Market basket analysis is a form of association rule learning within data mining that identifies statistically significant relationships between items in transactional datasets to predict item co-occurrence. The theory relies on specific metrics, including support (the probability of an itemset occurring), confidence (conditional probability given the antecedent exists), and lift (a measure of correlation relative to independence). By quantifying these probabilities against a minimum threshold, the method determines which products are likely purchased together, enabling cross-selling strategies based on probabilistic inference rather than temporal or causal sequences.
Prerequisite chain context: requires Association Rules LHS RHS Frequency Interpretation.