Weed composition may vary because of natural environment, managementpractices, and their interactions. In this study we presented a systematicapproach for analyzing the relative importance of environmental andmanagement factors on weed composition of the most conspicuous species insugarcane. A data-mining approach represented by k-meanscluster and classification and regression trees (CART) were used foranalyzing the 11 most frequent weeds recorded in sugarcane cropping systemsof northern Argentina. Data of weed abundance and explanatory factorscontained records from 1976 sugarcane fields over 2 consecutive years. The k-means method selected five different weed clusters.One cluster contained 44% of the data and exhibited the lowest overall weedabundance. The other four clusters were dominated by three perennialspecies, bermudagrass, johnsongrass, and purple nutsedge, and the annualitchgrass. The CART model was able to explain 44% of the sugarcane's weedcomposition variability. Four of the five clusters were represented in theterminal nodes of the final CART model. Sugarcane burning before harvestingwas the first factor selected in the CART, and all nodes resulting from thissplit were characterized by low abundance of weeds. Regarding the predictivepower of the variables, rainfall and the genotype identity were the mostimportant predictors. These results have management implications as theyindicate that the genotype identity would be a more important factor thancrop age when designing sugarcane weed management. Moreover, the abioticcontrol of crop–weed interaction would be more related to rainfall than theenvironmental heterogeneity related to soil type, for example soilfertility. Although all these exploratory patterns resulting from the CARTdata-mining procedure should be refined, it became clear that thisinformation may be used to develop an experimental framework to study thefactors driving weed assembly.