Using Built-In Supervised Learning Tasks

ML.NET has a wide range of supervised ML tasks available in its CLI and the NuGet libraries. These include the following:

• Classification: This is where the model is built to determine whether an entity belongs to a specific category based on its parameters. For example, we might want to build a model that identifies a type of car based on its dimensions, the number of seats, engine power, etc. ML.NET supports binary classification, multi-class classification, and image classification.

• Regression: Unlike classification, which deals with categorical labels, regression is used to predict a calculated value. For example, we can train the model by using item descriptions and prices in a shop and then use it to predict a price for an item it has never seen before.

• Ranking: Ranking models are trained to understand what characteristics make entities rank better compared to other entities. For example, such a model would be able to score someone’s writing style on a scale from 1 to 10.

• Recommendations: Recommendation engines are used to recommend items to users that other users with similar preferences gave a high rank to. So this type of ML task deals both with finding the similarities between the users and understanding the subjective preferences of each group. In a way, it's a combination of classification and ranking.

• Forecasting: Forecasting models look for patterns in the data and not just individual data points. The task is similar to a regression, but it also involves time series. Therefore, the job of the model will not only be to predict calculated values based on some characteristics of a specific item but also to factor in how the value has been changing historically.