Code of the model

Let us code the model of customer segmentation.

First we will import all important libraries in our juypter notebook

Now we will import the dataset and once check if the file is imported

Before moving further we can see that there is a spelling mistake of gender which is written as Genre. So let’s change that and check if it has been corrected.

Now we are all set to move to visualization of data

Let’s create a distribution plot for the 3 categories i.e. Age , Annual Income and Spending Score.

In this program snippet, A far loop has been created [because we need to represent the 3 variables]. Then we have implemented a subplot function. This represents number of rows, columns and the index which we will implement. Now let us run the program.

So in the first data plot we can see that age group 20-40 perform the highest shopping among all the all the age groups.

In the 2nd plot we can see that the people who have their annual income between $50,000 and $70,000 perform more shopping.

In the 3rd plot we can see that the most people have the score 50.

Now let us compare the number of female and male customers.

We will use the countplot function of sns to represent the comparison

From this we can conclude that majority of customers is female.

Now let us create a violin plot of age, annual income , and spending score based on gender using the seaborn library.

In the program/code the x-axis will represent the age, annual income and spending score whereas the y-axis will represent the gender.

Let us now the run the code.

The above violin plots represents the distribution of age, annual income and spending score based on gender

In the first violin plot we can see that female density is more that of male in the age group 30

In the second violin plot we can see that female density is more at $60,000 p.a. and male density is also around $60,000-$65,000

In the third violin plot, the density of both male and female density is around 50.

Before moving ahead lets describe our data

As we can see from the description that minimum age group is 18, maximum is 70 and so on we can derive conclusions from this which will help us to provide efficiency in the visualization

Let us now divide the age group into five categories to understand which age group has more density of customers.

In the program , we have divided the age into 5 categories.

On x-axis we will represent the category of age

On y-axis we will represent the number of customers in each category by using the length function.

Also for visualization of this data we will use bar plot

Let us now run the code:

From this bar graph we can easily understand that the age group 26-35 has maximum customers

For better understanding of data let us look into the relationship between annual income and spending score.

From this plot we can see that customers who have an annual income of $40,000 to $60,000 have a spending score of 60.

Moving ahead let us divide the annual income into 5 categories as we did earlier with the age category.

In the program, we have divided the annual income into five categories.

The x-axis represents the range of income.

The y-axis will represent the number of customers.

Now let us run the code:

From the bar graph we can see that most customers have annual income of $60,000-$90,000

Similarly we will do with the spending scores i.e. dividing it into 5 categories

Let us run the code:

We can see from the bar graph that maximum customers have a spending score of 41-60

In this way we have analysied the data thoroughly.

From here clustering process starts.

Now first we will find the optimum number of clusters that we need to create.

We will first cluster the data of age and spending score

First import the Kmeans and now we will draw a graph to find the optimum number of clusters for the 2 variables we considered for a range of 1-11 in X1

Let us run the code:

From the graph we can see that, it becomes little bit constant after 4.

So let the number of optimum clusters = 4

We can see that clusters here have been divided into 4 parts i.e. 0,1,2 and 3

Now we will check the centroids of clusters

Now let us visualize the cluster using scatterplot.

Let us now run the code:

In this we can see that centroid has been given black color and rest all labels are given a different color.

Similarly we will make the cluster of annual income and spending score.

First finding the optimum number of clusters through graph

Let us run the code:

From graph let us take the number of optimum clusters to be 5

It has been divided into 5 clusters i.e. 0,1,2,3 and 4

Let us check the centroid

Now let us plot the clusters.

Now before making the 3d model let us first remove the customer id from the data set otherwise the model will give error as

ValueError: could not convert string to float: 'Male'

Finally let us cluster all the 3 variables i.e. age, annual income and spending scores.

First finding the optimum number of clusters through graph

Let us now run the code:

From graph let the optimum number of clusters be 5

It has been divided into 5 clusters i.e. 0,1,2,3 and 4

Let us check the centroid

Now let us plot the clusters. We will make the 3d model of customer segmentation by using mpl_toolkits

Let us now run the code:

So in this way we have a created a 3d model of customer segmentation which is useful for making any business efficient and profitable