Class TimeSeriesIsolationForestOptions<T>

Namespace

AiDotNet.Models.Options

Assembly

AiDotNet.dll

Configuration options for Time Series Isolation Forest anomaly detection.

public class TimeSeriesIsolationForestOptions<T> : TimeSeriesRegressionOptions<T>

Type Parameters

T

The numeric type used for calculations (typically double or float).

Inheritance

object

ModelOptions

RegressionOptions<T>

TimeSeriesRegressionOptions<T>

TimeSeriesIsolationForestOptions<T>

Inherited Members

TimeSeriesRegressionOptions<T>.LagOrder

TimeSeriesRegressionOptions<T>.IncludeTrend

TimeSeriesRegressionOptions<T>.SeasonalPeriod

TimeSeriesRegressionOptions<T>.AutocorrelationCorrection

TimeSeriesRegressionOptions<T>.ModelType

TimeSeriesRegressionOptions<T>.LossFunction

RegressionOptions<T>.DecompositionMethod

RegressionOptions<T>.UseIntercept

ModelOptions.Seed

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

Time Series Isolation Forest extends the classic Isolation Forest algorithm to handle temporal data by incorporating lag features, rolling statistics, and seasonal decomposition.

For Beginners: Isolation Forest works by randomly isolating observations. Anomalies are easier to isolate because they are "few and different" - they end up in shorter branches of the isolation trees.

For time series, we enhance this by considering:

• Lag Features: How the current value relates to recent past values
• Rolling Statistics: Moving averages, standard deviations
• Seasonal Patterns: Accounting for regular patterns like daily/weekly cycles
• Trend: Long-term direction of the data

This makes it effective for detecting:

• Sudden spikes or drops (point anomalies)
• Values that are unusual given the context (contextual anomalies)
• Unusual patterns over time (collective anomalies)

Constructors

TimeSeriesIsolationForestOptions()

Creates a new instance with default values.

public TimeSeriesIsolationForestOptions()

TimeSeriesIsolationForestOptions(TimeSeriesIsolationForestOptions<T>)

Creates a copy of the specified options.

public TimeSeriesIsolationForestOptions(TimeSeriesIsolationForestOptions<T> other)

Parameters

other TimeSeriesIsolationForestOptions<T>

Properties

ContaminationRate

Gets or sets the expected proportion of anomalies in the dataset.

public double ContaminationRate { get; set; }

Property Value

double

Remarks

For Beginners: This is used to set the threshold for what counts as an anomaly. A value of 0.1 means you expect about 10% of points to be anomalies. If unsure, start with 0.05-0.1 and adjust based on results.

LagFeatures

Gets or sets the number of lag features to include.

public int LagFeatures { get; set; }

Property Value

int

Remarks

For Beginners: Lag features capture how the current value relates to past values. For hourly data, lag=24 captures the daily pattern. For daily data, lag=7 captures the weekly pattern.

MaxDepth

Gets or sets the maximum depth of each isolation tree.

public int? MaxDepth { get; set; }

Property Value

int?

Remarks

For Beginners: Limiting depth speeds up the algorithm without much accuracy loss. A depth of ceil(log2(sample_size)) is typically sufficient since anomalies are isolated early. If null, uses the theoretical limit.

NumTrees

Gets or sets the number of isolation trees in the forest.

public int NumTrees { get; set; }

Property Value

int

Remarks

For Beginners: More trees generally give more stable anomaly scores, but take longer to build. 100-200 trees is usually sufficient.

RandomSeed

Gets or sets the random seed for reproducibility.

public int? RandomSeed { get; set; }

Property Value

int?

RollingWindowSize

Gets or sets the window size for rolling statistics (mean, std, min, max).

public int RollingWindowSize { get; set; }

Property Value

int

Remarks

For Beginners: Rolling statistics help detect if a value is unusual compared to its recent neighbors. A window of 10-30 is typical.

SampleSize

Gets or sets the number of samples to use when building each tree.

public int? SampleSize { get; set; }

Property Value

int?

Remarks

For Beginners: A smaller sample size (e.g., 256) actually works better because anomalies are easier to isolate in smaller samples. This also makes the algorithm faster. If null, defaults to min(256, n_samples).

UseSeasonalDecomposition

Gets or sets whether to decompose the series into trend and seasonal components.

public bool UseSeasonalDecomposition { get; set; }

Property Value

bool

Remarks

For Beginners: Seasonal decomposition helps detect anomalies that deviate from expected seasonal patterns. Enable this for data with clear daily, weekly, or yearly patterns.

UseTrendFeatures

Gets or sets whether to include trend-based features.

public bool UseTrendFeatures { get; set; }

Property Value

bool

Remarks

For Beginners: Trend features help detect anomalies that deviate from the long-term direction of the data.