Machine learning, a branch of artificial intelligence (AI), is increasingly used in various industries. No wonder, this technology can carry out various automation to accelerate business processes and optimal service to consumers. In machine learning, some components allow computers to learn data patterns and then make predictions or decisions without user guidance. One of the components is an algorithm.

A machine learning algorithm is a set of mathematical instructions or procedures used to develop machine learning models. The better the algorithm used, the better the predictions and decisions made by machine learning. Like humans, the more and better information and knowledge they get, the smarter they will be. Therefore, the selection of machine learning algorithms must be according to their needs.

Based on its use, in general, machine learning algorithms are divided into 3 categories, namely supervised learning, unsupervised learning, and reinforcement learning. Here is the discussion.

#1 Supervised Learning

Supervised learning is a type of machine learning algorithm that uses labeled data. Like humans, when toddlers are introduced to various fruits with their names, for example, this is a banana, this is an apple, this is an orange, and so on. Thus, toddlers will be able to classify which ones are bananas, oranges, and apples. Likewise, machine learning with supervised learning algorithms can carry out classification, regression, and anomaly detection tasks.

Machine learning is used to predict patterns. The pattern already has complete sample data. Thus, the pattern formed is the result of studying the complete data.

If we enter new data, after extracting transform load (ETL) we will get feature info from the new sample. Then, these features are compared with the pattern classification of the model obtained from the labeled data. After the process of comparing all labels is complete, the label that has the most percentage will be taken as the final prediction.

This supervised learning algorithm is divided into several categories based on the purpose of its formation, classification and numerical regression/prediction. For classification purposes, the algorithm consists of neural networks, random forest, KNN, SVM, decision trees, random forest, etc. Meanwhile, for the regression purposes it consists of linear regression, neural networks, decision trees, SVM, etc.

For example, a company wants to know whether a customer will buy a product or not. The data owned is purchase history, activity history on the website, gender, age, and the amount of customer income. This data is labeled “yes” or “no” based on whether or not the customer purchased the product. With the available data the supervised learning algorithm creates a model to predict whether the customer will buy the product or not.

#2 Unsupervised learning

The second type of machine learning algorithm is unsupervised learning where the data used has no labels. Without true and false labels or outputs, this type of algorithm has no supervisor to help determine right or wrong. The purpose of the unsupervised learning algorithm is to find patterns or groups in data such as clustering and dimension reduction.

For example, unsupervised learning algorithms are used to determine market segmentation. A company has data in the form of gender, occupation, age, address, frequency of purchases, and quantity of product purchases for the last 1 year. Companies do not need to provide correct labels or outputs or label customers with age, gender, or other labels. We also do not need to determine the number of groups and criteria for each group. The unsupervised learning algorithm will learn the patterns from the characteristics of each data and then group them by itself.

Based on its purpose, unsupervised learning algorithms are divided into clustering and association. For example, clustering consists of K-Means clustering and Hierarchical clustering. For associations purpose is Association rules.

#3 Reinforcement Learning

In reinforcement learning, no data set is given as in supervised learning and unsupervised learning. The model learns through interaction with the environment. There are 2 components in this algorithm, “agent” and “environment”. Agents learn independently how to interact with the environment to achieve goals.

The goal of the reinforcement learning algorithm is to maximize the reward from the environment. Examples of its use are in games and robotics. For example, a company wants to create a robot that can walk and avoid obstacles that it encounters. So this algorithm will create a model that can maximize the reward when the robot manages to pass through obstacles without colliding.

Another example is in a chess game. Machine learning is the agent while the opponent (game user) is the environment. The machine will learn by itself how to win matches based on the experience he gets. For example, the experience when he managed to eat the opponent’s fortress or get a check. From the experience gained, the machine learns strategic patterns (dos and don’ts) to win.

Those are the types of machine learning algorithms. Today machine learning skills are much needed by various industries, including the financial industry. More and more are interested in mastering machine learning to compete for jobs with high salaries. This skill will add value to your branding to get a job with a more satisfying salary value.

Well, it’s not too late. You can learn about machine learning in Katalis.App right now. Check the complete info here.