The 3 most important forms of machine learning which are of business use or practical use are

• Supervised learning
• Unsupervised learning
• Reinforcement learning
• 
  

Supervised learning is the most mature and easiest form of Machine Learning. The models or machines are trained by feeding datasets. After the training, the given model or machine can start predicting when new data is given to it.

In unsupervised learning, the model learns through observation and finds structures in the data. It finds patterns and relationships in the dataset by creating clusters in it. However, it cannot label all the clusters after separating them into different groups.

Reinforcement learning is the ability of an agent to interact with the environment and find out what is the best outcome. It uses the hit and trial method. During the process, the model is rewarded or penalized with a point for a correct or a wrong answer. The model itself trains on the basis of the positive reward points gained. After the training, the model gets ready to predict the new data presented to it.