We got a taste of Machine Learning in my previous post. We also got to know about the different types of machine learning techniques. We discussed how some problems can be of supervised nature and how can we solve them using regression and classification algorithms. In this post, we’ll talk about the second type of machine learning problem solving technique, called unsupervised machine learning.
Unsupervised Machine Learning:
It is the type of machine learning where the main aim of our algorithm is to find a relevant pattern in our data set. The data that we have does not contain any labels, so building any model on the top of it won’t be a feasible thing to do. Hence, the task at hand is to try and identify any occurring pattern for the data to make more sense. Unsupervised cases can be observed in many daily life scenarios like grouping people with similar interests on social media, also, called as clustering.
I’ve tried to mention a few most common unsupervised learning algorithms.
- Clustering – As the name suggests, this technique involves clustering of data points based on the similarity with other data points. There are various algorithms that can be used to to this, like k-means clustering.
(source: imperva.com) Example of clustering
Machine Learning is nothing but the ability of a machine to learn without a need a to program it again and again for the same type of problem. The algorithms used in machine learning learn to identify patterns hidden in a data set. These algorithms are then trained using these data sets till the algorithm becomes an expert in identifying the hidden patterns. Once our algorithm reaches it’s peak accuracy, we test it by providing it with test data to check the accuracy of our algorithm on the data it has never seen before. There are mainly three types of machine learning –
- Supervised Machine Learning
- Unsupervised Machine Learning
- Reinforcement Learning
In this post I’ll give a hint about various supervised machine learning algorithms.
Supervised Machine Learning :
The type of machine learning in which the data that we use in our problem has labels on them. This makes the learning process easier for our algorithm as it only has to learn the pattern. Every observation in our data set has an already defined attribute and the only task is to find patterns in the data. This is the easiest type of problem in machine learning. There are many types of tasks that we can perform using supervised machine learning. The most commonly used are described below :
- Regression : Regression is used for predicting the output by learning the pattern from an already available data set. It can of two types – Linear regression and Multivariate regression. It learns to map the input to output through the training data by fitting a straight line onto the data points. The accuracy of the learning is measured using an error function or a cost function. Lesser the value of cost / error function, more accurate is our algorithm. We can use certain methods like gradient descent or normal equation to find the most accurate fit for our algorithm.
- Classification: As the name suggests, in this type of machine learning problem, the algorithm learns classification using the training data. These are generally of two types, binary or multi class. The most commonly used classification algorithms are Naive Bayes and Logistic Regression. Generally, the algorithms learn the probability about whether the observation belongs to certain class or not. A good example of classification is the task of classifying whether the mail in your inbox is a spam mail or not.
Hence, in this post, I’ve introduced you to the basic terms of supervised machine learning algorithms. In the next post, I’ll discuss more about the unsupervised machine learning algorithms.