Abstract: TSK (Takagi-Sugeno-Kang) fuzzy system has been widely used in regression, classification and decision-making, and has showed a great precision and interpretability. But for the modeling scene with monotonic data, the modeling effect of TSK fuzzy system is not ideal enough. In order to overcome the above challenge, this paper proposes a novel monotonic TSK fuzzy system (MC-TSK). The MC-TSK satisfies monotonicity by adding monotonic constraints to the original TSK fuzzy system model. MC-TSK does not require that all the monotonic relationships between features and the decision attribute are consistent, which relaxes the assumption that consistent monotonicity is used in the most existing methods for dealing with monotonic classification problems. The extensive experimental results show that MC-TSK has better classification performance and maintains the characteristics of interpretability.

Key words: TSK fuzzy system, monotonic classification, monotonic constraints, classification performance, interpretability

摘要： TSK（Takagi-Sugeno-Kang）模糊系统已被广泛应用于回归、分类和决策等方面，并展现出了良好的精度和可解释性，但是对于存在单调数据的建模场景，TSK模糊系统的建模效果还不够理想。针对此，提出了一个新颖的单调TSK模糊系统（MC-TSK）。通过在原始的TSK模糊系统模型上添加单调约束使MC-TSK满足单调性。MC-TSK不要求特征和输出之间存在的单调关系是一致的，这放宽了用于处理单调分类问题时大多数现有方法中使用的一致单调性的假设。较广泛的实验结果表明，与已有的处理单调分类的方法相比，MC-TSK具有更好的分类性能并且保持了可解释性等方面的特点。

关键词: TSK模糊系统, 单调分类, 单调约束, 分类性能, 可解释性