Abstract
AbstractIn machine learning, hyperparameter tuning is strongly useful to improve model performance. In our research, we concentrate our attention on classifying imbalanced data by cost-sensitive support vector machines. We propose a multi-objective approach that optimizes model’s hyper-parameters. The approach is devised for imbalanced data. Three SVM model’s performance measures are optimized. We present the algorithm in a basic version based on genetic algorithms, and as an improved version based on genetic algorithms combined with decision trees. We tested the basic and the improved approach on benchmark datasets either as serial and parallel version. The improved version strongly reduces the computational time needed for finding optimized hyper-parameters. The results empirically show that suitable evaluation measures should be used in assessing the classification performance of classification models with imbalanced data.
Highlights
Classification problems may be encountered in different domains
9: end while 10: return similarit y in the literature with our results. They are related to medical diagnosis represented as binary classification problems and have different sample sizes, attributes, and imbalance ratio (IR), defined as m/M (Amin et al 2016), where m is the number of the minority instances and M is the number of majority instances
As in other machine learning (ML) techniques, their performance depends on hyperparameters
Summary
Classification problems may be encountered in different domains. One of these is the disease diagnosis, which establishes the presence or absence of a given disease according to referred symptoms and results of medical exams. We tested in (Guido et al 2021) two evaluation model metrics, i.e., accuracy and G-Mean, on two imbalanced benchmark datasets by optimizing hyper-parameters of support vector machines by genetic algorithms (GAs). They performed experimental analysis on class imbalance, cost-sensitive learning with a given class and example costs and showed that their proposed algorithm provides superior generalization performance compared to conventional methods. Qi et al (2013) proposed a new Cost-Sensitive Laplacian SVM and tested its effectiveness via experiments on public datasets They evaluate the algorithms performance by the Average Cost. Noia et al (2020) applied SVM, k-Nearest Neighbors and k-means as clustering techniques to predict the probability of contracting a given disease starting from both workplace-related (using Ateco and Istat codes) and workerrelated characteristics (i.e., age at hiring, age at disease certification, gender, employment duration) They used a GA to find the best values of the used methods. The most used evaluation measures are accuracy, precision, recall, F-score, and the Receiver Operating Characteristic (ROC)
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
