Unsupervised learning algorithm

In unsupervised learning algorithm we don't supervise the model, but we let the model work on its own to discover information that may not br visible to the human eyes. It means, the unsupervised algorithm train on the dataset and draws conclusion on unlabelled data.

Fig 1. Supervised VS Unsupervised


Generally speaking, unsupervised algorithm has more difficult algorithms than supervised algorithm since  we know little to no information about the data, or the outcomes that are to be expected.

This algorithm further grouped into Clustering and Association problems. As shown in below.

Fig 2. Types of algorithm

Mainly we'll focus about clustering algorithm.
Now let's see more about Clustering.

Clustering:
    
Fig 3.This is how cluster looks like
  • It is an important concept when it comes to unsupervised learning algorithm. It mainly deals with finding a structure or pattern in a collection of uncategorised data.
  • Clustering algorithm will process your data and find natural cluster(group) if they exist in the data.
  • You can also modify how many cluster your algorithm identify. It allows you to adjust the granularity of these group.
Clustering VS Classification: 

Fig 4. Classification example
 We'll understand this with the help of following example.
This is a classification where we have labeled dataset. After training data we build a model and predict value(result)
 
Fig 5. Clustering example
In clustering we have given a unlabelled dataset and we didn't predict any value rather than we need to construct structure or pattern(group) of similar data.

Why clustering?
  • Exploratory data analysis
  • Summary generation
  • Outlier detection
  • Finding duplicates
  • Pre-processing step
Types of clustering algorithm
  1. Partition based clustering
    • Relatively efficient
    • E.g K-means, K-median, Fuzzy C-mean
  2. Hierarchical clustering
    • Produces tree of cluster
    • E.g Agglomerative, Divisive
  3. Density based clustering
    • Produces arbitrary shaped cluster
    • E.g DBSCAN 
For any machine learning algorithm(implementation code) visit this link: https://github.com/vishalbarad/Machine-learning

Comments