Abstract: Existing methods for tabular data prediction primarily focus on classical classification and regression tasks. However, there is a type of data in the tabular data domain where the labels have an ordinal relationship, and its prediction task is called tabular ordinal classification. Current methods for tabular ordinal classification mainly rely on retrieving similar features and augmenting the sample feature representation by fusing the similar features with the ordinal distance between classes. However, existing methods neglect the full utilization of label ordinal knowledge. To address this, a method based on ordinal label entropy optimization is proposed, which effectively guides the model to learn ordinal information by mining the ordinal entropy embedded in the label order knowledge. Specifically, first establish an ordinal entropy calculation module that quantifies the ordinal entropy based on the ranking differences between the predicted and true labels. Through step-by-step analysis and derivation, the ordinal label entropy is designed as a novel rank loss function, which is introduced as a regularization term into the model. This encourages the model to learn the ordinal relationship between labels and reduces the information loss caused by unordered predictions. Then, the ordinal-entropy optimized ranking loss function is combined with the original loss function of the model to jointly improve the model's predictive ability. Finally, experimental results on multiple ordinal tabular datasets show that this method outperforms various baseline models, fully demonstrating the effectiveness and advantages of the ordinal entropy optimization model in tabular ordinal classification tasks.

摘要： 现有表格数据预测方法主要聚焦于传统分类和回归的研究，然而在表格数据领域中存在一种标签具有有序关系的数据类型，其预测任务被称为表格有序分类。目前表格有序分类方法主要采用检索相似特征的方式，通过相似特征与类间有序距离融合增强样本特征表示。但现有方法忽略了标签有序知识的充分利用，因此提出一种基于标签有序熵优化的方法，通过挖掘标签有序知识中蕴含的有序熵，有效指导模型学习有序信息。具体而言，首先建立有序熵计算模块，利用预测标签与真实标签之间的等级顺序差异量化有序熵。通过逐步分析和推导，将标签有序熵设计为一种新颖排序损失函数，作为正则项引入模型，鼓励模型学习标签等级顺序关系，以减少无序预测带来的信息损失。然后，将有序熵优化排序损失函数与模型原有损失函数相结合，共同提升模型的预测能力。最后，在多个有序表格数据集上的实验结果显示，该方法相较于多种基线模型取得了性能提升，充分证明了有序熵优化模型在表格有序分类任务中的有效性与优势。