A classification model

Published

August 17, 2022

Natalie has been a software developer for quite some time. Although her job keeps her motivated, she wants to take her career to the next level.

A critical step she wants to take is introducing machine learning into her work. She started learning some of the fundamentals and is now ready to apply what she’s learned.

She researched one of her company’s problems and learned she needed to build a supervised learning classification model. She had enough labeled data, so it seemed like a good fit.

Based on this, which of the following better describes what Natalie needs to accomplish?

  1. She needs to train a model that returns a numerical prediction for each sample of data.

  2. She needs to train a model that clusters the data into different groups based on their characteristics.

  3. She needs to train a model to predict the class of every sample of data out of a predefined list of classes.

  4. She needs to train a model that returns the optimal policy that maximizes the potential outcomes of her problem.

3

Natalie’s problem requires her to predict the class of every sample of data out of a predefined list of classes. That’s the goal of machine learning classification models.

The first choice refers to a regression model. Here we want the model to output a single, continuous value. For example, imagine we want to return the predicted price of a house or the predicted highest temperature for the weekend.

The second choice refers to a clustering model. These unsupervised learning techniques are helpful when we don’t have labels for our data and want the algorithm to group every sample into dynamically generated groups.

The fourth choice is a loose description of a reinforcement learning approach, where we want an agent to learn the optimal policy that maximizes a reward function.

Recommended reading