Abstract: Several session mining on sequential time intervals achieve a rule measure sequence. The estimation of the measure sequence acquires some statistical parameters, which are used to evaluate the interestingness and discover evolutional regularity about the rule. The diffusion estimation method applicable to the sequence with small samples is proposed, based on the principle of information diffusion. The algorithms of parameter estimation are discussed for the sequence in fluctuant or ascend/descend trend, using the measure sequence regarded as incomplete sample. Experiments show the validity, robustness and simplicity for application.

Key words: Data mining, Information diffusion, Statistical estimation

摘要： 对若干连续期间的数据挖掘将形成规则的度量值序列。对度量值序列的参数估计，可获得度量值的基本特征参数，用作评价规则的兴趣度，掌握规则的演化规律。基于信息扩散原理，提出了适用于小样本的度量值扩散估计方法。以度量值序列作为非完备知识样本，讨论在波动和上升(下降)趋势下的序列参数计算。实验表明，该方法准确简便、抗干扰性好。

关键词: 数据挖掘, 信息扩散, 统计估计

CLC Number: 

• TP 311