Overview of different types of machine learning models
Machine learning is a branch of artificial intelligence in which we build methods and techniques by using statistical models and algorithms so that machines learn about particular data domains. We use machine learning to retrieve insights from patterns in data. In this article, we will explore mainly three types of machine learning models or techniques, which are categorized based on the problems they solve.
Contents
Types of machine learning models
- Supervised Machine Learning
- Unsupervised Machine Learning
- Reinforced Machine Learning
Supervised Machine Learning model
In supervised machine learning, we train the ML model by providing inputs(which are independent variables) and outputs(which are dependent on the inputs). For instance, let’s say we have a set of data on book pricing; the data has three fields as shown below and let’s say we want to predict the price.
| No. of pages | Binding | Price |
|---|---|---|
| 600 | Paperback | 20 |
| 1000 | Paperback | 22 |
| 1050 | Hardcover | 40 |
The dataset has three features namely “no. of pages”, binding, and price. In supervised learning, we divide these features into two categories. The first category is an independent feature, and the second one is a dependent feature. In the above dataset, “no. of pages” and “binding” are independent features or input features, and the “price” is a dependent feature or output feature that depends on the other two features. As the input values change, the price also changes.
If we create and train a machine-learning model on this dataset with a significant number of rows, the model will be able to predict the price of unknown input values.
Types of supervised machine learning
Supervised machine learning can be applied to mainly two types of problems solving models. The first one is regression, and the other one is classification.
Regression model
If the output feature(or dependent feature) has continuous values i.e. range of numerical values, then the problem statement fits the regression model. The example shown in the above table fits this model.
Classification model
If the output feature has categorical values, then the problem fits the classification model. The following dataset on loan approval fits the classification model because the values in the output feature(loan approved) fall under two categories.
| Study time | Loan approved | |
|---|---|---|
| 25 | 45000 | 1 |
| 30 | 60000 | 1 |
| 45 | 80000 | 0 |
| 40 | 60000 | 0 |
Note: If the output feature(dependent feature) in the classification model has two categories, the classification is called binary classification, and if there are more than two categories, the model is called multiclass classification.
Unsupervised Machine Learning model
Contrary to supervised machine learning, in unsupervised learning, we do not know the dependent/output feature, which means we do not predict the output. Rather, we try to find out the clusters of similar data based on the input features. For instance, let’s say we have a dataset of customers’ purchasing as shown below; the dataset has only the independent fields/features. In such a dataset, we try to create groups of similar customers based on the values of the features.
| Customer ID | Purchase Frequency | Time since the last purchase |
|---|---|---|
| 1 | 2 | 15 |
| 2 | 6 | 30 |
| 3 | 3 | 22 |
| 4 | 1 | 10 |
In the following diagram, a range of 100 data points has been grouped into three clusters represented by red, blue, and green colors. The centers of each cluster have been circled. You can visualize the clustering by going to this website.
Reinforcement Machine Learning Model
Reinforced learning is a completely different type of machine learning model. In this type, the machine learning model learns something itself by getting rewards for favorable behaviors and punishments for unfavorable behaviors. This model has an agent and an environment. The agent takes action on the environment from one state and goes to a new state based on the reward or feedback given by the environment. Thus, the model learns by trial and error.
This model is based on human and animal behavior. For instance, we train dogs by giving them treats when they follow our commands, and not giving any treats when they disregard us.
Conclusion
In this article, we explored various types of machine learning models, learned about when they are used, and understood the differences between them.
There are three main models of machine learning. Supervised learning has input/dependent features and an output/independent feature. On the other hand, unsupervised learning has only independent features and our goal is to group the dataset into clusters. Finally, reinforcement learning implements the reward and punishment to train an agent.
In the next series of articles, we will discuss various algorithms in each model of machine learning.
If you have any questions or suggestions, post them in the comment section below.
