Supervised Learning is a type of machine learning algorithm used to identify patterns in data sets by analyzing data labeled with pre-determined variables. It is a form of predictive analytics that poses a question and then trains the data set to correctly solve it. Supervised learning is used to extract insight and information from data sets that would otherwise have been overlooked, allowing an organization to uncover trends and patterns that can enable better decision-making.
Overview
Supervised learning is a subcategory of machine learning algorithms used to forecast and classify data. Supervised learning is different from unsupervised learning, which identifies patterns in data sets without predetermined criteria or labels. Through supervised learning, algorithm tasks can be set up to learn from data sets by comparing outputs with already existing data points, all with the goal of solving a posed problem. Problems posed in supervised learning are referred to as tasks. For example, a supervised learning algorithm could be set up to help identify if an email is spam or not.
Supervised learning models are effectively used to identify whether an input into a machine learning algorithm is in a specific class or not. These types of algorithms are used extensively in fields such as finance, retail, and law to identify data that is not easily defined by rules and criteria established by humans.
Uses
Supervised learning algorithms are used to apply a particular function to recognize data points when no such data exists or to identify patterns that may not be seen by humans. These types of algorithms are particularly useful when a data set contains a large number of input variables or when the degree of certainty of the output needs to be high.
Supervised learning algorithms are also used when data does not fit into an already-defined schema because the amount of variance is beyond the capacity of humans. This type of learning algorithm is used to identify outliers and anomalies, as well as to identify correlations between sections of data where no human-defined causal explanation is known.
Supervised learning is also used to identify and predict fraud in e-commerce transactions. In this type of task, the algorithm is designed to distinguish between valid and fraudulent payments by comparing them with previously identified cases of fraud.
Key Features
• Supervised learning requires no predetermined criteria or labels
• It is used to solve a posed problem
• Good for large data sets with many input variables
• Useful to identify outliers or anomalies
• Can be used to identify and predict fraud
Example of Use
The banking industry uses supervised learning algorithms extensively to help prevent fraud. Banks use deep learning algorithms that independently identify and categorize patterns in hundreds of variables at once, including card type, card number, timestamp and location for every transaction. The algorithm is able to compare a transaction against known cases of fraud and alert the customer or the bank if a fraudulent transaction is suspected. Supervised learning algorithms are also used to detect suspicious accounts and flag them for further investigation. By using supervised learning algorithms, banks are more able to identify fraud before it occurs, thereby minimizing financial losses and reputational damage.
« Back to Glossary Index