Scikit Learn Model Predict — Machine Learning/Python Models

Scikit-learn Model Predict

Machine Learning / Python Models

Runs a saved scikit-learn estimator on the input, once per sample: y = model.predict(x). The estimator is loaded from its file when the run starts and called in the embedded Python interpreter, the way MATLAB’s Scikit-learn Model Predict block calls it.

Ports

• Input – x, the features. A vector (a column or a row) of n is one observation, handed to predict as a (1, n) array; an m×n matrix is m observations, one per row. It is cast to the Input Data Type first.
• Output – y, what predict returns: one value per observation, so [1,1] for a vector input and [m,1] for m observations. A 2-D result keeps its shape. Class labels arrive as numbers; an estimator that predicts text cannot drive a port.

Parameters

• Model File – the file the estimator was saved to. Empty runs a built-in example estimator, y = 0.5·(the sum of the features) + 0.25, which needs no scikit-learn. A relative path is read from the directory the app was started in.
• Model Load Command – how the file is read:
  • pickle.load() – a pickle file.
  • joblib.load() – a joblib file; needs joblib.
  • skops.io.load() – a skops file, loaded as trusted; needs skops.
  Pickle and joblib files run code as they load: open only files you trust.
• Input Data Type – the numpy type the input is cast to before predict: float32 (the default, as in MATLAB’s block), float64, float16, float, int8, int16, int32, int64, int, uint8, uint16, uint32 or uint64. The cast rounds (float32) or truncates (the integer types), and the estimator sees the cast values.
• Preprocessing File – optional: a Python file defining preprocess(model, inputs), which receives the list of input arrays and returns the list predict gets.
• Postprocessing File – optional: a Python file defining postprocess(model, outputs), which receives predict’s list and returns the list the output takes.
• Sampling Time (s) – zero or less inherits the solver's rate; a positive value runs the block at that period.

Code export

Python only – the generated module loads the estimator from the same file and calls it exactly as the simulation does. The other nine targets (MATLAB, Java, Rust, C, C++, VHDL, Verilog, SystemVerilog, PLC Structured Text) are not supported: none can carry a Python interpreter, so an export to any of them fails loudly for this block.

Simulink bridge

Both directions, to statsPycoex/Scikit-learn Model Predict. Model File ↔ ModelFile, Model Load Command ↔ ModelLoadCommand (its three options 1:1), Preprocessing File ↔ PreprocessingFilePath, Postprocessing File ↔ PostprocessingFilePath, and Input Data Type crosses as the type column of InputTable. Input and output permutations have no setting here and are reported, not crossed. A block on the built-in example has no file to name and is reported on export. Sampling Time (s) → SampleTime, as on every block.

Notes

• Discrete by nature: the estimator is called once per sample, never integrated.
• Before the first sample predict runs once on inputs of all ones, as in MATLAB’s block; the output size is taken from it and must not change during the run.
• Needs the Python runtime, and scikit-learn wherever a real model file is named; a model that fails to load or to predict is reported once, with the block’s path, and stops the run.