Aim of study: Aassessing the existence of consistent co-occurrence between tree species that characterize seasonal tropical forests, using the association rules analysis (ARA), that is a novel data mining methodology; and evaluate evaluating the taxonomic and functional similarities between associated species.Area of study: forty-four seasonal forest sites with permanent plots (40.2 ha of total sample) located in Southeast Brazil, from which we obtained species occurrences.Material and methods: we applied association rules analysis (ARA) to the dataset of species occurrence in sites considering the criteria of support equal to or greater than 0.63 and confidence equal to or greater than 0.8 to obtain the first set of associations rules between pairs of species. This set was then submitted to Fisher’s criteria exact p-value less than 0.05, lift equal to or greater than 1.1 and coverage equal to or greater than 0.63. We considered these criteria to be able to select non-random and consistent occurring associations.Main results: We obtained a final result of 238 rules for semideciduous forest and 11 rules for deciduous forests, composed of species characteristic of vegetation types. Co-occurrences are formed mainly by non-confamilial species, which have similar functional characteristics (potential size and wood density). There is a difference in the importance of co-occurrence between forest types, which tends to be less in deciduous forests.Research highlights: The results point to out the feasibility of applying ARA to ecological datasets as a tool for detecting ecological patterns of coexistence between species and the ecosystems functioning.Keywords: data mining; coexistence; semideciduous forests; deciduous forests; biotic interaction.