Ordered data analytics is frequently mentioned in many areas including theoretical research, commercial analysis, and industrial application. In 2001, Yao and Sai proposed a method for mining ordering rules in an ordered information table. The foundation of an ordered information table is a standard information table adding order relations. In 2012, Yao proposed a theory of three-way decision based on three pair-wise disjoint regions in rough set theory. The theory is then generalized as thinking in threes. This thesis focuses on building an ordered information table with two types of attributes. Order relations are binary relations that reveal the user preference between attribute values or objects. The two important properties of an order relation are asymmetry and negative transitivity. They lead to that an order relation is a weak order. Attributes in a table, which are classified by different measurements, come with different manifestations as numerical attributes and categorical attributes. Order relations and orderings on these two types of attributes are discussed. By combing order relations and an information table, an ordered information table can be obtained. Several studies are working on ordered information. Roy proposed an outranking relation which satisfies reflexivity but does not necessarily satisfy a transitivity property. Greco et al. extended the rough set approach by introducing a dominance relation which is called dominance-based rough set theory. The degree of preference hasn’t received enough attention. Motivated by a three-way interpretation of fuzzy set and shadowed set, we propose three-way approximations of an order relation. The main purpose of constructing three-way approximations is to describe the preference between every two objects by three regions of object pairs. Following the evaluation-based trisection methods, we consider the degree of preference and set up a pair of thresholds in order to separate all object pairs into three sets. Object pairs, that are located in different regions, have different order relationships. The original ordered information table is then converted into a three-valued table. In this case, ordered description language is introduced in this thesis, two types of rules are discussed, and a rough set approach is used for mining ordering rules.