The most famous albania cell phone number list non-hierarchical cluster analysis method, which groups data without creating a tree diagram, is the "k-means" method.
Suppose you have some data scattered on a coordinate system and you want to split it into two clusters.
First, two random points (centroids) are plotted on the coordinate system, and clusters are generated from those closest to each centroid.
In the end, two clusters are generated for each two distinct centroids, achieving our goal.
The method of hierarchical cluster analysis explained so far is visually easy to understand, but it is not suitable for analyzing large amounts of data because it requires a huge amount of calculations.
In contrast, non-hierarchical cluster analysis such as the k-means method is suitable for analyzing large amounts of data because it does not require calculating the distance between all clusters and requires fewer calculations.
Five steps to perform cluster analysis
Now that you understand how to perform cluster analysis, we will explain the steps to perform cluster analysis.
Step 1: Set the purpose of your analysis
Step 2: Choose a cluster analysis method
Step 3: Define how to calculate similarity
Step 4: Decide how to form clusters
Set the analysis objective
The first step is to set the purpose of the analysis. In marketing, it is important to be clear about "why do we do analysis?"
You need a clear goal, such as understanding trends from customer purchasing data or reading latent preferences that cannot be determined from survey results alone.
Choosing a cluster analysis method
The next step is to decide on a cluster analysis method.
Specify either "hierarchical cluster analysis" or "non-hierarchical cluster analysis", as already explained.
Basically, when the amount of data analyzed is small, hierarchical cluster analysis is performed, and when the amount of data analyzed is large, non-hierarchical cluster analysis is performed.
Hierarchical cluster analysis produces easy-to-understand results, but care should be taken when using this method as accuracy decreases as the amount of data increases.
Define how similarity is calculated
The next step is to define what constitutes a "similar set" that will be the basis for clustering. In the case of four sets of data, "A, B, C, and D," when grouping similar sets, it is important to first consider "what is similar?"
The mathematical standard for determining "what is similar?" is "similarity" in cluster analysis, and this determines the "type of distance" between clusters.
The most common types of cluster distances are as follows:
Euclidean distance
Manhattan Distance
Minkowski distance
Chebyshev distance
Explaining how to calculate each one would take too long, so I will skip that, but the most orthodox method is the Euclidean distance, which uses straight-line distance.
In reality, it is necessary to select an appropriate distance for each case and the characteristics of the data.
Decide how to form clusters
The next step is to decide how to form the clusters.
When dividing the clusters, the results will vary depending on which method you choose: Ward's method, the group average method, or the k-means method, which have already been explained.
However, since marketing often involves analyzing huge amounts of data, the basic method is to use the "k-means" method of non-hierarchical cluster analysis.
In many cases, Ward's method is used only when the amount of data is small and the goal is to make the data visually easy to understand.