In Machine Learning we generally talk about three types of learning
- Supervised Learning
- Unsupervised Learning
- Reinforcement Learning
In this post We’ll be discussing Supervised learning.
Supervised learning is learning a model based on a set of labeled example. The technical description of a labeled example maybe expressed as a vector containing parameters paired with a desired output value. But a friendlier way of saying that is that supervised learning models are given example of what worked or what not.
So they have some experience and use that experience to make current and future decisions.
There are plenty if problems machine learning models have to deal with.
Spam detection in an email is a common example where input is an email message or email metadata and output is determination that an email is spam or not. Or the probability that an email is an spam so somewhere in between.
In Medical field input is patient’s medical information and output would be medical diagnoses.
Another example might be the prediction of a home’s selling price with input being house’s details- How big, how new, how many rooms, how many bathroom and so on… Output would be the selling price of the home.
Classification is used to assign a meaningful placement of an element. That is- a categorization. So given an input , an agent could assign a class to an element. The output is usually a categorical and numerical data can be transformed into categorical data for classification. To use a spam detection example, an agent could classify an email message to determine that it’s spam or not — so a binary classification.
Common models used for classification are decision tree, logistic regression and support vector machine.
Regression means that given an input , an estimation of a parameter is provided. It can be seen as the function of an input and output is typically numerical data. Some Regression Models map to categorical data, for example- Logistic Regression.
An example might be home price detection where the output is function of the input mapping of a number.
Common models used in regression are the Linear Regression model, Multiple regression model, Decision Tree Models.