Özet:

Abstract The rapid advancement of contemporary technology and smart systems has resulted in a large influx of big data. A phenomenon known as the class imbalance problem limits learning from many real-world datasets. When one class (the majority class) contains disproportionately more instances than the other class, the dataset is unbalanced (the minority class). Because of these datasets, traditional machine learning algorithms struggle to perform well on classification tasks. To compensate for the imbalance, NPC employs an innovative hybrid machine learning approach for grading the training samples.Both local and global data are used to generate the grades. The contribution of this article is a totally new classifier for efficiently dealing with the imbalance issue without the requirement for manually-set parameters or expert knowledge. To address this problem, in this research a novel approach Hybrid Support Vector Machine is designed by incorporating three major steps like pre-processing, dimension reduction and classification. Initially, the pre-processing phase is enabled by the data normalization process. The extensive sets of features are reduced using dimension reduction process and are achieved by using Quantum Theory-based Particle Swarm Optimization (QPSO). With this technique, a better solution can be obtained for classifying the big data; therefore, the existing problems related to accuracy metrics can resolved. Finally, a hybrid optimized support vector machine technique is proposed to accomplish the big data classification task. The suggested technique is compared to sample algorithms on unbalanced datasets in order to demonstrate the algorithm's efficacy.

Anahtar Kelimeler: