A robust missing value imputation method for noisy data

Abstract

Missing data imputation is an important research topic in data mining. The impact of noise is seldom considered in previous works while real-world data often contain much noise. In this paper, we systematically investigate the impact of noise on imputation methods and propose a new imputation approach by introducing the mechanism of Group Method of Data Handling (GMDH) to deal with incomplete data with noise. The performance of four commonly used imputation methods is compared with ours, called RIBG (robust imputation based on GMDH), on nine benchmark datasets. The experimental result demonstrates that noise has a great impact on the effectiveness of imputation techniques and our method RIBG is more robust to noise than the other four imputation methods used as benchmark.

An approach for multimodal biometric fusion under the missing data scenario

“…As it can be seen, the RIBG estimates the missing values of individual matchers based on the behaviour of other matchers [2]. Future work will focus on addressing the issue of insufficient number of matchers in the multimodal identification system.…”

“…Missing information may be in the form of missing modalities in the template or the query or incomplete score information from individual matchers. Data reduction methods, which delete all incomplete vectors, are not suitable in such cases [2]. Imputation methods, on the other hand, which substitute the missing scores with predicted values are a better solution as (a) they do not delete any of the score vectors, which may contain useful information for identification, and (b) their application could be followed by a standard score fusion scheme.…”

Using Density based Score Fusion for Multimodal Identification Systems under the Missing Data Scenario

“…The RIBG algorithm [2] is a recently proposed imputation approach. It is based on the Group Method of Data Handling (GMDH), which is applied in a great variety of areas for data mining and knowledge discovery, forecasting and systems modeling, optimization and pattern recognition.…”

“…Comparative studies on 9 benchmark dataset from the UCI (University of California at Irvine) have shown that the RIBG algorithm outperforms other imputation techniques, including regression imputation, EM imputation, grey-based nearest neighbor imputation and multiple-imputation based on fully Not selected candidate models Selected candidate models conditional specification. We refer interested readers to [2] for a detailed description of the RIBG algorithm.…”

“…Towards this end, we have made the following two contributions. Firstly, we present a RIBG algorithm (Robust Imputation Based on Group Method of Data Handling) [2], which proved to be an efficient imputation technique in many data mining applications. Secondly, we propose a density based score fusion, where the distributions of genuine and impostor match scores are modeled as a finite Gaussian mixture model to achieve optimal performance of biometric systems at any desired operating point.…”