The paper delves into the challenge of classification using dispersed data gathered from independent sources. The examined approach involves local models as ensembles of decision trees or random forests constructed based on local data. In the proposed model, a conflict analysis is used to identify the coalitions of local models. Two variants of forming coalitions were checked – unified and diverse – and two different strategies for generating final decisions were explored, allowing one or two of the strongest coalitions to make decisions. The diverse coalition approach is a wholly new and innovative strategy. The methods were tested and compared with corresponding accuracy-based weighted variants. The proposed approach improves classification performance, with weighted variants outperforming unweighted ones in balanced accuracy. Diverse model coalitions are especially effective for challenging and heterogeneous datasets.