{"id":80080,"date":"2020-10-26T09:00:55","date_gmt":"2020-10-26T03:30:55","guid":{"rendered":"https:\/\/techvidvan.com\/tutorials\/?p=80080"},"modified":"2020-10-26T09:00:55","modified_gmt":"2020-10-26T03:30:55","slug":"machine-learning-k-means-clustering","status":"publish","type":"post","link":"https:\/\/techvidvan.com\/tutorials\/machine-learning-k-means-clustering\/","title":{"rendered":"K-Means Clustering in Machine Learning"},"content":{"rendered":"

In this article, we will specifically focus on one of the popular algorithms of clustering i.e., K-means clustering<\/strong>. This is one of the most used clustering algorithms.<\/p>\n

Here, we will mainly look at how the algorithm works, what are its benefits and shortcomings. We will discuss its various areas of usage. So, let\u2019s begin.<\/p>\n

What is K-means Clustering?<\/h3>\n

The algorithm that we will now dive into comes under unsupervised learning. Here, we deal with data that isn\u2019t labelled and unsupervised learning generally, uses input vectors<\/strong> to draw information from the datasets.<\/p>\n

Well, the premise of the k-means clustering is that it divides the dataset into similar and non-similar data and it clusters them.<\/p>\n

Let\u2019s say, if there are two different varieties of data in a dataset, then we will have two clusters. Here, \u2018k\u2019 is a fixed value. The number we equate \u2018k\u2019 with, is the number of clusters the algorithm will form. Also, in mathematics, there is a concept known as vector quantization<\/strong>.<\/p>\n

K-means clustering is a part of this concept. Also, the K-means algorithm is a part of the much bigger Expectation-Maximization algorithm.<\/p>\n

We have another technique known as the elbow point technique used for determining the value of k. We will study about it in this article.<\/p>\n

Note:- The second meaning of K is that it is a hyper-parameter. A hyper-parameter is a variable whose value we set before the training phase of the algorithm.<\/strong><\/p>\n

Well, it\u2019s always better to understand with a real-life example.<\/p>\n

If we have a dataset of let\u2019s say, junk-food eaters and healthy food eaters, the value of k here will be 2. Since we have two clusters that can form over here. We have shown it graphically in x-y axes terms as well. This way we can also find categories of data like only junk-food eaters, only healthy-food eaters.<\/p>\n

Also, varying categories like junk-food eaters who also consume healthy food and vice versa. The numbers can vary, but that\u2019s exactly the aim here.<\/p>\n

To understand various hidden patterns in the data. If we can picture these:<\/p>\n

\"k-means<\/a><\/p>\n

This plot shows the distribution of the data. This has numerous combinations of data available in it. Now, let\u2019s see how the cluster forms.<\/p>\n

\"Cluster<\/a><\/p>\n

The cross-marks are the centroids<\/strong> of the clusters. The algorithm here identifies the centroids of the clusters. Centroids can be both imaginary and real if taken, but the main aim is to find the centroids and to cluster data around them.<\/p>\n

How does K-means clustering work?<\/h3>\n

Well, here let\u2019s try to understand how exactly this algorithm works. Let\u2019s see a basic explanation first and then dive into the deeper concepts.<\/p>\n