MetricComputationEngine<T>

Models & Types · AiDotNet.Evaluation.Engines

Core engine for computing evaluation metrics across all task types.

For Beginners

This is the heart of model evaluation. Give it predictions and actuals, and it computes all relevant metrics automatically based on task type.

How It Works

This engine provides a unified interface for computing classification, regression, and time series metrics with confidence intervals and proper handling of edge cases.

Constructors

Constructor Summary
MetricComputationEngine(EvaluationOptions<T> options) Initializes the metric computation engine with default or custom options.

Methods

Method Summary
IReadOnlyList<string> GetAvailableMetricNames(bool classification, bool regression, bool probabilistic) Gets all available metric names.
MetricCollection<T> ComputeClassificationMetrics(ReadOnlySpan<T> predictions, ReadOnlySpan<T> actuals, ReadOnlySpan<T> probabilities, int numClasses) Computes all classification metrics.
MetricCollection<T> ComputeRegressionMetrics(ReadOnlySpan<T> predictions, ReadOnlySpan<T> actuals) Computes all regression metrics.
MetricWithCI<T> ComputeMetric(string metricName, ReadOnlySpan<T> predictions, ReadOnlySpan<T> actuals, ReadOnlySpan<T> probabilities, int numClasses) Computes a specific metric by name.
void RegisterClassificationMetric(IClassificationMetric<T> metric, string name) Registers a classification metric.
void RegisterProbabilisticMetric(IProbabilisticClassificationMetric<T> metric, string name) Registers a probabilistic classification metric.
void RegisterRegressionMetric(IRegressionMetric<T> metric, string name) Registers a regression metric.