Metodología De Evaluación Del Desempeño De Métodos De Imputación Mediante Una Métrica Tradicional Complementada Con Un Nuevo Indicador (original) (raw)

2020, European Scientific Journal ESJ

Missing Values (MV), values not observed in the dataset, constitute a common obstacle faced by researchers in real-world contexts. Data imputation techniques allow estimating them using different algorithms, through which an important characteristic can be imputed to a particular instance. Most of the articles published in this field deal with new imputation methods, however, few studies address the evaluation of existing methods in order to provide more appropriate guidelines for imputation of data. The objective of this work is to show a methodology for evaluating the performance of imputation methods using a traditional metric complemented with a new indicator, based on the normalized average of the Root Mean Squared Error (RMSE). From a complete data set, 63 data sets were generated with MV. These were imputed using the methods of imputation by means, k-NN, k-Means and hot-deck. The performance of the imputation methods was evaluated using the traditional metric complemented with a new proposed indicator. The results show that the error for the k-Means imputation method is the lowest considering all data sets. The work environment developed to perform the amputation and subsequent imputation experiments was appropriate and allows the incorporation of other amputation mechanisms and other imputation methods in the future, being an essential part of the proposed methodology.

Sign up for access to the world's latest research.

checkGet notified about relevant papers

checkSave papers to use in your research

checkJoin the discussion with peers

checkTrack your impact