Machine learning world easy-to-understand overview for beginners
A complete explanation of the most widely practical and efficient field, that nowadays has an impact on every industry
Machine learning has become one of the most rapidly evolving and popular fields of technology in recent years.
Basically, Machine learning is a part of the Artificial intelligence field, which is mainly defined as a technic that gives the possibility to predict the future based on a massive amount of past known or unknown data. ML algorithms can be broadly divided into supervised learning, unsupervised learning, and reinforcement learning. In this article, I will cover all of them.
But I also want truly define that ML isn’t represent some kind of unsecured AI technologies, super brain or dark magic, it’s clear combination of programming skills, enough amount of data, cloud solutions, theory of algorithms and math — that’s all we should have to be able to work in this branch.
Are you interested in how Netflix suggests movies that are precisely desirable for you to see, or maybe how Amazon suggests the products which you probably want to buy with, or maybe how the company can predict approximate profit for the next couple of years, or how we can save humans lives applying ML on breast cancer prediction? Strictly, everything that I said earlier is based on Machine learning algorithms and, of course, strong math and theory of algorithms behind them.
How is it actually looks in a real life process of ML investigation?
During the machine learning tasks, you will work with data preprocessing(basically work with data: cleaning, observing, splitting), modelling(choosing the right model, building, training, prediction) and evaluation(tracking metrics, making final verdict, trying to find a useful application in a real word)
Supervised learning
Shortly, supervised learning deals with different regression models and classification models, when we have labeled data. It means that we can see clearly what we are going to predict.
Regression is a technique to predict the real continuous value. For example, financial and sales forecasting, marketing and healthcare field and such more.
- Simple linear regression
- Multiple linear regression
- Polynomial regression
- Decision Tree regression
- Support Vector regression
- Random Forest regression
Classification is a technique to predict a category. For instance, Email can be divided into normal email, import email, promotion email or spam — so hear we have 4 categories to work with.
- Logistic Regression
- K-Nearest Neighbors (K-NN)
- Support Vector Machine (SVM)
- Kernel SVM
- Naive Bayes
- Decision Tree Classification
- Random Forest Classification
I will not go too deep about these algorithms in this article, but it’s worth it for you to do it yourself. It’s a fantastic world, trust me! You can also look at my GitHub portfolio to see the actual applications of some of them.
Unsupervised learning
Unsupervised learning is applied with clustering models with unlabeled data, so our goal is to detect new features and patterns.
For example, we have a list of online e-commerce users with some information such as the amount of spending, amount of salary, and age, and our goal is to discover which group of people has more chances to continue spending money so that we can target marketing strategies directly on them.
The most common unsupervised algorithms are clustering, dimensionality reduction, and association rule mining.
Clustering is similar to classification, but the basis is different. In Clustering you don’t know what you are looking for, and you are trying to identify some segments or clusters in your data. When you use clustering algorithms on your dataset, unexpected things can suddenly pop up like structures, clusters and groupings you would have never thought of otherwise. Great example of this tecnique is K-means clustering algorithm.
Association rule mining involves identifying relationships between items in a dataset. This technique is commonly used in retail for market basket analysis, which involves identifying which products are frequently purchased together. The most commonly used algorithms are Apriori and Eclat.
Reinforcement learning
And the last one is reinforcement learning. With Reinforcement learning the model will be optimized for feedback to previous actions and experiences. Usually the optimization is based on rewards.
It is also used for Artificial Intelligence when training machines to perform tasks such as train the robot to walk in the right form. Reward(1) or punishment(0). In this technique algorithm tried huge amount of the random set of actions and discover which suits the best.Reinforcement learning is the closest part of Ml to Artificial intelligence field and probably the hardest one to understand.
The real-world examples of applications are game playing such as chess, Go, and Atari games, robotics, autonomous vehicles and recommender systems.
Overall, reinforcement learning is a powerful technique for training agents to make decisions in complex and dynamic environments.
