Machine Learning

We can break down ML into four parts:

• Supervised machine learning
• Unsupervised machine learning
• Semi-supervised machine learning
• Reinforcement machine learning

Supervised Machine Learning

In this type of learning, the model is trained on the data for which we already have the correct labels or output. In short, we try to map the relationship between input data and output data in such a way that it can generalize well on unseen data as well. Supervised machine learning is the major category of machine learning that drives a lot of applications and value for businesses.
The training of the model takes place by comparing the actual output with the predicted output and then optimizing the function to reduce the total error between the actual and predicted.
This type of learning is predominantly used in cases where historical data is available and predictions need to be made on future data. The further categorization of supervised learning is based on types of labels being used for prediction.
If the nature of the output variable is numerical, it falls under regression, whereas if it is categorical, it is in the classification category.
Classification refers to the case when the output variable is a discrete value or categorical in nature. Classification comes in two types.

• Binary classification - Exactly two categories
• Multiclassification - More than two categories

Another property of supervised learning is that the model’s
performance can be evaluated.

Unsupervised Learning
Unsupervised Learning is different from supervised learning in terms of the output labels. In unsupervised learning, we build the models on similar sort of data as of supervised learning except for the fact that this dataset does
not contain any label or outcomes column. Essentially, we apply the model on the data without any right answers. In unsupervised learning, the machine tries to find hidden patterns and useful signals in the data that can be later used for other applications. The main objective is to probe the data and come up with hidden patterns and a similarity structure within the dataset.
One of the use cases is to find patterns within the customer data and group the customers into different clusters. It can also identify those attributes that distinguish between any two groups. From a validation perspective, there is no measure of accuracy for unsupervised learning. The clustering done by person A can be totally different from that of person B based on the parameters used to build the model.

There are different types of unsupervised learning.

• K-means clustering
• Mapping of nearest neighbor