A glimpse of Machine Learning

Machine learning is a process of imitating humans with the help of algorithms. In simple terms, it involves copying the learning process used by humans to learn a task.

The algorithm used for the above process is known as the machine learning algorithm.

Let’s consider a human trying to learn a language; initially, he will try to understand the basics, then with repetitive practice, he will master the task of learning. Likewise, a machine learning algorithm will learn the task by copying the learning procedure used by humans.

Machine learning algorithms are mainly divided into four parts:

Fig 1.0 - Classification of machine learning algorithm

To understand in more detail let's first try to understand the excel sheet below.

Photo 1: Input and Output

As one can see the sheet is divided into 2 parts i.e. Input and Output.

   Input is a factor that may or may not affect the output. In the above example age, education, capital, hours per week are the factors that may affect an individual's income. 

Generally in machine learning, we provide data (including input and output), a machine learns from the examples, and finally produces output for the unknown data. Different types of machine learning algorithms are as follows:

Supervised Learning: In this type of learning, input and output are provided to the model to predict the output of unknown data. Supervised learning consists of both input(X) and output(Y)

Unsupervised Learning: Unsupervised learning consists of only input(X) and no output variable (Y), here machine learns from grouping the data.

Semi-supervised Learning: In this type of learning, some part of input (X) data may have the output(Y)

Deep Learning: This is a subset of machine learning where data is passed through several layers of neural networks. For example, a dataset of images of cats and dogs (mixed) is provided as input and the machine accurately classifies the image as cat or dog.

 A dataset in supervised learning can be classified into two types i.e. regression type and classification type

   A model with continuous values in output falls into regression category whereas data with discrete labels in output is categorized as classification type. In the above photo 1 (Input vs. output), the output in the dataset is classified into income more than 50K and less than 50k hence it is a classification type dataset. The price range of houses is a great example of a regression-type dataset.

Different types of classification and regression algorithms:

Classification algorithms:

• Naive Bayes 
• Logistic regression
• Decision Tree  
• Random Forest
• K-Nearest Neighbours
• SVMs

Regression algorithms: 

• Simple Linear Regression
• Multiple Linear Regression
• Logistic Regression
• Ridge Regression
• Lasso Regression

**Continuous data – data that changes and is not fixed. Example- Bodyweight

**Discrete data – data that is fixed. It can also be categorical. Example- male and female

Summary: 

Machine learning involves obtaining maximum accuracy from the data provided. 

As humans learn from practice - to master any art, likewise machines also learn from experience/model. Algorithms develop several models and the model with maximum accuracy is considered as the best model.

Link - https://analyticaluniverse.blogspot.com/2021/06/a-glimpse-of-machine-learning.html