基于动态规划方法优化关联规则发现

为了得到准确可信任的关联规则,将关联规则的发现归纳为多阶段决策问题,利用动态规划方法对关联规则发现进行优化分析.通过条件概率分析,计算出了动态规划状态转移方程和最优期望代价方程,并得到了关联规则发现的决策策略.该策略不需要每一步计算条件概率,其实现平稳方便.最后给出了一个应用例子,并通过模拟实验将该方法与增量关联规则挖掘进行了比较分析,实验结果证明了该方法的有效性.

Optimal Strategy for Association Rules Mining Based on Dynamic Programming

In order to improve the trustiness and accuracy of association rules mining,the association rules mining is reduced to the multi-step decision process,and an optimal strategy is proposed based on dynamic programming.The equation of state and optimal value function used to achieve the optimal strategy is figured out through the analysis of conditional probability of the process. The strategy is realized placidly and easily,without the computation of conditional probability in each step.The experimental results show that the mining based on dynamic programming is superior to traditional incremental updating algorithms for mining association rules.