CoOccurrence

CoOccurrence refers to the phenomenon in which two or more items appear together within a defined dataset or context. This can include words in a document, events within a time window, species observed in an ecological sample, or products found in a shopping basket. Co-occurrence is used to identify associations and dependencies that are not necessarily causal.

In text mining and natural language processing, co-occurrence is used to build matrices where the rows and columns represent terms and the cells record how often two terms appear within a given window or sentence. Common measures include raw co-occurrence counts, pointwise mutual information, chi-squared statistics, normalized PMI, and cosine similarity between co-occurrence vectors. These data underpin tasks such as word embeddings, clustering, topic modeling, and network analyses.

Co-occurrence concepts also appear in other domains, including market basket analysis, epidemiology for co-occurring symptoms, and ecological studies of species associations. Co-occurrence networks visualize relationships as graphs with nodes representing items and edges weighted by co-occurrence strength, aiding interpretation and further analysis such as community detection or centrality measures.

Methodological considerations for co-occurrence analysis include the choice of window size or context, normalization approaches, baseline frequency, and sampling bias. Interpreting co-occurrence requires caution to distinguish statistical association from causation. The term serves as a general descriptor rather than a single statistic, and its precise meaning depends on the data type, context, and analytical goals.