Skip to main content

Table 3 Types of methods addressing error detection and quantification (RQ2) only, along with its addressed errors, respective papers, and the total number of papers that proposed that method

From: Sensor data quality: a systematic review

MethodErrors addressedPapersTotal
Principal component analysisOutliers, bias, drift, stuck-at-zero[27, 32, 47, 48, 50, 59, 69]7
Artificial neural networkOutliers, bias, drift, constant values, noise, stuck-at-zero, uncertainty[34, 36, 54, 70, 78, 81]6
Ensemble classifiersOutliers, drift, constant values, noise, uncertainty[33, 35, 37, 79]4
Support vector machineOutliers[57, 58]2
ClusteringOutliers[39, 45]2
Ontology/knowledge-based systemsUncertainty (inaccurate data), missing data (incomplete data)[25, 26]2
Univariate autoregressive modelsOutliers[40]1
Statistical generative modelsOutliers[49]1
Grey prediction modelOutliers, noise, constant values[52]1
Particle filteringBias, scaling[71]1
Association rule miningOutliers[56]1
Bayesian networkOutliers, noise[44]1
Euclidean distanceOutliers[42]1
Hybrid methods
 Polynomial predictive filter and fuzzy rulesOutliers[53]1
 Dempster–Shafer theory and mathematical modellingDrift, noise[75]1