[Roughly] Clustering by KMeans

Overview

• Purpose: Cluster customers from purchase history, etc. to understand their characteristics
• Method: Clustering with KMeans
• Output: Number of people in each cluster, characteristics of each customer cluster
• Area of use: Marketing

approach

1. Store data in dataframe
2. Determine the number of clusters
3. Clustering with KMeans
4. Output the clustering result

code

import pandas as pd
import numpy as np
from sklearn.cluster import KMeans
from collections import Counter

##Prepare dataframe df###

num_clus = 4 #Set the number of clusters
kmeans = KMeans(n_clusters=num_clus, random_state=0).fit(df)

print(Counter(kmeans.labels_)) #Output the number of people in each cluster

df['cluster_id']=kmeans.labels_ #Add cluster number to original dataframe

for i in range(0,num_clus): #Output the average value of each cluster
    print(df[df['cluster_id']==i].mean())