using the same method. This dataset is anticipated to provide valuable geographical context for our analysis, linking each ramen review to its corresponding continent.

2. Initial Exploration of the Data
   • A primary inspection of both datasets is conducted using the glimpse() function. This function offers a quick overview of the data structure, including the types of variables and the first few entries in each column. This step is crucial for familiarizing ourselves with the dataset’s layout, identifying any apparent inconsistencies, and planning subsequent data processing steps.

# Reading the data
ramen <- readr::read_csv("../data/ramen-ratings.csv", na = c("", "Unrated"))
country_continent <- readr::read_csv("../data/Countries-Continents.csv")

# Initial glimpse (or str) at the data
glimpse(ramen)

## Rows: 2,580
## Columns: 7
## $ `Review #` <dbl> 2580, 2579, 2578, 2577, 2576, 2575, 2574, 2573, 2572, 2571,~
## $ Brand      <chr> "New Touch", "Just Way", "Nissin", "Wei Lih", "Ching's Secr~
## $ Variety    <chr> "T's Restaurant Tantanmen", "Noodles Spicy Hot Sesame Spicy~
## $ Style      <chr> "Cup", "Pack", "Cup", "Pack", "Pack", "Pack", "Cup", "Tray"~
## $ Country    <chr> "Japan", "Taiwan", "USA", "Taiwan", "India", "South Korea",~
## $ Stars      <dbl> 3.75, 1.00, 2.25, 2.75, 3.75, 4.75, 4.00, 3.75, 0.25, 2.50,~
## $ `Top Ten`  <chr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA,~

glimpse(country_continent)

## Rows: 194
## Columns: 2
## $ Continent <chr> "Africa", "Africa", "Africa", "Africa", "Africa", "Africa", ~
## $ Country   <chr> "Algeria", "Angola", "Benin", "Botswana", "Burkina", "Burund~

Overview of Ramen Reviews

Take a look to the first row :

head(ramen)

## # A tibble: 6 x 7
##   `Review #` Brand          Variety                Style Country Stars `Top Ten`
##        <dbl> <chr>          <chr>                  <chr> <chr>   <dbl> <chr>    
## 1       2580 New Touch      T's Restaurant Tantan~ Cup   Japan    3.75 <NA>     
## 2       2579 Just Way       Noodles Spicy Hot Ses~ Pack  Taiwan   1    <NA>     
## 3       2578 Nissin         Cup Noodles Chicken V~ Cup   USA      2.25 <NA>     
## 4       2577 Wei Lih        GGE Ramen Snack Tomat~ Pack  Taiwan   2.75 <NA>     
## 5       2576 Ching's Secret Singapore Curry        Pack  India    3.75 <NA>     
## 6       2575 Samyang Foods  Kimchi song Song Ramen Pack  South ~  4.75 <NA>

var_names <- colnames(ramen)

In the ramen, there are 2580 ramen reviews and 7 columns.

For each review, there are informations the following informations :Review #, Brand, Variety, Style, Country, Stars, Top Ten.

Count of Reviews by Brand

# Calculate the number of unique brands
unique_brands_count <- length(unique(ramen$Brand))

# Count the number of reviews for each brand
brand_reviews_count <- ramen %>%
  group_by(Brand) %>%
  summarize(ReviewCount = n()) %>% 
  arrange(desc(ReviewCount))


# Calculate the number of brands with less than 5 reviews
Brand_reviews_less_5 <- sum(brand_reviews_count$ReviewCount < 5)

# Calculate the percentage of brands with less than 5 reviews
Brand_reviews_less_5_percent <- (Brand_reviews_less_5 / unique_brands_count) * 100

There are a total of 355 unique brands, but 233 of these brands have fewer than 5 reviews, which constitutes 65.6338028% of all brands.

Data Cleaning and Preprocessing

1. Removal of Sparse Columns
   • The Top.Ten column, characterized by a high frequency of missing values, was identified as a candidate for removal. This step is crucial to enhance the dataset’s quality and focus on more reliable variables.
2. Filtering Out Uncommon Ramen Styles
   • We filtered out ramen styles that are rare in our dataset (‘Bar’, ‘Box’, and ‘Can’). This decision aligns with our objective to concentrate on more common and representative ramen styles for a more focused analysis.
3. Conversion of Rating Scale
   • The Stars column, representing the ramen quality rating, was transformed from its original format to a numeric scale. This transformation is vital for enabling quantitative analysis and comparison across different ramen products.
4. Exclusion of Incomplete Records
   • Our dataset cleansing included the removal of any records with missing data across any columns, ensuring completeness and reliability in the subsequent stages of our analysis.

Data Enrichment and Integration

1. Augmentation with Geographic Data
   • By performing an inner join with the country_continent dataset, we have enriched our ramen reviews with continent information. This enrichment allows us to explore geographical trends and patterns in ramen preferences and ratings.

Text Data Transformation

1. Unpacking the Variety Column
   • A key step in our text analysis involved breaking down the Variety column, which contains descriptive text about each ramen product, into individual words. This process was essential for uncovering common themes and descriptors used in the product variety.
2. Identification of Common Words
   • We filtered the words to retain only those appearing 100 or more times, categorizing them as “common words.” This focus helps us understand the most prevalent terms and descriptors in our dataset, shedding light on popular characteristics or ingredients in ramen varieties.
3. Pattern Creation for Advanced Filtering
   • Using the identified common words, we crafted a regex pattern to filter our dataset further. This pattern allows us to retain reviews where the Variety description includes any of these common terms, thereby focusing our analysis on the most representative and frequently mentioned attributes in ramen varieties.

Determining the Third Quartile of Reviews

1. Counting Reviews per Brand
   • The analysis begins with the creation of a new data frame ramen_nb_commentaire which focuses on the count of reviews for each ramen brand. This is achieved by selecting the Brand column from the ramen data, grouping the data by Brand, and then tallying the number of occurrences for each brand. This process helps in understanding how many reviews each brand has received, which is crucial for identifying popular or frequently reviewed brands.
2. Calculating the Third Quartile of Review Counts
   • The script calculates the third quartile (75th percentile) of the review counts across all brands using the quantile function. The value nb_commentaire_075 represents the threshold above which the top 25% of brands fall in terms of the number of reviews. Brands that have a number of reviews greater than or equal to this threshold are considered to be more commonly reviewed or possibly more popular.

Filtering Brands Based on Review Counts

1. Identifying Brands Above the Threshold
   • A second data frame ramen_brand is created, which once again tallies the number of reviews per brand but also filters out those brands that do not meet the third quartile threshold. This is a key step as it narrows down the dataset to brands that are more commonly reviewed, ensuring that the analysis focuses on brands that have a significant presence in the dataset.
2. Refining the Main Dataset
   • The final step involves filtering the main ramen dataset to include only those brands that are identified in the ramen_brand data frame. This is done using the %in% operator, which checks if the Brand in the ramen dataset is present in the ramen_brand dataset. This filtering step ensures that subsequent analyses are concentrated on the most frequently reviewed brands, providing a more focused and potentially more insightful examination of the data.

Rationale Behind Excluding These Styles

1. Focus on Mainstream Ramen Varieties
   • The decision to exclude these particular styles—Bar, Box, and Can—likely stems from an analytical focus on more traditional or mainstream types of ramen. These excluded styles might represent a minor portion of the market, specialty products, or formats that are not central to the primary interest of the analysis.
2. Enhancing Data Relevance
   • By removing less common or less relevant styles, the analysis becomes more streamlined and focused on the types of ramen that are more commonly consumed or reviewed. This helps in ensuring that the findings and insights derived from the dataset are applicable to a broader audience and reflect more general trends in ramen consumption and preference.

# Handling missing values and outliers
ramen <- ramen %>%
  select(-`Top Ten`) %>% # Removing Top.Ten due to high NA
  filter(Style != "Bar", Style != "Box", Style != "Can") %>% # Filtering out rare styles
  mutate(Stars = as.numeric(as.character(Stars))) %>% # Convert Stars to numeric
  drop_na() # Drop rows with NA

# Joining with country_continent data
ramen <- inner_join(ramen, country_continent, by = "Country")

# Transforming text data in Variety
ramen_words <- ramen %>% 
  unnest_tokens(word, Variety) %>% 
  count(word, sort = TRUE)

# ramen <- ramen %>% unnest_tokens(word, Variety) %>% 
#   dcast(Review.. + Brand + Style + Country + Stars  ~ word)

# Filter for common words (e.g., n >= 100 most frequent)
top_words <- ramen_words %>% filter(n >= 100)

# Create a regular expression pattern that matches any of the top words
pattern <- top_words$word %>% 
  paste(collapse = "|") # Collapse into a single string separated by '|'

# Filter ramen reviews where Variety contains any of the top words
ramen <- ramen %>%
  filter(str_detect(Variety, pattern))

# Determining the 3rd quartile of review counts by brand
ramen_brand_reviews <- ramen %>%
  count(Brand) %>%
  filter(n >= quantile(n, 0.75))

# Filtering the main dataset to include only brands above the 3rd quartile of review counts
ramen <- ramen %>%
  filter(Brand %in% ramen_brand_reviews$Brand)

# Focus on Mainstream Ramen Varieties
ramen <- ramen %>% 
  filter(!Style %in% c("Bar", "Box", "Can"))

There is a significant reduction in the number of reviews from over 2500 to just 49 after applying the filtering based on the third quartile of review counts per brand.

Summary of Findings from the Ramen Reviews Analysis

Our comprehensive analysis of the ramen reviews dataset has led to several insightful findings:

1. Brand Diversity: The dataset features a significant number of unique ramen brands, indicating a diverse and competitive market.

2. Review Concentration: A substantial proportion of brands have fewer than 5 reviews, suggesting market dominance by a few popular brands.

3. Variety in Flavors and Styles: The wide array of ramen flavors and styles, primarily from Asia, highlights the culinary diversity and broad consumer preferences in the ramen industry.

4. High Median Ratings: The generally high median star ratings across the dataset suggest overall customer satisfaction with the ramen products.

5. Brand Performance Insights: Sorting brands by median ratings revealed differences in brand performance, providing valuable insights for consumers and manufacturers about quality and preferences.

6. Market Segmentation Potential: The analysis points to possible market segments, useful for targeted marketing and product development in the ramen industry.

In conclusion, this analysis not only sheds light on the diverse and complex landscape of the ramen market but also opens avenues for further in-depth studies on consumer behavior, market trends, and competitive strategies within the food industry.