Category Archive Machine Learning

BySlide Scope

K means clustering algorithm example using Python

K Means Clustering is an algorithm of Unsupervised Learning. You can apply this algorithm on datasets without labeled output data.Only Input data is there an we have a goal of finding regularities in data to group or cluster like items together.

You can copy the code an run it line by line in Jupyter Notebook.

Watch the videos given in the bottom of this post to understand the process clearly.

What is a Cluster – Datapoints aggregated together because of certain similarities


import numpy as np
import matplotlib.pyplot as plt
# Import the algorithm from scikitlearn https://scikit-learn.org
from sklearn.cluster import KMeans
# Get the dataset of wine https://archive.ics.uci.edu/ml/datasets/wine
names = ['Class', 'Alcohol', 'Malic acid', 'Ash', 'Alcalinity of ash', 'Magnesium', 'Total phenols', \
'Flavanoids', 'Nonflavanoid phenols', 'Proanthocyanins', 'Color intensity', 'Hue', 'OD280/OD315',\
'Proline'] data = pd.read_csv('https://archive.ics.uci.edu/ml/machine-learning-databases/wine/wine.data', names = names)
data.head(100)
data['Class'].value_counts().plot(kind='bar')
data.plot.scatter(x = 'Alcohol', y = 'OD280/OD315', figsize=(8,5))



data.plot.scatter(x = 'Alcohol', y = 'OD280/OD315', c= 'Class', figsize=(8,5), colormap='jet')
data.iloc[:,[12,1]].head()
# kmeans = Kmeans().fit(data)
# kmeans = KMeans(n_clusters = 2)
# kmeans.fit(X)
# kmeans.cluster_centers_
# kmeans.labels_
kmeans = KMeans(n_clusters=3, init = 'random', max_iter = 1, random_state = 5).fit(data.iloc[:,[12,1]])


centroids_df = pd.DataFrame(kmeans.cluster_centers_, columns = list(data.iloc[:,[12,1]].columns.values))
fig, ax = plt.subplots(1, 1)
data.plot.scatter(x = 'Alcohol', y = 'OD280/OD315', c= kmeans.labels_, figsize=(12,8), colormap='jet', ax=ax, mark_right=False)
centroids_df.plot.scatter(x = 'Alcohol', y = 'OD280/OD315', ax = ax, s = 80, mark_right=False)
kmeans = KMeans(n_clusters=3, init = 'random', max_iter = 150, random_state = 5).fit(data.iloc[:,[12,1]])
centroids_df = pd.DataFrame(kmeans.cluster_centers_, columns = list(data.iloc[:,[12,1]].columns.values))
fig, ax = plt.subplots(1, 1)
data.plot.scatter(x = 'Alcohol', y = 'OD280/OD315', c= kmeans.labels_, figsize=(12,8), colormap='jet', ax=ax, mark_right=False)
centroids_df.plot.scatter(x = 'Alcohol', y = 'OD280/OD315', ax = ax, s = 80, mark_right=False)

k means clustering with centroid
What is K means clustering ?
You can watch the theory here :

Applying K means clustering on wine dataset :