# 10 Common Machine Learning Algorithms that You Must Know

With more and more data, machine learning is becoming incredibly powerful to make more accurate predictions or personalized suggestions. However, there is no one machine learning algorithm that works best for every problem; especially, if it’s for supervised learning (i.e. predictive modeling). In this post, we will go through the basic statistical models and most common machine learning algorithms that you must know as a beginner. For the absolute beginners, please refer to this introductory article on machine learning and artificial intelligence.

# Machine Learning Algorithms for Beginners

Co-authored by Pavan Somwanshi

## What are Machine Learning Algorithms?

Machine Learning algorithms are the brains behind any model, allowing machines to learn, making them smarter.

The way these algorithms work is, they’re provided with an initial batch of data, and with time, as algorithms develop their accuracy, additional data is introduced into the mix.

This process of regularly exposing the algorithm to new data and experience improves the overall efficiency of the machine.

ML algorithms are vital for a variety of tasks related to classification, predictive modeling, and analysis of data.

### Related Articles:

Top Countries with Growing Shortage of AI, Machine Learning, and Deep Learning Talent

Best Universities for Masters in Machine Learning and AI in UK

Masters in Machine Learning / Artificial Intelligence for Non-CS Graduate in Mid-30s as Reapplicant

MS Data Science (with Supervised / Reinforcement Learning Specialization) in USA with Scholarships

## 4 Broad Types of Machine Learning Algorithms based on Learning Style

### Supervised Learning

• The word “supervise” literally means to observe and direct the execution of a task. In the context of Machine Learning, a data scientist supervises a model that has been assigned to perform a particular task or a project.
• This algorithm consists of a target/outcome variable (or dependent variable) which is to be predicted from a given set of predictors (independent variables). Using these set of variables, we generate a function that map inputs to desired outputs.
• The initial process here is to train the model with a labeled dataset, that acts as sample data which the model can use to make predictions about possible outcomes.
• This type of machine learning is ideal for Binary classification, Multi-class Classification, Regression Modelling, and Ensembling.
• Examples of Supervised Learning include Regression, Decision Tree, Random Forest, KNN, Logistic Regression, etc.

### Unsupervised Learning

• Just like supervised learning the workings of unsupervised learning are too literal to its meaning. In tasks involving unsupervised learning, we don’t supervise the models but let them perform a function on their own.
• Unsupervised models make predictions based on unlabeled data. It works on the unlabeled data to find patterns it can use to further group the data.
• This makes the algorithms in Unsupervised learning more complex as there is very little information to work with & the expected outcomes are uncertain.
• In this algorithm, we do not have any target or outcome variable to predict / estimate. It is used for clustering populations in different groups, which is widely used for segmenting customers in different groups for specific intervention.
• Unsupervised learning is best for Clustering, Anomaly detection, Association mining & density reduction.
• Examples of Unsupervised Learning include Apriori algorithm, K-means.

### Semi-Supervised Learning

• Semi-supervised learning is, for the most part, just what it sounds like: a training dataset with both labeled and unlabeled data.
• This method is particularly useful when extracting relevant features from the data is difficult, and labeling examples is a time-intensive task for experts.
• Example problems are classification and regression.
• Example algorithms are extensions to other flexible methods that make assumptions about how to model the unlabeled data.

### Reinforcement Learning

• Reinforcement Learning refers to a kind of Machine Learning method in which the agent receives a delayed reward in the next time step to evaluate its previous action.
• Using this algorithm, the machine is trained to make specific decisions. It works this way: the machine is exposed to an environment where it trains itself continually using trial and error.
• The machine learns from past experience and tries to capture the best possible knowledge to make accurate business decisions.
• Example of Reinforcement Learning is Markov Decision Process.

## Common Machine Learning Algorithms based on Similarity

### Linear Regression

In linear regression, we establish a relationship between independent and dependent variables by fitting the best line. This best fit line is known as regression line and is represented by a linear equation Y= a *X + b.

The best way to understand linear regression is to relive this experience of childhood. Let us say, you ask a child in fifth grade to arrange people in his class by increasing order of weight, without asking them their weights! What do you think the child will do?

S/he would likely look (visually analyze) at the height and build of people and arrange them using a combination of these visible parameters. This is linear regression in real life!

### Logistic Regression

The technique of Logistic Regression, borrowed from statistics, is primarily used as a method for binary classification problems. Binary classification for those of you who don’t know is where our output predictions can only take one of the two possible values Eg. True or False, Yes or No.

A real-world example of this is when an algorithm decides to classify a mail as either spam or not spam.

Logistic regression algorithms help us predict the probability of occurrence of an event by fitting the data into a Logistic curve (common S-shaped curve with equation). I suppose this helps explain why the outcome is binary, but that’s not the only type of Logistic regression.

Two other types of Logistic Regressions (LR) are, Multinomial LR which has 3 or more possible outcomes with no order, and Ordinal LR with the same outcomes but with a natural ordering.

### Decision Tree

Decision Tree (also known as Classification & Regression Tree) is a very important and one of the most popular algorithms in Machine learning. A decision tree is a graphical representation that displays all possible outcomes of a decision. In the graphical representation, each fork represents a test and each branch represents an outcome. Both of these are derived by taking the attributes before them into consideration.

There are different types of decision trees:

1. Classification trees are used when the response variable is categorical in nature.
2. Regression trees are used when the response variable is continuous or numerical in nature.

The visual representation aids for better representation. It works well classifying for both categorical and continuous dependent variables. The algorithm barely makes errors and the knowledge of possible outcomes under different decisions helps the data scientist. This is why the use of the Decision Tree algorithm is favored.

Not just in Machine learning, Decision trees are frequently used in various other domains as they are easy to learn and practice.

It works well classifying for both categorical and continuous dependent variables.

### Random Forest

The Random Forest algorithm is one of the most popular machine learning algorithms. It is an ensemble model that uses bagging as an ensemble method and is primarily used for solving regression and classification problems. For those unfamiliar, Bagging is when you train a bunch of individual models simultaneously.

With this algorithm, you create multiple subsets of the entire dataset, use these subsets to train an equal number of decision trees such that one subset trains one tree. Further, each decision tree makes a prediction and the prediction with maximum votes is considered as the final prediction.

A few advantages of using this algorithm are, it’s highly accurate, can be implemented with just a few lines of code making it easy to use & the algorithm runs effectively on large databases.

Each tree is planted & grown as follows:

1. If the number of cases in the training set is N, then sample of N cases is taken at random but with replacement. This sample will be the training set for growing the tree.
2. If there are M input variables, a number m<<M is specified such that at each node, m variables are selected at random out of the M and the best split on these m is used to split the node. The value of m is held constant during the forest growing.
3. Each tree is grown to the largest extent possible. There is no pruning.

Recommended Article: Best Online Courses on Machine Learning and AI

### Support Vector Machine (SVM)

SVM is a supervised ML algorithm used for both classification and regression problems. Basically, we plot each data item as a point in n-dimensional space where n is the number of features present. Here, the value of each feature is the value of a particular coordinate.

In this algorithm, we plot each data item as a point in n-dimensional space (where n is the number of features you have) with the value of each feature being the value of a particular coordinate.

For example, if we only had two features like the Height and Hair length of an individual, we’d first plot these two variables in two-dimensional space where each point has two coordinates (these coordinates are known as Support Vectors).

Classification of genes is a prominent real-life usage example of this algorithm.

### Naive Bayes

Based on the Bayes theorem, Naive Bayes operates assuming the presence of a particular feature in a class to be unrelated with the presence of any other feature.

The way it operates is by calculating the probability of each class and conditional probability for each class given a certain value, which it later uses to make predictions for new data.

In simple terms, a Naive Bayes classifier assumes that the presence of a particular feature in a class is unrelated to the presence of any other feature. For example, a fruit may be considered to be an apple if it is red, round, and about 3 inches in diameter.

This model is highly useful and quite easy to build. Naive Bayes assumes that each input variable is independent, and that’s where the name ‘naive’ comes from.

Gmail uses this algorithm to determine whether a particular email is a spam or not.

### k- Nearest Neighbors (kNN)

This algorithm is a super simple one that can be used to classify data. For example, if you want to determine what kind of cell “cell D” is, you can use the already present data about numerous cells and accordingly come to a conclusion.

The way you do this is, first you start with a dataset of known categories(of cells). You then add the new uncategorized cell on the plot that maps all the other cells. Further, you classify the new cell based on the “nearest neighbors” and come to a conclusion.

K Nearest Neighbors uses distance-based measurements to get the most accurate prediction. Euclidean, Manhattan, Minkowski, and Hamming distance are the various distances that can be used. Except for the last function mentioned(which is used for categorical variables), all the functions are continuous functions.

This algorithm is really useful for non-linear data and delivers high accuracy but the only drawback it has is that it demands high computational memory storage.

It can be used for both classification and regression problems. However, it is more widely used in classification problems in the industry.

### K-Means

It is a type of unsupervised algorithm which solves the clustering problem. Its procedure follows a simple and easy way to classify a given data set through a certain number of clusters (assume k clusters). Data points inside a cluster are homogeneous and heterogeneous to peer groups.

### Neural Network

Neural Networks are of core importance in the fields of Machine Learning, Deep Learning, and Artificial Intelligence.

Just like neurons in our brains signal each other, Neural networks operate in a similar fashion. Going through a ton of data over a period of time for training purposes, neural networks fine-tune themselves to offer precise accuracy.

Once this gets done, data scientists make use of these algorithms which allows them to effectively classify huge amounts of data with high accuracy.

Boltzmann machine, Recurrent neural networks, Multilayer Perceptrons & Deep Belief Network are some widely used deep learning models based on neural networks.

Google search engine is a popular example that makes intensive use of Neural Networks. The use of this algorithm is not limited to data science, as chess players frequently use chess engines to analyze their games and these chess engines deploy neural networks in order to learn.

As the algorithm learns and recognizes new techniques and patterns, knowledge of this helps the chess player to improve their performance.

### Gradient Boosting Machine Learning Algorithms

Gradient Boosting Algorithm is a boosting algorithm that is used when we deal with huge amounts of data. Boosting, according to Wikipedia, is “an ensemble meta-algorithm for primarily reducing bias, and also variance in supervised learning, and a family of machine learning algorithms that convert weak learners to strong ones”.

There are four popular machine learning algorithms under the Gradient Boosting Algorithms category:

1. GBM
2. XGBoost
3. LightGBM
4. CatBoost

#### XGBoost

XGBoost is the most common and effective one from the above list. This algorithm is preferred because it can accelerate other algorithms by multiple folds and with higher accuracy than any other GBR method. It supports multiple objective functions such as regression, classification, and ranking.

The XGBoost has an immensely high predictive power, which makes it the best choice for accuracy in events. It possesses both a linear model and the tree learning algorithm, making the algorithm almost 10x faster than existing gradient booster techniques.

The support includes various objective functions, including regression, classification, and ranking.

One of the most interesting things about the XGBoost is that it is also called a regularized boosting technique. This helps to reduce overfit modeling and has massive support for a range of languages such as Scala, Java, R, Python, Julia, and C++.