LDCOFDetector<T>
Models & Types · AiDotNet.AnomalyDetection.DistanceBased
Detects anomalies using LDCOF (Local Density Cluster-Based Outlier Factor).
Inheritance: ModelBase<T, Matrix<T>, Vector<T>> → AnomalyDetectorBase<T> → LDCOFDetector<T>
Implements: IGeneratedParameterRegistrar<T>
For Beginners
LDCOF combines clustering with density-based outlier detection. It first clusters the data, then computes outlier scores based on how a point’s density compares to its cluster’s density. Points in sparse regions of dense clusters are flagged.
How It Works
The algorithm works by:
- Cluster the data using k-means
- Compute local density for each point
- Compare point density to cluster average density
- Large deviations indicate outliers
When to use:
- When data has natural cluster structure
- For detecting local outliers within clusters
- When global methods miss cluster-specific anomalies
Industry Standard Defaults:
- Number of clusters: 8
- Number of neighbors (k): 10
- Contamination: 0.1 (10%)
Reference: Amer, M., Goldstein, M. (2012). “Nearest-Neighbor and Clustering based Anomaly Detection Algorithms for RapidMiner.” Workshop on Open Source Data Mining Software.
Constructors
| Constructor | Summary |
|---|---|
LDCOFDetector(int numClusters, int numNeighbors, double contamination, int randomSeed) |
Creates a new LDCOF anomaly detector. |
Properties
| Property | Summary |
|---|---|
int NumClusters { get; } |
Gets the number of clusters. |
int NumNeighbors { get; } |
Gets the number of neighbors for density estimation. |
Methods
| Method | Summary |
|---|---|
Vector<T> ScoreAnomalies(Matrix<T> X) |
|
void Fit(Matrix<T> X) |