基于深度森林的量表数据挖掘方法

Deep forest based inventory data mining method

在数据挖掘领域中,量表是间接获取样本属性数据的重要工具。针对量表数据离散、稀疏、二值化的特点,导致其难以进行分析挖掘的问题。文中采用了基于深度森林的量表数据挖掘方法,实验分别对老年健康综合评估数据库中的两个量表进行对比分析。实验结果表明,在所选取的两个量表中,提取到的关键属性数量相比于原始量表属性数量分别下降了30%和40%,且比基线模型下降了16%和18%。同时,提出的方法可在保证分类性能基本不变的情况下,进一步降低提取到的关键属性数量。

In data mining,inventory is one of the important equipment for acquiring data.However,the crucial attribute analysis often fails on inventory data due to the discrete,sparse and binary charac teristics of inventory.In this paper,we propose a deep forest based inventory data mining method.The experiments are carried out on two inventory datasets from the elderly health comprehensive evaluating database.The results show that in the selected two datasets,the extracted crucial attributes are reduced by 30%and 40%respectively compared with the original ones.Moreover,the number of crucial attributes extracted by proposed model is reduced by 16%and 18%compared with baselines.Besides,it also indicates that the proposed model can further reduce the number of crucial attributes while maintaining the performance of classification.