AI-ordlista
Den kompletta ordlistan över AI
IQR Method
Non-parametric technique identifying outliers as those located outside 1.5 times the interquartile range above the third quartile or below the first.
Grubbs' Test
Statistical hypothesis test identifying univariate outliers by assuming a normal distribution and testing whether the extreme value is significantly different.
Connectivity-Based Outlier Factor
Measure based on the connectivity of k-nearest neighbors assessing the degree of isolation of a point relative to the overall structure of the neighborhood graph.
Angle-Based Outlier Detection
Method exploiting angular properties between points in the data space, where anomalies exhibit lower angular variances than normal points.
Histogram-Based Outlier Score
Multidimensional histogram-based approach evaluating the rarity of an observation based on the frequency of the histogram cells it occupies.
k-NN Outlier Detection
Method using the distance to k-nearest neighbors as an anomaly score, where points with the highest average distances are considered outliers.
Subspace Outlier Detection
Technique identifying anomalies in different dimensional subspaces where they may be more detectable than in the full dimensional space.
Feature Bagging
Ensemble method creating multiple detectors on different random subsets of features, aggregating their results to improve detection robustness.
PCA Anomaly Detection
Use of principal component analysis to detect anomalies as points with high reconstruction scores after projection onto principal components.