Understanding Machine Learning Algorithms

If you want to get started with Data Science or Machine Learning, you need to have a basic understanding of all the algorithms which are commonly used. To become a data scientist, it is not the right choice to start coding immediately. 

In this blog, you will learn everything about the basic and most common algorithms used in Machine Learning models. This blog is a theoretical blog which will help you to understand which algorithms are suitable for what kind of problems. Stay tuned for the practical knowledge of these algorithms.

Machine Learning is basically divided into SUPERVISED and UNSUPERVISED learning. Let's learn more about them.

RELATED POST: LOAN PREDICTION PROBLEM PROJECT

SUPERVISED learning: It is a type of learning algorithm in which we teach the machine by giving the correct answers. There is a function that maps input variables (X) to an output variable (Y) based on examples present on the dataset. In easy terms, the dataset consists of inputs as well as outputs that we want to predict. This helps the machine to develop a function that maps input to the output correctly. Input variables are independent variables that are used to predict the output value that is called the dependent variable. The algorithm learns on a labeled dataset with the answer key to evaluate its accuracy. After the modeling of the algorithm with the labeled dataset, we test the model by using it to predict the unlabeled dataset. 

SUPERVISED LEARNING IS LIKE LEARNING WITH A TEACHER.

                                                           Y = f(X)

Examples - 

a) Real-life example:

1. You are very much familiar with the fact that when you see a photo of a person for the first time you always ask your parents or someone else about their name and identity. What happens when you see the same person again in another photo but in the same or other dress and pose? Yes, you recognize them easily. That's supervised learning. 

2. Another real-life example is when you see an apple for the first time you don't know that it is known as an apple. But when somebody teaches you that it is known as apple, you learn about it by its features such as red color, round in shape, and depression on its top.

b) Pattern recognition, speech recognition, Bioinformatics, and a lot more.

Types of supervised learning algorithms - As supervised learning is the most common type of learning, there are plenty of algorithms that come handy when you work on machine learning models. This learning is broadly divided into two types:

1. CLASSIFICATION

It is a type of supervised learning which specifies the class to the dataset. Classification algorithms classify the dataset into different classes. If the target variable is divided into more than two classes then it is known as a multiclass classification problem. In classification type of problems, the output format is just like 0/1, Yes/No, True/False.

Example - Whether a bank should provide a loan to a company or not? The output (target) variable consists only of 'YES' or 'NO' answers. Yes means that yes the bank should give the loan and No means the bank shouldn't. There are many examples of classification problems in real life. One of them is whether it will rain today or not. 

Algorithms:

•  Logistic Regression Algorithm
•  Naive Bayes Algorithm
•  K-Nearest Neighbours Algorithm
•  Random Forest Algorithm
• Decision Trees Algorithm
• Support Vector Machines

2. REGRESSION

In this type of supervised learning, the output (target) value is continuous. Regression does not divide the values into classes instead, it provides continuous numerical values bases on independent variables. Basically, regression is used where our target value is numerical (in very easy terms) and not in classes like 0/1.

Example - To predict the price of houses based on data such as house area, number of rooms, etc.

Algorithms:

• Simple Linear Regression Algorithm
• Multiple Regression 
• Polynomial Regression
• Logistic Regression
• -Ridge Regression
• LASSO Regression

UNSUPERVISED learning:  It is just the opposite of supervised learning. It consists of an unlabeled dataset. The algorithm tries to learn and extract features on its own. The dataset does not consist of answer keys or we can say it does not have output values so that it can easily learn based on the provided data. This algorithm needs more computing because we use it to predict values by extracting features on its own. This learning is best suitable when one wants to find patterns but don't know exactly what he/she is looking for. 

UNSUPERVISED LEARNING IS LIKE LEARNING WITHOUT A TEACHER (SELF STUDY).

Examples - 

Let's assume that you are a Data Scientist in a company. You have thousands of data fields based on your customer behavior and shopping data. Now the company wants you to draw patterns and gain insights from that data. Here, you only know that the company needs patterns and insights but they haven't told you what do they want. The unsupervised learning comes into action where you will group (cluster) the people based on the data to find relevant insights.

Types of Unsupervised learning algorithms - This learning is broadly divided into two types:

1. CLUSTERING

This type of unsupervised learning groups the data points. We can classify each data point by assigning it to a certain group. It divides the data points into a number of groups such that the data points in the same group are similar to other data points in the same group and different to data points in other groups. There are different methods by which you can cluster your data.

• Centroid based clustering
• Density-based clustering
• Distribution based clustering
• Hierarchical clustering

Example - Market segmentation to characterize relevant customer base and audience.

Algorithms: The most common clustering algorithms are-

• K-Means (Most widely used)
• DBSCAN (Density-Based Spatial Clustering of Applications with Noise)

2. ASSOCIATION

In this type of unsupervised learning, the relationships are manually represented in form of rules or frequent itemsets. These rules are used to create interesting insights from data. In easy terms, association rules define, 'If X buys an item 'A' then how likely Y is going to buy the same item 'A'.

Example - Determine which kind of people tend to visit together in a market for example a kid mostly visits with his/her mother.

Algorithms: 

• Apriori

This was the very basic introduction of machine learning algorithms which I tried to explain in my own words. Further, I will be writing about practical approaches and posting some projects so that we can learn to code together. 

I apologize if some any of my fact or explanation is wrong. Do provide us feedback, it will help us to grow together and become Atmanirbhar.

Depending upon the interest of the viewers, we can further publish blogs regarding various platforms, websites, and technologies for you guys to land your dream job, help us know your demands by filling the Google form and join this Telegram channel for more updates.

Thank you for reading and stay tuned.