蒙医方剂数据挖掘关键技术研究

The traditional Chinese medicine prescription is the treasure of Chinese culture. It contributes a lot to human health. The rule of prescription plays an important role on shorten the period of the development of new Traditional Chinese medicine. There are many results on utilizing data mining technology to find the rule of prescription. But there is not a complete and mature prescription database. The technology and the aim of mining are traditional and single. It can not fit the development of medicine research. The Mongolian Medicine is a branch of the traditional Chinese medicine. Compared with the traditional Chinese medicine, it has some same aspects but also has unique principle and treatment. At present, data mining in Mongolian medicine prescription is still empty. This project will establish the Mongolian medicine prescription database which has rational structure and exact data. Beside using the classical mining algorithm, some other mining algorithms are used according to the feature of Mongolian medicine prescription, such as the association rule optimization algorithm, local rule mining algorithm, decision tree rule mining algorithm and unrelated dataset rule deduction.These algorithms can find more effective association rules that classical algorithm can not find. This project will provide the software for Mongolian medicine prescription data mining using both Mongolian and Chinese. The mining algorithms are advanced and the software has certain extension and generalization. This project will provide some supports to Mongolian medicine research.

中医药学是中华文化的瑰宝，为人类的健康做出了重大贡献，方剂配伍规律对缩短新药的研发周期具有重要的作用。采用数据挖掘技术发现方剂配伍规律的研究成果很多，但目前没有比较全面和成熟的方剂数据库，使用的挖掘技术和挖掘目标比较传统、单一，不适合日益发展的药物研发要求。 蒙医药学与中医药相比，既有相同之处，又有独特的原理和疗效，蒙医方剂数据挖掘工作目前几乎空白。 项目首先建立结构合理、数据准确的蒙医方剂数据库，使用和研究更加合理的数据预处理方法，实现数据可用性。在使用经典挖掘算法进行数据挖掘的同时，针对蒙医方剂的特点，研究关联规则优化算法、局部规则挖掘算法、决策树规则挖掘算法、不相关数据集间规则演绎等算法，高效率发现常规算法无法发现或很难发现的更多有效关联规则。 项目最终提供蒙汉两种语言的软件，实现蒙医方剂数据挖掘，算法先进，软件具有一定的扩展性和通用性，为蒙药研发提供技术支持。

依托本项目开展研究工作，设计了蒙医方剂数据库，构建了蒙医方剂数据预处理技术，建立了适合蒙医方剂数据挖掘的数据库；给出了一种基于决策树的数据修补算法；给出了一种基于关联规则的条件函数依赖发现及数据修复算法；对大数据环境下相容数据集的关联规则数据挖掘研究，给出了一种实现相容数据集挖掘的关联规则直接合并算法；给出了一种基于自然语言的Apriori关联规则可视化挖掘方法；给出了一种基于精简二元矩阵的蒙医方剂关联规则挖掘算法；实现了基于模糊C均值聚类算法的蒙医方剂类别划分方法；给出了基于WPF的蒙文数据库应用系统解决方案；实现了基于决策树的蒙医方剂药物与主治的关系挖掘。.最后设计了一套蒙医方剂数据挖掘系统，并基于WPF技术实现了平行蒙语版，系统能够建立蒙医方剂数据库，应用关联规则算法、聚类分析算法、决策树算法等对蒙医方剂进行数据挖掘，系统具有通用性和扩展性，可广泛应用。