Optimizing Wine Marketing: A Cluster Analysis
Intan M Sari
2024-08-23
1 Introduction
This project involves applying unsupervised clustering to customer data from a wine farmer and seller. The data are the results of a chemical analysis of wines grown in the same region in Italy but derived from three different cultivars. The analysis determined the quantities of 13 constituents found in each of the three types of wines.
By grouping wines based on their similarities, I aim to provide a more customer-centric approach to product offerings and marketing, tailoring them to the individual needs of different customer segments.
2 Business Question
The goal of this analysis is to identify the optimal marketing strategies and branding approaches for each distinct customer segment.
3 Preparation
For our analysis, we will employ the Wine dataset sourced from Kaggle. The following descriptions outline the dataset’s features:
| Column Name | Data Type | Description |
|---|---|---|
| Alcohol | double | Alcohol content |
| Malic_Acid | double | Malic acid content |
| Ash | double | Ash content |
| Ash_Alcanity | double | Ash alkalinity |
| Magnesium | integer | Magnesium content |
| Total_Phenols | double | Total phenols content |
| Flavanoids | double | Flavanoids content |
| Nonflavanoid_Phenols | double | Nonflavanoid phenols content |
| Proanthocyanins | double | Proanthocyanins content |
| Color_Intensity | double | Color intensity |
| Hue | double | Hue |
| OD280 | double | Optical density at 280 nm |
| Proline | integer | Proline content |
| Customer_Segment | integer | Customer segment (presumably a numerical identifier) |
3.1 Importing Libraries
Next, we must import the necessary libraries.
library(tidyr)
library(factoextra)
library(cluster)
library(FactoMineR)
library(stringr)
library(tibble)
library(tidyverse)
library(cluster)
library(ggplot2)
library(grid)
library(gridExtra)
library(GGally)3.2 Importing Data
Following that, we incorporate our data into the wine
dataset.
wine <- read.csv("data_input/Wine.csv")
head(wine)4 Data Cleansing and Exploratory Data Analysis (EDA)
Before proceeding with data cleaning and feature engineering, we’ll
conduct a preliminary exploration of the imported dataset. This involves
examining the initial and final rows using the head() and
tail() functions. This step will provide valuable insights
into the dataset’s structure and content.
head(wine)tail(wine)In this step, we are going to inspect our dataset by checking its data types, and check for missing values.
Result
Structure with Glimpse
To find out the suitable data type, it is checked first with the `glimpse() function.
wine %>%
glimpse()#> Rows: 178
#> Columns: 14
#> $ Alcohol <dbl> 14.23, 13.20, 13.16, 14.37, 13.24, 14.20, 14.39, …
#> $ Malic_Acid <dbl> 1.71, 1.78, 2.36, 1.95, 2.59, 1.76, 1.87, 2.15, 1…
#> $ Ash <dbl> 2.43, 2.14, 2.67, 2.50, 2.87, 2.45, 2.45, 2.61, 2…
#> $ Ash_Alcanity <dbl> 15.6, 11.2, 18.6, 16.8, 21.0, 15.2, 14.6, 17.6, 1…
#> $ Magnesium <int> 127, 100, 101, 113, 118, 112, 96, 121, 97, 98, 10…
#> $ Total_Phenols <dbl> 2.80, 2.65, 2.80, 3.85, 2.80, 3.27, 2.50, 2.60, 2…
#> $ Flavanoids <dbl> 3.06, 2.76, 3.24, 3.49, 2.69, 3.39, 2.52, 2.51, 2…
#> $ Nonflavanoid_Phenols <dbl> 0.28, 0.26, 0.30, 0.24, 0.39, 0.34, 0.30, 0.31, 0…
#> $ Proanthocyanins <dbl> 2.29, 1.28, 2.81, 2.18, 1.82, 1.97, 1.98, 1.25, 1…
#> $ Color_Intensity <dbl> 5.64, 4.38, 5.68, 7.80, 4.32, 6.75, 5.25, 5.05, 5…
#> $ Hue <dbl> 1.04, 1.05, 1.03, 0.86, 1.04, 1.05, 1.02, 1.06, 1…
#> $ OD280 <dbl> 3.92, 3.40, 3.17, 3.45, 2.93, 2.85, 3.58, 3.58, 2…
#> $ Proline <int> 1065, 1050, 1185, 1480, 735, 1450, 1290, 1295, 10…
#> $ Customer_Segment <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1…Data Cleansing & Missing Value
Here we need to change data type for Customer_Segment column from
integer to factor
wine_cl <-
wine %>%
mutate(Customer_Segment = as.factor(Customer_Segment))Next we check for missing values using is.na
wine_cl %>%
is.na() %>%
colSums()#> Alcohol Malic_Acid Ash
#> 0 0 0
#> Ash_Alcanity Magnesium Total_Phenols
#> 0 0 0
#> Flavanoids Nonflavanoid_Phenols Proanthocyanins
#> 0 0 0
#> Color_Intensity Hue OD280
#> 0 0 0
#> Proline Customer_Segment
#> 0 0Insights:
The wine dataset stands out for having no missing data
at all. This means that each data point in the dataset has values for
every variable that is defined.
Descriptive Statistics
To gain a comprehensive understanding of the dataset, we can use:
summary(wine_cl)#> Alcohol Malic_Acid Ash Ash_Alcanity
#> Min. :11.03 Min. :0.740 Min. :1.360 Min. :10.60
#> 1st Qu.:12.36 1st Qu.:1.603 1st Qu.:2.210 1st Qu.:17.20
#> Median :13.05 Median :1.865 Median :2.360 Median :19.50
#> Mean :13.00 Mean :2.336 Mean :2.367 Mean :19.49
#> 3rd Qu.:13.68 3rd Qu.:3.083 3rd Qu.:2.558 3rd Qu.:21.50
#> Max. :14.83 Max. :5.800 Max. :3.230 Max. :30.00
#> Magnesium Total_Phenols Flavanoids Nonflavanoid_Phenols
#> Min. : 70.00 Min. :0.980 Min. :0.340 Min. :0.1300
#> 1st Qu.: 88.00 1st Qu.:1.742 1st Qu.:1.205 1st Qu.:0.2700
#> Median : 98.00 Median :2.355 Median :2.135 Median :0.3400
#> Mean : 99.74 Mean :2.295 Mean :2.029 Mean :0.3619
#> 3rd Qu.:107.00 3rd Qu.:2.800 3rd Qu.:2.875 3rd Qu.:0.4375
#> Max. :162.00 Max. :3.880 Max. :5.080 Max. :0.6600
#> Proanthocyanins Color_Intensity Hue OD280
#> Min. :0.410 Min. : 1.280 Min. :0.4800 Min. :1.270
#> 1st Qu.:1.250 1st Qu.: 3.220 1st Qu.:0.7825 1st Qu.:1.938
#> Median :1.555 Median : 4.690 Median :0.9650 Median :2.780
#> Mean :1.591 Mean : 5.058 Mean :0.9574 Mean :2.612
#> 3rd Qu.:1.950 3rd Qu.: 6.200 3rd Qu.:1.1200 3rd Qu.:3.170
#> Max. :3.580 Max. :13.000 Max. :1.7100 Max. :4.000
#> Proline Customer_Segment
#> Min. : 278.0 1:59
#> 1st Qu.: 500.5 2:71
#> Median : 673.5 3:48
#> Mean : 746.9
#> 3rd Qu.: 985.0
#> Max. :1680.0and also we can visualize its column using boxplot
function
boxplots <- list()
for (i in 1:ncol(wine_cl)) {
col_name <- colnames(wine_cl)[i]
boxplots[[i]] <- ggplot(wine_cl, aes(x = "", y = .data[[col_name]])) +
geom_boxplot() +
labs(title = paste(col_name))
}
grid.arrange(grobs = boxplots, ncol = 5) 
Insights:
General Observations:
- Skewness: A number of variables (such as Alcohol, Malic_Acid, Ash,
Ash_Alcanity, Magnesium, Total_Phenols, Flavanoids, Proanthocyanins,
Color_Intensity, Hue, OD280, Proline) seem to have a right-skewed
distribution, indicating a longer tail on the right side. This implies
that there are possibly some outliers causing the mean to be shifted
towards the upper range.
- Variability: Some variables have wide ranges (e.g., Proline), suggesting there could be notable diversity in the data.
Specifics Observations:
- Segments of customers: The Customer_Segment variable contains three separate values (1, 2, and 3), indicating a possible division of data into three groups or clusters.
- Alcohol Percentage: The alcohol percentage varies between 11.03% and
14.83%, with an average of 13.00%. This indicates that the typical wine
in the dataset has a moderate level of alcohol.
- Acidity is measured by the variables Malic Acid and Ash Alcanity.
The variables’ distributions indicate that there is a variety of acidity
levels present in the wines.
- Phenolic compounds in wine such as Total_Phenols, Flavanoids,
Nonflavanoid_Phenols, and Proanthocyanins are crucial for flavor and
color. The variable distributions indicate various levels of phenolic
compounds.
- Intensity of Color: The Color_Intensity variable assesses the
strength of the color of the wine. The distribution indicates a variety
of color intensities, with certain wines exhibiting higher color
intensity than others.
- Proline, an amino acid, can be found in wine. The Proline distribution indicates variability in Proline levels among wines, with some wines containing notably higher levels than others.
5 PCA (Principal Component Analysis)
5.1 Data Preprocessing
The following code automatically identifies the positions of numeric and categorical variables, making it adaptable to datasets with many columns.
quanti <- wine_cl %>%
select_if(is.numeric) %>%
colnames()
quantivar <- which(colnames(wine_cl) %in% quanti)
quali <- wine_cl %>%
select_if(is.factor) %>%
colnames()
qualivar <- which(colnames(wine_cl) %in% quali)wine_pca <- PCA(
X = wine_cl,
scale.unit = T, # scaling data
quali.sup = qualivar,
graph = F,
ncp = 13
)
wine_pca#> **Results for the Principal Component Analysis (PCA)**
#> The analysis was performed on 178 individuals, described by 14 variables
#> *The results are available in the following objects:
#>
#> name description
#> 1 "$eig" "eigenvalues"
#> 2 "$var" "results for the variables"
#> 3 "$var$coord" "coord. for the variables"
#> 4 "$var$cor" "correlations variables - dimensions"
#> 5 "$var$cos2" "cos2 for the variables"
#> 6 "$var$contrib" "contributions of the variables"
#> 7 "$ind" "results for the individuals"
#> 8 "$ind$coord" "coord. for the individuals"
#> 9 "$ind$cos2" "cos2 for the individuals"
#> 10 "$ind$contrib" "contributions of the individuals"
#> 11 "$quali.sup" "results for the supplementary categorical variables"
#> 12 "$quali.sup$coord" "coord. for the supplementary categories"
#> 13 "$quali.sup$v.test" "v-test of the supplementary categories"
#> 14 "$call" "summary statistics"
#> 15 "$call$centre" "mean of the variables"
#> 16 "$call$ecart.type" "standard error of the variables"
#> 17 "$call$row.w" "weights for the individuals"
#> 18 "$call$col.w" "weights for the variables"5.2 Data Visualizing
Subsequently, we will examine each factor plot separately and pinpoint the five most extreme outliers.
plot.PCA(
x = wine_pca,
choix = "ind",
select = "contrib 5",
invisible = "quali",
habillage = "Customer_Segment"
)
Insights:
- Based on the visual representation, the points corresponding to Customer IDs 4, 15, 19, 147, and 178 appear to be outliers. These points are located far from the main clusters of data, suggesting that they have unique characteristics or behaviors that deviate significantly from the norm.
- Clustering: The plot shows distinct clustering among the customer segments. This suggests that the segments exhibit different patterns or characteristics based on the underlying variables captured by the principal components.
- Separation: Customer Segment 3 appears to be well-separated from the other segments, particularly along Dim 1. This could indicate that this segment is characterized by a unique combination of attributes that distinguishes it from the rest.
- Overlapping: Customer Segments 1 and 2 show some overlap, suggesting that they share similarities in terms of the variables represented by the principal components. However, there’s still a degree of separation, indicating that they also have distinct characteristics.
- Importance of Dimensions: The percentages associated with Dim 1 (36.20%) and Dim 2 (19.21%) indicate the relative importance of these components in explaining the variation in the data. Dim 1 is more influential, suggesting that the variations captured by Dim 1 are more significant in differentiating the customer segments.
For a more comprehensive understanding of variable distribution, the
fviz_contrib() function can be employed. This tool offers a
visual representation of how rows or columns contribute to the principal
components (PCs) in a Principal Component Analysis (PCA), providing
valuable insights into the data’s underlying structure.
fviz_contrib(X = wine_pca,
axes = 1, # = PC1
choice = "var")
Notes:
- The red dotted horizontal line indicates the anticipated contribution level when the principal component extracts information evenly from all variables.
- To calculate it, we divide 100% by the original number of variables.
- In this specific instance, the calculation is 100 divided by 13, resulting in 7.7%.
- If a variable’s contribution surpasses this threshold, it indicates that the variable is influential in forming that particular principal component.
Correlation between Variable
Next, we’ll determine the relationships between variables using the
ggcorr() function.
ggcorr(wine_cl, label = TRUE, hjust = 1)
Insights:
Strong Positive Correlations:
- Total_Phenols and Flavanoids: These two variables have a very strong positive correlation, as indicated by the dark red square at their intersection. This suggests that wines with higher levels of Total_Phenols tend to also have higher levels of Flavanoids.
- Flavanoids and OD280: Another strong positive correlation exists between Flavanoids and OD280.
- Total_Phenols and OD280: A strong positive correlation is also observed between Total_Phenols and OD280.
Negative Correlations:
- Color_Intensity and Proanthocyanins: There is a moderate negative correlation between Color_Intensity and Proanthocyanins. This suggests that wines with higher Color_Intensity tend to have lower levels of Proanthocyanins.
Chemical Composition: The strong positive correlations between Total_Phenols, Flavanoids, and OD280 might suggest that these compounds are chemically related or co-occur in wine.
Sensory Attributes: The negative correlation between Color_Intensity and Proanthocyanins could indicate that wines with a more intense color might have different sensory characteristics related to Proanthocyanins.
6 Clustering
Next, we need to remove Customer_Segment from our
dataframe and store it in new dataframe called
wine_clean.
wine_clean <-
wine_cl %>%
select(-Customer_Segment)The resulting data frame now only contains columns that are of
integer, double, or numeric data types. Next, we need to ensure that
different variables contribute equally to the analysis, this can be done
by using scale() function. The scale()
function standardizes numerical data by subtracting the mean of each
column and dividing by its standard deviation. This transforms the data
to have a mean of 0 and a standard deviation of 1.
wine_z <- scale(wine_clean)
summary(wine_z)#> Alcohol Malic_Acid Ash Ash_Alcanity
#> Min. :-2.42739 Min. :-1.4290 Min. :-3.66881 Min. :-2.663505
#> 1st Qu.:-0.78603 1st Qu.:-0.6569 1st Qu.:-0.57051 1st Qu.:-0.687199
#> Median : 0.06083 Median :-0.4219 Median :-0.02375 Median : 0.001514
#> Mean : 0.00000 Mean : 0.0000 Mean : 0.00000 Mean : 0.000000
#> 3rd Qu.: 0.83378 3rd Qu.: 0.6679 3rd Qu.: 0.69615 3rd Qu.: 0.600395
#> Max. : 2.25341 Max. : 3.1004 Max. : 3.14745 Max. : 3.145637
#> Magnesium Total_Phenols Flavanoids Nonflavanoid_Phenols
#> Min. :-2.0824 Min. :-2.10132 Min. :-1.6912 Min. :-1.8630
#> 1st Qu.:-0.8221 1st Qu.:-0.88298 1st Qu.:-0.8252 1st Qu.:-0.7381
#> Median :-0.1219 Median : 0.09569 Median : 0.1059 Median :-0.1756
#> Mean : 0.0000 Mean : 0.00000 Mean : 0.0000 Mean : 0.0000
#> 3rd Qu.: 0.5082 3rd Qu.: 0.80672 3rd Qu.: 0.8467 3rd Qu.: 0.6078
#> Max. : 4.3591 Max. : 2.53237 Max. : 3.0542 Max. : 2.3956
#> Proanthocyanins Color_Intensity Hue OD280
#> Min. :-2.06321 Min. :-1.6297 Min. :-2.08884 Min. :-1.8897
#> 1st Qu.:-0.59560 1st Qu.:-0.7929 1st Qu.:-0.76540 1st Qu.:-0.9496
#> Median :-0.06272 Median :-0.1588 Median : 0.03303 Median : 0.2371
#> Mean : 0.00000 Mean : 0.0000 Mean : 0.00000 Mean : 0.0000
#> 3rd Qu.: 0.62741 3rd Qu.: 0.4926 3rd Qu.: 0.71116 3rd Qu.: 0.7864
#> Max. : 3.47527 Max. : 3.4258 Max. : 3.29241 Max. : 1.9554
#> Proline
#> Min. :-1.4890
#> 1st Qu.:-0.7824
#> Median :-0.2331
#> Mean : 0.0000
#> 3rd Qu.: 0.7561
#> Max. : 2.9631K-means
Next, we need to partitions data into K clusters based on the mean distance to cluster centers using K-mean algorithm.
Parameters:
- x: dataset
- centers: number of centroids (k)
Note: It is necessary to use set.seed() due to the
random initialization in the initial stages of the K-means
algorithm.
RNGkind(sample.kind = "Rounding")
set.seed(100)
wine_km <- kmeans(
x = wine_z,
centers = 5
)next bit is the result of k-means() function:
wine_km$centers#> Alcohol Malic_Acid Ash Ash_Alcanity Magnesium Total_Phenols
#> 1 -0.1886092 -0.3973055 0.88933287 0.3084403 1.11034100 0.4408200
#> 2 0.1766166 0.9039567 0.21536153 0.5494898 -0.07712756 -0.9873154
#> 3 -0.7932129 -0.2670132 -1.06563358 -0.2679823 -0.68011948 0.2781082
#> 4 1.0751808 -0.3606243 0.16642266 -0.8929012 0.46102024 0.9849139
#> 5 -0.9985113 -0.4250204 0.08279355 0.6568276 -0.57973353 -0.5680217
#> Flavanoids Nonflavanoid_Phenols Proanthocyanins Color_Intensity Hue
#> 1 0.5438510 -0.3483549 0.5933360 -0.5275837 0.7570958
#> 2 -1.2236663 0.7114800 -0.7591372 0.9516989 -1.1867156
#> 3 0.2668683 -0.5684285 0.1678066 -0.8107192 0.2991794
#> 4 1.0567190 -0.6716820 0.6866569 0.3596910 0.4267835
#> 5 -0.3042323 0.8751497 -0.4437352 -0.9381988 0.5304372
#> OD280 Proline
#> 1 0.61524013 0.1902360
#> 2 -1.28577141 -0.3952058
#> 3 0.50311030 -0.7139802
#> 4 0.79065585 1.3330286
#> 5 -0.09609123 -0.7561714Optional step when we don’t have a predetermined value for K based on business decisions.
The quality of clustering results can be evaluated using three metrics:
- Within Sum of Squares (
$withinss): This measures the sum of squared distances from each data point to its assigned cluster centroid. - Between Sum of Squares (
$betweenss): This measures the sum of squared distances, weighted by cluster size, from each cluster centroid to the overall mean. - Total Sum of Squares (
$totss): This measures the total sum of squared distances from each data point to the overall mean.
wine_km$tot.withinss#> [1] 1106.418wine_km$betweenss / wine_km$totss#> [1] 0.519158Insights:
- Within-Cluster Variation: The value of 1106.418 for
tot.withinsssuggests that there is a moderate level of within-cluster variation. This means that data points within their respective clusters are reasonably close together, but there might be some room for improvement in terms of compactness. - Explained Variance Ratio (EVR): The explained variance ratio of 0.519158 indicates that the clustering is moderately effective in separating the data into distinct groups. This means that the clusters are reasonably well-separated, but there might be some overlap between them. The higher ratio close to 1, the better because it represent actual data distribution.
7 Parameter Tuning: Determining the Optimal Number of Clusters ( \(k\) )
To determine optimum number of clusters (\(k\)), objectively we can use a method called Elbow method. The elbow method is a common technique used to determine the optimal number of clusters (\(k\)) in K-means clustering. It involves plotting the Within-Cluster Sum of Squares (WSS) against the number of clusters. The “elbow” point in the plot, where the rate of decrease in WSS starts to slow down significantly, is often considered the optimal number of clusters.
elbow_plot <- fviz_nbclust(wine_z,
FUNcluster = kmeans,
method = "wss")
k_optimal <- 3
elbow_plot + geom_vline(xintercept = k_optimal, linetype = "dashed")
Insight :
Based on the visualization provided, the optimal \(k\) value for your dataset is 3. This means that dividing the data into 3 clusters will provide good and representative clustering results.
Re-run K-means with the Optimal Number of Clusters (\(k\)):
RNGkind(sample.kind = "Rounding")
set.seed(100)
wine_km_opt <- kmeans(x = wine_z,
centers = 3)8 Cluster Interpretation and Profiling
Next, we need to analyze the characteristics of each cluster and understand how the data points are grouped based on their similarity.
wine_km_opt$cluster#> [1] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
#> [38] 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 3 2 3 3 3 3 3 3 3 3 3 3 3 1
#> [75] 3 3 3 3 3 3 3 3 3 2 3 3 3 3 3 3 3 3 3 3 3 1 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
#> [112] 3 3 3 3 3 3 3 2 3 3 1 3 3 3 3 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2
#> [149] 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2The output wine_km_opt$cluster shows the cluster
assignments for each data point in the wine_z dataset. The
values in the output represent the cluster labels assigned to each
observation.
wine$kelompok <- wine_km_opt$cluster
wine_cluster <-
wine %>%
group_by(kelompok) %>%
summarise_all(mean)
wine_clusterCluster Profiling:
Based on above code bit and the result we can profile our clusters as follows:
Cluster 1: Premium Wine
- Highest: Alcohol, Ash, Magnesium, Total Phenols, Flavanoids, Proanthocyanins, Hue, OD280, Proline
- Lowest: Malic Acid, Ash Alcanity, Nonflavanoid Phenols
- Label: Premium Wine (premium quality)
Cluster 2: Unique Wine
- Highest: Malic Acid, Ash Alcanity, Nonflavanoid Phenols, Color Intensity
- Lowest: Total Phenols, Flavanoids, Proanthocyanins, Hue, OD280
- Label: Unique Wine (have a distinctive flavor)
Cluster 3: Fresh Wine
- Highest: None
- Lowest: Alcohol, Malic Acid, Ash, Magnesium, Color Intensity, Proline
- Label: Fresh Wine (have a light flavor)
9 Conclusion
Based on the cluster analysis, we can identify three distinct customer segments
Cluster 1: Premium Wine
- Characteristics: High levels of key wine components, suggesting premium quality and complexity.
- Marketing Strategies: Target high-end wine connoisseurs and collectors. Emphasize the wine’s exceptional quality, complexity, and unique flavor profile. Consider premium packaging and distribution channels.
- Branding Approaches: Position the brand as a luxury, exclusive product. Use elegant and sophisticated branding elements.
Cluster 2: Unique Wine
- Characteristics: Distinctive flavor profile with higher levels of Malic Acid, Ash Alcanity, and Nonflavanoid Phenols.
- Marketing Strategies: Target consumers who appreciate unique and unconventional wines. Highlight the wine’s distinctive taste and its ability to pair well with specific foods.
- Branding Approaches: Create a brand identity that reflects the wine’s unique character. Use bold and eye-catching visuals.
Cluster 3: Fresh Wine
- Characteristics: Lighter flavor profile with lower levels of most components.
- Marketing Strategies: Target consumers who prefer lighter, more refreshing wines. Emphasize the wine’s freshness and drinkability.
- Branding Approaches: Create a brand identity that conveys a sense of lightness and freshness. Use vibrant colors and imagery.
10 Reference
Sadegh Jalalian. Wine Customer Segmentation. Retrieved from https://www.kaggle.com/datasets/sadeghjalalian/wine-customer-segmentation/data