Publication Date: 2023/06/07
Abstract: The development of real-world databases presents computing difficulties for a single computer. Cloud-based systems, on the other hand, can handle massive quantities of data management activities for large-scale real-world data set calculations. The study focuses on a new Generalized Flow inside the cloud computing platform, Microsoft Azure Machine Learning Studio (MAMLS), which analyses multi-class and binary classification data sets to maximise overall classification accuracy. To begin, each data set is split into training and testing sets. Following that, the training data is utilised to create classification model parameters. Reduce the dimensionality of your data to enhance classification accuracy. Data-centered information increases overall classification accuracy by reducing multi-class classification to a series of hierarchical binary classification problems. Finally, the performance of the improved classification model is tested and appraised. The proposed study assessed algorithm performance utilising 82,332 test samples from a recent data set, UNSW NB-15. It took 6 seconds to train 1,75,341 network instances using the suggested two-class forest decision model. At 99 percent, 94.49 percent, 91.79 percent, and 90.9 percent, the multi-level forest decision- making model recognised attack types such as generics, feats, shellcodes, and worms, respectively.
Keywords: Azure Machine Learning; Decision Forest; Intrusion Detection; Locally Deep SVM; Mutual Information; UNSW NB-15.
DOI: https://doi.org/10.5281/zenodo.8012114
PDF: https://ijirst.demo4.arinfotech.co/assets/upload/files/IJISRT23JUN207.pdf
REFERENCES