Skip to content

Stockertim Dictionary · Artificial intelligence

Predictive model

Also called predictive models, predictive analytics model

In short

A predictive model is a model trained on past data to estimate a specific outcome, such as the chance that a loan defaults or a payment is fraudulent.

  • Intermediate
  • Technology
  • Checked

In detail

A predictive model learns the link between inputs and a known outcome from historical records. A credit model, for example, studies past borrowers' income, repayment history and existing debts alongside whether each one later defaulted, and produces a score or probability for a new applicant.

This differs from a generative model, which produces new text, images or code rather than estimating one outcome. A predictive model gives a narrow answer to a narrow question, which makes it easier to test: its estimates can be compared with what actually happened. Common methods range from logistic regression and decision trees to neural networks.

Predictive models matter in finance because they sit behind decisions on lending, fraud alerts, insurance pricing and collections. Their quality depends on whether the past data resemble the present. When customer behaviour or the economy shifts, a model trained on older patterns can lose accuracy, and if the training data carried unfair patterns the model can repeat them.

For this reason firms validate predictive models before use, monitor their results against real outcomes and keep a person accountable for decisions that affect customers.

Worked example

Northwind Bikes' finance arm scores 2,000 loan applications with a predictive model; one group receives an estimated default probability of 5%.

  1. The group holds 400 applicants with an estimate of 5%.
  2. If the model is well calibrated, expected defaults: 400 x 0.05 = 20
  3. A year later 30 of the 400 have defaulted.
  4. Observed rate: 30 / 400 = 7.5%

The model underestimated defaults for this group, so the team investigates before relying on it further.

In practice

Banks, lenders and insurers use predictive models for credit scoring, fraud detection, churn estimates and claims triage. Model risk teams track how estimated probabilities compare with actual outcomes over time.

Watch out

  • A probability is not a verdict: a 5% default estimate means some such borrowers still default.
  • Predictive does not mean it sees the future; it projects past patterns, which can break when conditions change.

Learn it properly

Official sources

Facts last checked on 3 November 2026.

Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.

English