On the Performance of Oversampling Techniques for Class Imbalance Problems
- 1 Citations
- 2.6k Downloads
Abstract
Although over 90 oversampling approaches have been developed in the imbalance learning domain, most of the empirical study and application work are still based on the “classical” resampling techniques. In this paper, several experiments on 19 benchmark datasets are set up to study the efficiency of six powerful oversampling approaches, including both “classical” and new ones. According to our experimental results, oversampling techniques that consider the minority class distribution (new ones) perform better in most cases and RACOG gives the best performance among the six reviewed approaches. We further validate our conclusion on our real-world inspired vehicle datasets and also find applying oversampling techniques can improve the performance by around 10%. In addition, seven data complexity measures are considered for the initial purpose of investigating the relationship between data complexity measures and the choice of resampling techniques. Although no obvious relationship can be abstracted in our experiments, we find F1v value, a measure for evaluating the overlap which most researchers ignore, has a strong negative correlation with the potential AUC value (after resampling).
Keywords
Class imbalance Minority class distribution Data complexity measures1 Introduction
The classification problem under class imbalance has caught growing attention from both academic and industrial field. Due to recent advances, the progress in technical assets for data storage and management as well as in data science enables practitioners from industry and engineering to collect a large amount of data with the purpose of extracting knowledge and acquire hidden insights. An example may be illustrated from the field of computational design optimization where product parameters are modified to generate digital prototypes which performances are evaluated by numerical simulations, or based on equations expressing human heuristics and preferences. Here, many parameter variations usually result in valid and producible geometries but in the final steps of the optimization, i.e. in the area where the design parameters converge to a local/global optimum, some geometries are generated which violate given constraints. Under this circumstance, a database would contain a large number of designs which are according to specs (even if some may be of low performance) and a smaller number of designs which eventually violate pre-defined product requirements. By far, the resampling techniques have proven to be efficient in handling imbalanced benchmark datasets. However, the empirical study and application work in the imbalanced learning domain are mostly focusing on the “classical” resampling techniques (SMOTE, ADASYN, and MWMOTE etc.) [11, 15, 20], although there are many recently developed resampling techniques.
In this paper, we set up several experiments on 19 benchmark datasets to study the efficiency of six powerful oversampling techniques, including SMOTE, ADASYN, MWMOTE, RACOG, wRACOG and RWO-Sampling. For each dataset, we also calculate seven data complexity measures to investigate the relationship between data complexity measures and the choice of resampling techniques, since researchers have pointed out that studying the data complexity of the imbalanced datasets is of vital importance [15] and it may affect the choice of resampling techniques [20]. We also perform the experiment on our real-world inspired vehicle dataset. Results of our experiments demonstrate that oversampling techniques that consider the minority class distribution (RACOG, wRACOG, RWO-Sampling) perform better in most cases and RACOG gives the best performance among the six reviewed approaches. Results on our real-world inspired vehicle dataset further validate this conclusion. No obvious relationship between data complexity measures and the choice of resampling techniques is found in our experiment. However, we find F1v value, a measure for evaluating the overlap which most researchers ignore [15, 20], has a strong negative correlation with the potential AUC value (after resampling).
The remainder of this paper is organized as follows. In Sect. 2, the research related to our work are presented, also including the relevant background knowledge on six resampling approaches and data complexity measures. In Sect. 3, the experimental setup is introduced in order to understand how the results are generated. Section 4 gives the results of our experiments. Further exploration through data from a real-world inspired digital vehicle model is presented in Sect. 5. Section 6 concludes the paper and outlines further research.
2 Related Works
Many effective oversampling approaches have been developed in the imbalanced learning domain and the synthetic minority oversampling technique (SMOTE) is the most famous one among all. Currently, more than 90 SMOTE extensions have been published in scientific journals and conferences [6]. Most of review paper and application work are based on the “classical” resampling techniques and do not take new resampling techniques into account. In this paper, we briefly review six powerful oversampling approaches, including both “classical” ones (SMOTE, ADASYN, MWMOTE) and new ones (RACOG, wRACOG, RWO-Sampling) [2, 3, 5, 7, 24]. The six reviewed oversampling techniques can be divided into two groups according to whether they consider the overall minority class distribution. Among the six approaches, RACOG, wRACOG, and RWO-Sampling consider the overall minority class distribution while the other three not. Apart from developing new approaches to solve class-imbalance problem, various studies have pointed out that it is important to study the characteristics of the imbalanced dataset [13, 20]. In [13], authors emphasize the importance to study the overlap between the two-class samples. In [20], authors set up several experiments with the KEEL benchmark datasets [1] to study the relationship between various data complexity measures and the potential AUC value. It is also pointed out in [20] that the distinctive inner procedures of oversampling approaches are suitable for particular characteristics of data. Hence, apart from evaluate the efficiency for the six reviewed oversampling approaches, we also aim to investigate the relationship between data complexity measures and the choice of resampling techniques.
2.1 Resampling Technique
In the following, the six established resampling techniques SMOTE, ADASYN, MWMOTE, RACOG, wRACOG and RWO-Sampling are introduced.
SMOTE and ADASYN. The synthetic minority oversampling technique (SMOTE) is the most famous resampling technique [3]. SMOTE produces synthetic minority samples based on the randomly chosen minority samples and their K-nearest neighbors. The new synthetic sample can be generated by using the randomized interpolation scheme above for minority samples. The main improvement in the adaptive synthetic (ADASYN) sampling technique is that the samples which are harder to learn are given higher importance and will be oversampled more often in ADASYN [7].
MWMOTE. The majority weighted minority oversampling techniques (MWMOTE) improves the sample selection scheme and the synthetic sample generation scheme [2]. MWMOTE first finds the informative minority samples (\(S_{imin}\)) by removing the “noise” minority samples and finding the borderline majority samples. Then, every sample in \(S_{imin}\) is given a selection weight (\(S_w\)), according to the distance to the decision boundary, the sparsity of the located minority class cluster and the sparsity of the nearest majority class cluster. These weights are converted in to selection probability (\(S_p\)) in the synthetic sample generation stage. The cluster-based synthetic sample generation process proposed in MWMOTE can be described as, 1). cluster all samples in \(S_{imin}\) into M groups; 2). select a minority sample x from \(S_{imin}\) according to \(S_p\) and randomly select another sample y from the same cluster of x; 3). use the same equation employed in k-NN-based approach to generate the synthetic sample; 4). repeat 1)–3) until the required number of synthetic samples is generated.
RACOG and wRACOG. The oversampling approaches can effectively increase the number of minority class samples and achieve a balanced training dataset for classifiers. However, the oversampling approaches introduced above heavily reply on local information of the minority class samples and do not take the overall distribution of the minority class into account. Hence, the global information of the minority samples cannot be guaranteed. In order to tackle this problem, Das et al. [5] proposed RACOG (RApidy COnverging Gibbs) and wRACOG (Wrapper-based RApidy COnverging Gibbs).
In these two algorithms, the n-dimensional probability distribution of minority class is optimally approximated by Chow-Liu’s dependence tree algorithm and the synthetic samples are generated from the approximated distribution using Gibbs sampling. Instead of running an “exhausting” long Markov chain, the two algorithms produce multiple relatively short Markov chains, each starting with a different minority class sample. RACOG selects the new minority samples from the Gibbs sampler using a predefined lag and this selection procedure does not take the usefulness of the generated samples into account. On the other hand, wRACOG considers the usefulness of the generated samples and selects those samples which have the highest probability of being misclassified by the existing learning model [5].
RWO-Sampling. Inspired by the central limit theorem, Zhang et al. [24] proposed the random walk oversampling (RWO-Sampling) approach to generate the synthetic minority class samples which follows the same distribution as the original training data.
In order to add m synthetic examples to the n original minority examples (\(m < n\)), we first select at random m examples from the minority class and then for each of the selected examples \(\mathbf {x} = (x_1,\ldots , x_m)\) we generate its synthetic counterpart by replacing \(a_i(j)\) (the ith attribute in \(x_j\), \(j \in {1,2,\ldots ,m}\)) with \(\mu _i - r_i \cdot \sigma _{i}/ \sqrt{n}\), where \(\mu _i\) and \(\sigma _i\) denote the mean and the standard deviation of the ith feature restricted to the original minority class, and \(r_i\) is a random value drawn from the standard normal distribution. When \(m > n\), we can repeat the above process until we reach the required amount of synthetic examples. Since the synthetic sample is achieved by randomly walking from one real sample, so this oversampling is called random walk oversampling.
2.2 Data Complexity Measures
Complexity measures information. “Positive” and “Negative” indicate the positive and negative relation between measure value and data complexity respectively.
Measure | Description | Complexity |
---|---|---|
F1 | Maximum fisher’s discriminant ratio | Negative |
F1v | The directional-vector maximum fisher’s discriminant ratio | Negative |
F2 | Volume of overlapping region | Positive |
F3 | Maximum individual feature efficiency | Negative |
L1 | Sum of the error distance by linear programming | Positive |
L2 | Error rate of linear classifier | Positive |
L3 | Non-linearity of a linear classifier | Positive |
Feature Overlapping Measures. F1 measures the highest discriminant ratio among all the features in the dataset [14]. F1v is a complement of F1 and a higher value of F1v indicates there exists a vector that can separate different class samples after these samples are projected on it [19]. F2 calculates the overlap ratio of all features (the width of the overlap interval to the width of the entire interval) and returns the product of the ratios of all features [19]. F3 measures the individual feature efficiency and returns the maximum value among all features.
Linearity Measures. L1 and L2 both measure to what extent the classes can be linearly separated using an SVM with a linear kernel [19], where L1 returns the sum of the distances of the misclassified samples to the linear boundary and L2 returns the error rate of the linear classifier. L3 returns the error rate of an SVM with linear kernel on a test set, where the SVM is trained on training samples and the test set is manually created by performing linear interpolation on the two randomly chosen samples from the same class.
3 Experimental Setup
Information on datasets in 4 groups
Datasets | #Attributes | #Samples | Imbalance Ratio (IR) |
---|---|---|---|
ecoli{1, 2, 3, 4} | 7 | 336 | {3.36, 5.46, 8.6, 15.8} |
glass{0, 1, 2, 4, 5, 6} | 9 | 214 | {2.06, 1.82, 11.59, 15.47, 22.78, 6.38} |
vehicle{0, 1, 2, 3} | 18 | 846 | {3.25, 2.9, 2.88, 2.99} |
yeast{1, 3, 4, 5, 6} | 8 | 1484 | {2.46, 8.1, 28.1, 32.73, 41.4} |
The 19 collected datasets can be simply divided into 4 groups, ecoli, glass, vehicle and yeast (Table 2). IR indicates the imbalance ratio, which is the ratio of the number of majority class samples to the number of minority class samples. In this paper, we aim to study the efficiency of different oversampling approaches and investigate the relationship between data complexity measures and the choice of oversampling techniques. Therefore, we need to calculate the 7 data complexity measures (shown in Table 1) for each dataset. In our 20 experiments for each dataset, we calculate the 7 data complexity measures for every training set (using R package ECoL [14]). Since we use 5 stratified cross-validations, we average each data complexity measures for these 5 training sets and make it the data complexity measure for the dataset.
Confusion matrix for a binary classification problem
Positive prediction | Negative prediction | |
---|---|---|
Positive class | True Positives (TP) | False Negatives (FN) |
Negative class | False Positives (FP) | True Negatives (TN) |
4 Simulation Analysis and Discussions
AUC results for C5.0 decision tree.
Dataset | Baseline | SMOTE | ADASYN | MWMOTE | RACOG | wRACOG | RWO |
---|---|---|---|---|---|---|---|
ecoli1 | 0.9418 | 0.9407 | 0.9364 | 0.9399 | 0.9471 | 0.9390 | 0.9423 |
ecoli2 | 0.8598 | 0.9019 | 0.9140 | 0.9071 | 0.9144 | 0.8959 | 0.9168 |
ecoli3 | 0.7795 | 0.9049 | 0.8943 | 0.8991 | 0.9098 | 0.8757 | 0.9001 |
ecoli4 | 0.8172 | 0.9247 | 0.9080 | 0.9286 | 0.9169 | 0.8875 | 0.9038 |
glass0 | 0.8286 | 0.8476 | 0.8377 | 0.8435 | 0.8442 | 0.8463 | 0.8515 |
glass1 | 0.7082 | 0.7496 | 0.7338 | 0.7489 | 0.7500 | 0.7421 | 0.6986 |
glass2 | 0.7264 | 0.8024 | 0.8072 | 0.7925 | 0.7990 | 0.7862 | 0.7173 |
glass4 | 0.8468 | 0.9291 | 0.9273 | 0.9364 | 0.9255 | 0.8648 | 0.9322 |
glass5 | 0.9905 | 0.9904 | 0.9903 | 0.9905 | 0.9927 | 0.9915 | 0.9916 |
glass6 | 0.9332 | 0.9337 | 0.9310 | 0.9357 | 0.9384 | 0.9378 | 0.9340 |
vehicle0 | 0.9734 | 0.9743 | 0.9725 | 0.9731 | 0.9753 | 0.9751 | 0.9675 |
vehicle1 | 0.7639 | 0.8031 | 0.7974 | 0.7987 | 0.8035 | 0.8094 | 0.7810 |
vehicle2 | 0.9742 | 0.9715 | 0.9741 | 0.9741 | 0.9787 | 0.9777 | 0.9772 |
vehicle3 | 0.7756 | 0.8045 | 0.8006 | 0.8196 | 0.8169 | 0.8166 | 0.7956 |
yeast1 | 0.7317 | 0.7437 | 0.7386 | 0.7449 | 0.7585 | 0.7109 | 0.7166 |
yeast3 | 0.9357 | 0.9584 | 0.9591 | 0.9600 | 0.9647 | 0.9564 | 0.9450 |
yeast4 | 0.7592 | 0.9030 | 0.9001 | 0.8940 | 0.8669 | 0.8245 | 0.8286 |
yeast5 | 0.9574 | 0.9774 | 0.9768 | 0.9782 | 0.9775 | 0.9727 | 0.9782 |
yeast6 | 0.7472 | 0.8760 | 0.8825 | 0.8825 | 0.8802 | 0.8085 | 0.8851 |
According to our experimental results, although the data complexity measures cannot provide guidance for choosing the oversampling approaches, we find there is a strong correlation between the potential best AUC (after oversample) and some of the data complexity measures. From Fig. 1 and Table 5, it can be concluded that the potential best AUC value that can be achieved through oversampling techniques has an extreme negative correlation with the F1v value and linearity measures. In the imbalanced learning domain, there are many researchers focus on studying data complexity measures. In [14], the authors propose that the potential best AUC value after resampling can be predicted through various data complexity measures. In [10], the authors demonstrate that F1 value has an influence on the potential improvement brought by oversampling approaches. However, they did not consider the F1v measure, which has the strongest correlation with AUC value. Hence, we recommend using F1v to evaluate the overlap in imbalanced dataset.
5 Efficient Oversampling Strategies for Improved Vehicle Mesh Quality Classification
In this section, we propose the application of the reviewed methods on the quality prediction of geometric computer aided engineering (CAE) models. In CAE applications, engineers often discretize the simulation domains using meshes (undirected graphs), i.e. a set of nodes (vertices), where the equations that describe the physical phenomena are solved, and edges connecting the nodes to form faces and volumes (elements), where the solution between nodes is approximated. The meshes are generated from an initial geometric representation, e.g. non-uniform rational B-Splines (NURBS) or stereolithography (STL) representations, using numerical algorithms, such as sweep-hull for Delaunay triangulation [23], polycube [12] etc.
Results of hypothesis test.
Measure | Correlation coefficient | P-value | Correlation level |
---|---|---|---|
F1 | −0.4878 | 0.0341 | Medium |
F1v | −0.9048 | \(1.041\times 10^{-7}\) | Extreme |
F2 | 0.1018 | 0.6782 | None |
F3 | −0.7019 | 0.0008 | High |
L1 | −0.8913 | \(3.054\times 10^{-7}\) | Extreme |
L2 | −0.8471 | \(4.735\times 10^{-6}\) | Extreme |
L3 | −0.8693 | \(1.354\times 10^{-6}\) | Extreme |
The continuity of the surfaces is ensured by the mathematical formulation of the FFD up to the order of \(k-1\), where k is the number of planes in the direction of interest, but the mesh quality is not necessarily maintained. The designer can either avoid models with ill-defined elements by applying constraints to the deformations, which might be unintuitive, or eliminate them by performing regular quality assessments. Addressing this issue, we propose the classification of the deformation parameters with respect to the quality of the output meshes, based on a data set of labeled meshes. Further than reducing the risk generating infeasible meshes for CAE applications, our approach avoids unnecessary computation to generate the deformed meshes, which is aligned with the objective of increasing the efficiency of shape optimization tasks.
5.1 Generation of a Synthetic Data Set
Feasible meshes labeling rule.
Dataset | #Attribute | #Sample | #Warnings | Max skewness | Max aspect ratio | IR |
---|---|---|---|---|---|---|
set1 | 9 | 294 | <4 | <6 | <10 | 6.35 |
set2 | 9 | 294 | <4 | <6.2 | <10.5 | 2.54 |
set3 | 9 | 294 | <2 | <5.8 | <10.3 | 12.36 |
5.2 Results and Discussion
Experimental results (AUC) on digital vehicle dataset.
Dataset | Baseline | SMOTE | ADASYN | MWMOTE | RACOG | wRACOG | RWO |
---|---|---|---|---|---|---|---|
set1 | 0.7786 | 0.8412 | 0.8315 | 0.8354 | 0.8543 | 0.8406 | 0.8502 |
set2 | 0.6952 | 0.7575 | 0.7560 | 0.7651 | 0.7614 | 0.7421 | 0.7452 |
set3 | 0.6708 | 0.7780 | 0.7792 | 0.7660 | 0.7823 | 0.7534 | 0.7743 |
6 Conclusion and Future Work
- 1)
In our experiment, in most cases, oversampling approaches which consider the minority class distribution (RACOG, wRACOG and RWO-Sampling) perform better. For both benchmark datasets and our real-world inspired dataset, RACOG performs best and MWMOTE comes to the second.
- 2)
No obvious relationship between data complexity measures and the choice of resampling techniques can be abstracted from our experimental results. However, we find F1v value has a strong correlation with the potential best AUC value (after resampling) while rare researchers in the imbalance learning domain do not consider F1v value for evaluating the overlap between classes.
We only simply apply the oversampling techniques for our digital vehicle dataset and evaluate their efficiency in this paper. In future work, we will focus on adjusting the imbalance learning algorithms to solve the proposed engineering problem. Additionally, the effect of the interaction between various data complexity measures on the choice of resampling technique will be studied.
Footnotes
References
- 1.Alcalá-Fdez, J., et al.: KEEL data-mining software tool: data set repository, integration of algorithms and experimental analysis framework. J. Multiple-Valued Logic Soft Comput. 17(2), 255–287 (2011)Google Scholar
- 2.Barua, S., Islam, M.M., Yao, X., Murase, K.: MWMOTE-majority weighted minority oversampling technique for imbalanced data set learning. IEEE Trans. Knowl. Data Eng. 26(2), 405–425 (2012)CrossRefGoogle Scholar
- 3.Chawla, N.V., Bowyer, K.W., Hall, L.O., Kegelmeyer, W.P.: SMOTE: synthetic minority over-sampling technique. J. Artif. Intell. Res. 16, 321–357 (2002)CrossRefGoogle Scholar
- 4.Cordón, I., García, S., Fernández, A., Herrera, F.: Imbalance: oversampling algorithms for imbalanced classification in R. Knowl.-Based Syst. 161, 329–341 (2018)CrossRefGoogle Scholar
- 5.Das, B., Krishnan, N.C., Cook, D.J.: RACOG and wRACOG: two probabilistic oversampling techniques. IEEE Trans. Knowl. Data Eng. 27(1), 222–234 (2014)CrossRefGoogle Scholar
- 6.Fernández, A., García, S., Galar, M., Prati, R.C., Krawczyk, B., Herrera, F.: Learning from Imbalanced Data Sets. Springer, Cham (2018). https://doi.org/10.1007/978-3-319-98074-4CrossRefGoogle Scholar
- 7.He, H., Bai, Y., Garcia, E.A., Li, S.: ADASYN: adaptive synthetic sampling approach for imbalanced learning. In: 2008 IEEE International Joint Conference on Neural Networks (IEEE World Congress on Computational Intelligence), pp. 1322–1328. IEEE (2008)Google Scholar
- 8.Heft, A.I., Indinger, T., Adams, N.A.: Experimental and numerical investigation of the DrivAer model. In: ASME 2012 Fluids Engineering Division Summer Meeting, pp. 41–51. American Society of Mechanical Engineers Digital Collection (2012)Google Scholar
- 9.Knupp, P.: Measurement and impact of mesh quality. In: 46th AIAA Aerospace Sciences Meeting and Exhibit, p. 933 (2008)Google Scholar
- 10.Kong, J., Kowalczyk, W., Nguyen, D.A., Menzel, S., Bäck, T.: Hyperparameter optimisation for improving classification under class imbalance. In: 2019 IEEE Symposium Series on Computational Intelligence (SSCI). IEEE (2019)Google Scholar
- 11.Li, J., et al.: Adaptive swarm balancing algorithms for rare-event prediction in imbalanced healthcare data. PLoS ONE 12(7), e0180830 (2017)CrossRefGoogle Scholar
- 12.Livesu, M., Vining, N., Sheffer, A., Gregson, J., Scateni, R.: PolyCut: monotone graph-cuts for PolyCube base-complex construction. Trans. Graph. 32(6), 171:1–171:12 (2013). (Proc. SIGGRAPH ASIA 2013)CrossRefGoogle Scholar
- 13.López, V., Fernández, A., García, S., Palade, V., Herrera, F.: An insight into classification with imbalanced data: empirical results and current trends on using data intrinsic characteristics. Inf. Sci. 250, 113–141 (2013)CrossRefGoogle Scholar
- 14.Lorena, A.C., Garcia, L.P., Lehmann, J., Souto, M.C., Ho, T.K.: How complex is your classification problem? A survey on measuring classification complexity. ACM Comput. Surv. (CSUR) 52(5), 107 (2019)CrossRefGoogle Scholar
- 15.Luengo, J., Fernández, A., García, S., Herrera, F.: Addressing data complexity for imbalanced data sets: analysis of SMOTE-based oversampling and evolutionary undersampling. Soft. Comput. 15(10), 1909–1936 (2011). https://doi.org/10.1007/s00500-010-0625-8CrossRefGoogle Scholar
- 16.Menzel, S., Olhofer, M., Sendhoff, B.: Application of free form deformation techniques in evolutionary design optimisation. In: Herskovits, J., Mazorche, S., Canelas, A. (eds.) 6th World Congress on Structural and Multidisciplinary Optimization (WCSM 2006). COPPE Publication, Rio de Janeiro (2005)Google Scholar
- 17.Menzel, S., Sendhoff, B.: Representing the change - free form deformation for evolutionary design optimization. In: Yu, T., Davis, L., Baydar, C., Roy, R. (eds.) Evolutionary Computation in Practice. SCI, vol. 88, pp. 63–86. Springer, Heidelberg (2008). https://doi.org/10.1007/978-3-540-75771-9_4CrossRefGoogle Scholar
- 18.Olhofer, M., Bihrer, T., Menzel, S., Fischer, M., Sendhoff, B.: Evolutionary optimisation of an exhaust flow element with free form deformation. In: 4th European Automotive Simulation Conference, Munich (2009)Google Scholar
- 19.Orriols-Puig, A., Macia, N., Ho, T.K.: Documentation for the data complexity library in c++, vol. 196, pp. 1–40. Universitat Ramon Llull, La Salle (2010)Google Scholar
- 20.Santos, M.S., Soares, J.P., Abreu, P.H., Araujo, H., Santos, J.: Cross-validation for imbalanced datasets: avoiding overoptimistic and overfitting approaches [research frontier]. IEEE Comput. Intell. Mag. 13(4), 59–76 (2018)CrossRefGoogle Scholar
- 21.Sederberg, T.W., Parry, S.R.: Free-form deformation of solid geometric models. ACM SIGGRAPH Comput. Graph. 20(4), 151–160 (1986)CrossRefGoogle Scholar
- 22.Sieger, D., Menzel, S., Botsch, M.: On shape deformation techniques for simulation-based design optimization. In: Perotto, S., Formaggia, L. (eds.) New Challenges in Grid Generation and Adaptivity for Scientific Computing. SSSS, vol. 5, pp. 281–303. Springer, Cham (2015). https://doi.org/10.1007/978-3-319-06053-8_14CrossRefzbMATHGoogle Scholar
- 23.Sinclair, D.: S-hull: a fast radial sweep-hull routine for Delaunay triangulation. arXiv preprint arXiv:1604.01428v1 [cs.CG] (2016)
- 24.Zhang, H., Li, M.: RWO-sampling: a random walk over-sampling approach to imbalanced data classification. Inf. Fusion 20, 99–116 (2014)CrossRefGoogle Scholar