Page 177 - AI Computer 10

CATEGORIES OF MACHINE LEARNING BASED MODELS

On the basis of nature of input, machine learning based model can further be categorised into three types:
u Supervised Learning u Unsupervised Learning u Reinforcement Learning

Machine Learning

Supervised Unsupervised Reinforcement

Supervised Learning

Supervised Learning is a category of machine learning that uses Letter grade Percentage
labelled datasets to train algorithms to recognise patterns in the
input data and predict outcomes. Datasets play an important A 80 – 1100
role in AI models. A model is said to be supervised if it learns
from a labelled dataset. A label is basically used to categorise the B 65 – 79
various items in the dataset into distinct groups. In Supervised
learning, a human expert creates the training dataset by C 55 – 64
assigning labels to the collected data. Thus, the leaning is called D 50 – 54
Supervised.
Let us understand the concept of labelling with the help of some E 0 – 49
simple examples. In your class, you have seen that teachers use
a grading system for the marks/percentage secured in examinations. These grades are labels that categorise the
students according to their marks/percentage. The model accepts the marks of a students as input and predicts
the grade based on the training dataset.
In the banking sector, loan approval system is a great example of supervised learning where banks can train a
model using past loan applicants' data including various data items such as income, credit score, and proposed
loan amount, along with labels whether the loan was approved or rejected. On the basis of the data, the model
can assess new applicants to classify whether the application is to be approved or rejected.

Let us learn the concept of supervised learning in chatbots. The general steps used to build chatbots are:
Step 1: Identify the specific tasks the chatbot will handle.

Step 2: Gather conversational data relevant to the defined purpose. The data should ideally be in a structured
format, with user queries and appropriate responses clearly mapped out.

43
43