Efficient resampling methods for training support vector machines with imbalanced datasets

Batuwita, Rukshan; Palade, Vasile

doi:10.1109/ijcnn.2010.5596787

Cited by 85 publications

(51 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One common approach in literature to overcome this challenge is to undersample the majority class to balance the class distribution. 49,50 Since a random undersampling approach could discard potentially useful data and cause further problems, we decided to employ a stratified sampling scheme based on k-medoids clustering 51 to attain a positive/negative class ratio of 1:1. As mentioned earlier, we used half of the positive and negative examples for training/validation and reserved the other half of the data for testing.…”

Section: Resultsmentioning

confidence: 99%

Combining Human Computing and Machine Learning to Make Sense of Big (Aerial) Data for Disaster Response

et al. 2016

View full text Add to dashboard Cite

Aerial imagery captured via unmanned aerial vehicles (UAVs) is playing an increasingly important role in disaster response. Unlike satellite imagery, aerial imagery can be captured and processed within hours rather than days. In addition, the spatial resolution of aerial imagery is an order of magnitude higher than the imagery produced by the most sophisticated commercial satellites today. Both the United States Federal Emergency Management Agency (FEMA) and the European Commission's Joint Research Center ( JRC) have noted that aerial imagery will inevitably present a big data challenge. The purpose of this article is to get ahead of this future challenge by proposing a hybrid crowdsourcing and real-time machine learning solution to rapidly process large volumes of aerial data for disaster response in a time-sensitive manner. Crowdsourcing can be used to annotate features of interest in aerial images (such as damaged shelters and roads blocked by debris). These human-annotated features can then be used to train a supervised machine learning system to learn to recognize such features in new unseen images. In this article, we describe how this hybrid solution for image analysis can be implemented as a module (i.e., Aerial Clicker) to extend an existing platform called Artificial Intelligence for Disaster Response (AIDR), which has already been deployed to classify microblog messages during disasters using its Text Clicker module and in response to Cyclone Pam, a category 5 cyclone that devastated Vanuatu in March 2015. The hybrid solution we present can be applied to both aerial and satellite imagery and has applications beyond disaster response such as wildlife protection, human rights, and archeological exploration. As a proof of concept, we recently piloted this solution using very high-resolution aerial photographs of a wildlife reserve in Namibia to support rangers with their wildlife conservation efforts (SAVMAP project, http://lasig.epfl.ch/savmap). The results suggest that the platform we have developed to combine crowdsourcing and machine learning to make sense of large volumes of aerial images can be used for disaster response.

show abstract

Section: Resultsmentioning

confidence: 99%

Combining Human Computing and Machine Learning to Make Sense of Big (Aerial) Data for Disaster Response

et al. 2016

View full text Add to dashboard Cite

show abstract

“…The goal of our work was not to obtain the best possible results for splice site classification, which has already been successfully addressed by Sonnenburg et al (2007) using SVM and specialised kernels, but rather to explore semi-supervised learning as a possible solution for splice site prediction, and to study the effects of imbalanced distributions on semi-supervised learning algorithms. Batuwita and Palade applied SVM (Batuwita and Palade, 2010) with re-sampling methods on four imbalanced biological datasets and the Pageblocks dataset from the UCI, with up to 8K instances and no more than 1-to-50 imbalance degree. They proposed to first identify the most informative negative instances, and then randomly over-sample the positive instances in order to reach the same number of negative instances selected.…”

Section: Related Workmentioning

confidence: 99%

An empirical study of self-training and data balancing techniques for splice site prediction

Stănescu

Caragea

2017

IJBRA

View full text Add to dashboard Cite

Thanks to Next Generation Sequencing technologies, unlabelled data is now generated easily, while the annotation process remains expensive. Semi-supervised learning represents a cost-effective alternative to supervised learning, as it can improve supervised classifiers by making use of unlabelled data. However, semi-supervised learning has not been studied much for problems with highly skewed class distributions, which are prevalent in bioinformatics. To address this limitation, we carry out a study of a semi-supervised learning algorithm, specifically self-training based on Naïve Bayes, with focus on data-level approaches for handling imbalanced class distributions. Our study is conducted on the problem of predicting splice sites and it is based on datasets for which the ratio of positive to negative examples is 1-to-99. Our results show that under certain conditions semi-supervised learning algorithms are a better choice than purely supervised classification algorithms.

show abstract

“…Multi-category data clustering suffered from multiple feature evaluation, Selection of number of cluster for multi-level [1,3,6] Diversity of feature selection process. [12] Boundary value of cluster [9,13] V.…”

Section: IVmentioning

confidence: 99%

A Review of Clustering Technique Based on Different Optimization Function Using for Selection of Center Point

Firke¹

2017

IJRASET

View full text Add to dashboard Cite

The selection of center point is major issue in clustering technique. the center point of cluster decides the quality and validation of clustering technique. for the better selection of clustering technique used different optimization function such as genetic algorithm, particle swarm optimization algorithm and many more algorithm used for center selection process. In this paper present the review of clustering technique for automatic validation and cluster center selection. The process of clustering basically group the data based on feature attribute of data. the selection of features attribute of data based on the process of iteration. Keywords: -Clustering, EA, K-Means, SVM. I.INTRODUCTION Bunching is the critical stride for some errands in machine learning [5]. Each algorithmic run has its own inclination owing to the upgrades of different criteria. Unsupervised machine learning is naturally an advancement assignment; one is attempting to fit the best model to a specimen of information. The meaning of "best" is unlimited; speculation show with significance to the full universe of information focuses. However machine learning calculations don't comprehend this from the earlier, and instead of depend on heuristic estimations considering the norms of their copy and confinement, for example, integrity of fit with significance to the analysis raw numbers, demonstrate niggardliness, thus on [5]. Undertaking conclusions were custom fitted at last by amplifying/limiting an objective or favorable position. The measurements and unpredictability of change issues that can be clarified in a sensible time has been progressed by the approach of exceptional processing advancements.Cluster by nature are the collection of similar objects. Each group or cluster is homogeneous, i.e., objects belonging to the same group are similar to each other. Also, each group or cluster should be different from other clusters, i.e., objects belonging to one cluster should be different from the objects of other clusters. In order to increase the classification performance for imbalance data streams, many approaches are proposed by improving traditional classification algorithms, for example, the cost-sensitive learning, the resample, the improved SVM, etc. The cost-sensitive learning takes a full consideration for the performance of the minority classification, and can resolve effectively the imbalanced classification in real life. Adel proposed an online ensemble of neural network classifiers for nonstationary and imbalanced data streams [1].Clustering is the process of grouping similar objects, and this could be hard or fuzzy. In hard clustering algorithm, each element is allocated to a single cluster during its operation; however, in fuzzy clustering method, a degree of membership is assigned to each element depending on its degree of association to several other clusters. Clustering problem for unsupervised data exploration and analysis has been investigated for decades in the statistics, image retrieval, bioinformatics, data mi...

show abstract

Efficient resampling methods for training support vector machines with imbalanced datasets

Cited by 85 publications

References 11 publications

Combining Human Computing and Machine Learning to Make Sense of Big (Aerial) Data for Disaster Response

Combining Human Computing and Machine Learning to Make Sense of Big (Aerial) Data for Disaster Response

An empirical study of self-training and data balancing techniques for splice site prediction

A Review of Clustering Technique Based on Different Optimization Function Using for Selection of Center Point

Contact Info

Product

Resources

About