Abstract: In one direction S_p-rough sets，the knowledge[x]（R-element equivalence class[x]） with the attribute set α has such characteristic：If there are new attributes supplemented to α，then the elements in [x] will decrease；By using these characteristics and considering the reliability characteristics of the element transference random，the concepts of S_p-upper ladder knowledge，the generation of depending on reliability of S_p-upper ladder knowledge，and attribute dependence of S_p-upper ladder knowledge are presented.Attribute dependence mining theorem of Sp-upper ladder knowledge，and attribute dependence algorithm are also given.

Key words: S_p-upper ladder knowledge, attribute dependence, generation of depending on reliability, dependence mining, unit circle theorem, state recognition, mining algorithm

摘要： 单向S_p-粗集中，具有属性集α的知识[x]（R-元素等价类[x]）具有这样的特征：若α内被补充属性，则[x]内的元素个数被减少。利用这一特征，考虑属性补充的随机性，给出S_p-上阶梯知识，S_p-上阶梯知识的依信度生成，S_p-上阶梯知识属性依赖的原理，给出S_p-上阶梯知识的属性依赖挖掘定理，S_p-上阶梯知识的状态识别算法。

关键词: S_p-上阶梯知识, 属性依赖, 依信度生成, 依赖挖掘, 单位圆定理, 状态识别, 挖掘算法, ,