Search

The most common preprocessing techniques used to deal with datasets having high dimensionality and a low number of instances—or wide data—are feature reduction (FR), feature selection (FS), and resampling. This study explores the use of FR and resampling techniques, expanding the limited comparisons between FR and filter FS methods in the existing literature, especially in the context of wide data. We compare the optimal outcomes from a previous comprehensive study of FS against new experiments conducted using FR methods. Two specific challenges associated with the use of FR are outlined in detail: finding FR methods that are compatible with wide data and the need for a reduction estimator of nonlinear approaches to process out-of-sample data. The experimental study compares 17 techniques, including supervised, unsupervised, linear, and nonlinear approaches, using 7 resampling strategies and 5 classifiers. The results demonstrate which configurations are optimal, according to their performance and computation time. Moreover, the best configuration—namely, k Nearest Neighbor (KNN) + the Maximal Margin Criterion (MMC) feature reducer with no resampling—is shown to outperform state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

This paper studies the effects that combinations of balancing and feature selection techniques have on wide data (many more attributes than instances) when different classifiers are used. For this, an extensive study is done using 14 datasets, 3 balancing strategies, and 7 feature selection algorithms. The evaluation is carried out using 5 classification algorithms, analyzing the results for different percentages of selected features, and establishing the statistical significance using Bayesian tests. Some general conclusions of the study are that it is better to use RUS before the feature selection, while ROS and SMOTE offer better results when applied afterwards. Additionally, specific results are also obtained depending on the classifier used, for example, for Gaussian SVM the best performance is obtained when the feature selection is done with SVM-RFE before balancing the data with RUS. • Wide datasets usually suffer from unbalanced classes distributions. • Feature selection (FS) is commonly recommended for wide datasets. • We aim to find the best combination and order to apply FS and resampling. • 14 datasets, 5 classifiers, 7 FS, and 7 balancing strategies were tested. • The best configuration was SVM-RFE used before RUS for the SVM-G classifier. [ABSTRACT FROM AUTHOR]

Mainstream semi-supervised classification assumes that part of the available data are labelled. Here we assume that, in addition to the labels, we have pairwise constraints on the unlabelled data. Each constraint links two instances, and is one of Must Link (ML, belong to the same class) or Cannot Link (CL, belong to different classes). We propose an approach that uses the labelled data to train a classifier and then applies the ML and CL constraints in subsequent labelling. In our approach, a set of instances are labelled at the same time. Our case study is on animal re-identification. The dataset consists of five free-camera video clips of animals (koi fish, pigeons and pigs), annotated with bounding boxes and animal identities. The proposed approach combines the representations or classifiers predictions from the bounding boxes of consecutive frames. We demonstrate that our approach outperforms standard classifiers, constrained clustering, as well as inductive and transductive semi-supervised learning, using five feature representations. • A method for classification of instances with pairwise constraints is proposed. • Two constraint types: Must Link (ML) and Cannot Link (CL). • Case study: animal re-identification from 5 videos, 5 feature representations. • Constraints are obtained from objects in single and consecutive frames. • Improvements on (semi-)supervised classifiers and constrained clustering. [ABSTRACT FROM AUTHOR]