准确高效的动态干预决策发现

Accurate and Efficient Dynamic Interventions Discovery

干预决策是数据挖掘领域关注的重要问题,致力于评价干预措施对干预目标的影响或发现满足干预目标的最优干预措施,而朴素干预规则模型简单,无法精确表达干预知识,且效率较差。在模型设计中引入了马尔科夫链,提出了干预过程动态模型,设计并实现了基于干预力度的动态精确干预评价体系。在中国出生缺陷数据集上的实验表明,该方法可比较精确地发现干预规则。

Intervention decision is the hot topic concerned by data mining fields, which wants to evaluate the influence of intervention methods to intervention targets or discover the most effective method to achieve intervention targets, but naive intervention rule （NIR） does not express interventional knowledge accurately and efficiently due to its simplicity. This paper introduces the idea of Markov chains in intervention model which substantially improves the evaluation accuracy and accelerates the mining process, proposes a dynamic model of intervention process, and designs and realizes an accurate evaluating system of intervention process based on intervention intensity. Experimental results in the dataset of Chinese Birth Defects show the effectiveness of the proposed method.