Abstract: The identification of flotation conditions plays a crucial role in froth flotation engineering, relying solely on subjective identification based on manual experience, which leads to low accuracy and efficiency. Therefore, a weighted K−nearest neighbors (KNN) algorithm considering the interactions among froth image features was proposed for the recognition of flotation condition categories. Firstly, the correlation between bubble image features and flotation condition categories was quantified using information entropy, while also assessing the redundancy of this feature with respect to other features. Subsequently, the difference between the correlation of this feature with flotation condition categories and its redundancy with other features was calculated, and this difference was used as the weight of the feature. Furthermore, in the KNN algorithm, feature weighting was applied to the Euclidean distance to achieve feature weighting in the KNN algorithm. Next, the feature selection process was embedded into the training process of the feature−weighted KNN classification algorithm, and the feature subset with the highest classification accuracy was selected as the optimal feature subset. Finally, flotation condition recognition was performed based on the optimal feature subset. Experimental results demonstrated that compared to other benchmark classification algorithms, the proposed method achieved the best performance in terms of classification accuracy and time, thereby validating the effectiveness of the proposed flotation condition identification method in this study.