predict — Predict responses from a fitted model.

predict(mdl,Xnew) evaluates supported fitted model objects, including LinearModel from fitlm, ClassificationTree from fitctree, and supported Deep Learning network objects, on new predictor data.

Syntax

ypred = predict(mdl, Xnew)
[ypred, yci] = predict(mdl, Xnew)
[ypred, yci] = predict(mdl, Xnew, Name, Value)
label = predict(tree, Xnew)
[label,score,node,cnum] = predict(tree, Xnew)
[label,score] = predict(linearClassifier, Xnew)

Inputs

Returns

Returned values from predict depend on how many outputs the caller requests.

Errors

How predict works

• Typed-integer predictors and numeric controls for implemented statistical models are independently gated RunMat-mode extensions with checked double boundaries; supported deep-learning models retain their documented integer-input contract.
• Xnew can be a numeric matrix with the same predictor column count used during fitting.
• For table input, predictor variables are selected by the model's PredictorNames property.
• For LinearModel, [ypred,yci] = predict(...) returns pointwise confidence intervals for curve predictions by default.
• For ClassificationTree, [label,score,node,cnum] = predict(...) returns predicted class labels, posterior probabilities, one-based leaf node ids, and one-based class-number predictions.
• For dlnetwork, SeriesNetwork, and DAGNetwork compatibility objects, predict(net,Xnew) returns the score/activation matrix for the supported sequential feed-forward layer subset also used by forward.
• Deep Learning prediction accepts ObservationsIn with rows or columns; column-observation input is transposed before execution.
• Linear-model name-value options Alpha, Prediction, and Simultaneous are accepted. Prediction can be curve or observation; Simultaneous must be false because simultaneous intervals are not supported yet.
• Classification-tree prediction currently accepts matrix or table predictor data and raises explicit errors for unsupported pruning/subtree options.

Examples

Predict from a linear model

mdl = fitlm([0; 1; 2; 3], [1; 3; 5; 7]);
yhat = predict(mdl, [4; 5])

Expected output:

yhat is approximately [9; 11].

Return confidence intervals

[yhat, yci] = predict(mdl, [4; 5], 'Alpha', 0.01)

Expected output:

yci is a two-column matrix with lower and upper bounds.

Predict classes from a classification tree

tree = fitctree([0; 1; 2; 3], [0; 0; 1; 1], 'MaxNumSplits', 1, 'MinParentSize', 2);
[label,score] = predict(tree, [0.5; 2.5])

Expected output:

label is [0; 1], and score has one posterior-probability column per class.

Predict from a dlnetwork

layers = {featureInputLayer(2,'Name','in'); fullyConnectedLayer(2,'Name','fc'); softmaxLayer('Name','prob')};
net = dlnetwork(layers);
scores = predict(net, [1 2; 3 4])

Expected output:

scores has one row per observation.

Using predict with coding agents

Open a RunMat example with live inputs, then ask the agent to explain how predict changes the result.

FAQ

Related Stats functions

Open-source implementation

Unlike proprietary runtimes, every RunMat function is open-source. Read exactly how predict is executed, line by line, in Rust.

• View the source for predict in Rust on GitHub
• Learn how the RunMat runtime works
• Found a bug? Open an issue with a minimal reproduction.

About RunMat

RunMat is an open-source runtime that executes MATLAB-syntax code blazing on any GPU. It is licensed under the Apache 2.0 license.

• RunMat automatically optimizes your math for GPU execution on Apple, Nvidia, and AMD hardware. No code changes needed. Simulations that took hours now take minutes.
• Start running code in seconds. RunMat runs in the browser, on the desktop, or from the CLI. No license server, no IT ticket.

Getting started · Benchmarks · Pricing