An imputation method based on crow search algorithm for medical data
The absence of medical data can lead to a decrease in statistical power,which in turn severely affects the accuracy of diagnoses and may even result in misdiagnoses.Consequently,selecting effective imputation methods for missing data in medical issues is of great importance.In order to efficiently impute missing medical data and enhance the effectiveness of medical data mining,this paper proposes a medical data imputation method based on crow search algorithm(CSA).The data imputation model is designed,and the mapping between the individual encoding of the algorithm and the data imputation model is established.Subsequently,the CSA is applied to iteratively search the optimal imputation model,and the optimal imputation model is used to construct a complete medical dataset.Comparative experiments are conducted on four medical datasets with two traditional imputation methods-mean imputation(MI)and K nearest neighbor imputation(KNNI).Datasets with different missing rates are constructed artificially,and each imputation method is applied to the missing datasets.The accuracy of classification algorithms on the imputed datasets is used as an evaluation metric for the imputation methods.The results show that,compared to MI,the proposed method improves the average accuracy of classification algorithms on the four datasets by 3.7%,3.8%,11.1%,and 17.7%,respectively.Compared to KNNI,the proposed method increases the average accuracy of classification algorithms by 4%,14.8%,12.6%,and 21.7%,respectively.These findings indicate that the imputation method based on the CSA proposed in this paper can effectively complete the imputation of missing data and enhance the performance of data mining algorithms.
NETL
NSTL
万方数据
