AI-ordlista
Den kompletta ordlistan över AI
Numerical Variable Transformation
Application of mathematical transformations to continuous variables to improve their distribution and relationship with the target.
Categorical Variable Encoding
Converting textual or categorical data into a numerical format usable by machine learning algorithms.
Handling Missing Values
Imputation techniques and treatment of missing data to maintain dataset integrity.
Creation of Temporal Features
Extraction and generation of time-based variables from temporal or sequential data.
Text Feature Extraction
Transforming unstructured text into numerical vectors using techniques like TF-IDF, embeddings, and n-grams.
Geospatial Engineering
Creating variables from location data and geographic coordinates to capture spatial relationships.
Normalization and Standardization
Scaling variables for comparability and optimal convergence of learning algorithms.
Feature Selection
Identification and retention of the most relevant variables to improve performance and reduce complexity.
Dimensionality Reduction
Techniques such as PCA and t-SNE to compress information while preserving important variations.
Creation of Interactions
Generation of new features through multiplicative or additive combination between existing variables.
Transformation of Distributions
Application of logarithmic, Box-Cox or Yeo-Johnson transformations to normalize skewed distributions.
Image Feature Extraction
Conversion of visual data into numerical descriptors via histograms, textures, and local descriptors.
Audio Feature Engineering
Extraction of spectral and temporal features such as MFCC, chroma, and spectrograms from audio signals.
Temporal Aggregation
Creation of rolling statistics and aggregates over time windows to capture trends and patterns.
Polynomial Features
Generation of higher-order terms to capture non-linear relationships between variables.
Binning and Discretization
Dividing continuous variables into discrete intervals to capture non-linear effects and reduce noise.
Time Series Engineering
Creating lag features, moving averages, and seasonal decompositions to model temporal dependencies.
Graph Feature Extraction
Generation of descriptors from network structures such as centrality, clustering, and node embeddings.