Prediction Types

The first step in ML Studio - Choose what to predict - asks a single question: what kind of answer are you after? Your answer determines everything that follows, so ML Studio presents three clearly explained choices and helps you pick.

The three prediction type cards: Binary, Classification, and Regression

You do not need any data-science background. Each card explains what the type means, when to use it, and gives an example phrased in process terms.

The three types

Binary - "yes or no"

Predict whether something will or will not happen - exactly two possible answers.

Use this when:

  • The answer is one of two outcomes (yes / no, pass / fail, will / won't).
  • You want to flag at-risk cases so a team can act early.

Example in your process: "Will this case miss its SLA?" -> Yes or No, with a confidence.

Binary is the simplest and most common prediction, and the best place to start.

Classification - "which category"

Predict which one of several categories a case will fall into.

Use this when:

  • There are three or more possible outcomes.
  • You want the single most likely category, along with the runners-up.

Example in your process: "What is the final outcome?" -> Approved, Rejected, Refunded, or Escalated.

Classification learns the patterns that separate each category. Pick it when a simple yes / no is not enough and you need to know which kind.

Regression - "a number"

Predict a numeric value - an amount, a duration, or a count.

Use this when:

  • The answer is a number, not a category.
  • You care how much, how long, or how many.

Example in your process: "How many days late will this case be?" -> about 3.4 days.

Regression estimates a continuous value and tells you how close it usually gets. Pick it when the answer is measured, not labelled.

Not sure which to pick?

A quick rule of thumb:

If your answer is... Choose
yes / no Binary
one of a few named outcomes Classification
a number Regression

What the type controls

Your choice shapes the rest of the flow:

  • Which columns you can predict. On the next step, ML Studio only offers columns that fit the type you picked - two-value columns for Binary, categorical columns for Classification, and numeric columns for Regression. See Building a prediction.
  • How the result is measured. Each type is scored with the metrics that make sense for it - for example accuracy for categories, or typical error for numbers. See Training and results.
  • What gets written back onto your cases. Binary and Classification add a predicted label plus a confidence; Regression adds a predicted number. See Using your predictions.

Once you have chosen a type, click Choose Binary, Choose Classification, or Choose Regression to move on.