AI-ordlista
Den kompletta ordlistan över AI
Multi-criteria collaborative filtering
Extension of classic collaborative filtering that simultaneously considers multiple evaluation dimensions to predict user preferences and improve the accuracy of personalized recommendations.
Preference aggregation
Mathematical process combining a user's multiple ratings on different criteria to produce an overall preference score used in ranking recommendations.
Multi-dimensional utility matrix
Data structure representing user ratings along several preference axes, allowing for fine-grained modeling of complex tastes beyond a simple satisfaction score.
Criteria weighting
Technique assigning relative importance coefficients to each evaluation dimension to reflect the hierarchy of individual preferences and influence the recommendation algorithm.
Multi-objective utility function
Mathematical model transforming ratings on multiple criteria into a single utility value, integrating trade-offs between different preference dimensions.
Preference space
Multidimensional vector representation where each axis corresponds to an evaluation criterion, enabling visualization and calculation of similarities between complex user profiles.
Vector user profile
Composite mathematical representation of a user's preferences as a multidimensional vector, where each component encodes the preference intensity for a specific criterion.
Pareto-optimal recommendation
Set of items that cannot be improved on one criterion without degrading performance on at least one other criterion, constituting the best solutions in a multi-objective context.
Multi-criteria hybrid system
Architecture combining multiple recommendation techniques (collaborative, content-based, knowledge-based) while explicitly managing multiple user preference dimensions.
Criteria sensitivity analysis
Method evaluating the impact of variations in the relative importance of each criterion on final recommendations, enabling identification of the most influential dimensions.
Preference normalization
Process of standardizing evaluation scales of different criteria to make them comparable and mathematically manipulable in a coherent multi-dimensional framework.
Additive preference model
Approach where the overall utility of an item is calculated as the weighted sum of partial utilities on each criterion, assuming independence of preference dimensions.
Multi-criteria implicit feedback
Automatic inference of user preferences on multiple dimensions from observed behaviors (clicks, viewing time, purchases) without direct explicit evaluation.
Criteria slicing
Technique for segmenting preference dimensions into more granular sub-categories to refine personalization and capture nuances in user tastes.
Preference mapping
Graphical visualization of relationships between users and items in multi-criteria space, revealing preference clusters and complex behavioral patterns.
Preference elicitation
Interactive process of acquiring weights and relative importance of different criteria from users, often through comparative questions or analytical methods.
Composite scoring function
Mathematical algorithm combining individual scores of each criterion into a single relevance metric, incorporating personalization parameters and domain constraints.
Preference divergence
Measure quantifying the gap between a user's multi-criteria preferences and the characteristics of a recommended item, used to optimize overall fit.