EXPLAINABLE HYBRID DEEP LEARNING FOR TASK TYPE PREDICTION IN CLOUD COMPUTING
Cloud computing environments execute heterogeneous and dynamic workloads. Accurate task type prediction is required for efficient scheduling, resource allocation, and Quality of Service. Raw performance metrics remain difficult to classify because of noise, class imbalance, and overlap- ping resource patterns. This study proposes an explainable hybrid deep learning model for task type prediction. A three stage pipeline is applied for data preprocessing, model construction, and opti- mization. Feature selection, encoding, normalization, denoising, class balancing, and stratification improve data quality. A convolutional neural network extracts local feature patterns, while a bidi- rectional long short term memory network captures temporal task relationships. Model optimization is achieved through the Adam optimizer and focal loss. Experimental results demonstrate superior classification performance over baseline and recent hybrid models. Ablation analysis validates the preprocessing strategy and model architecture. We show that the proposed method provides reliable workload prediction for cloud computing environments.
Сведения
- Опубликовано
- 2026 · Т. 14, № 2, С. 126-144
- Язык
- Английский
- Цитирований
- 2
- ISSN
- 2306-6172 · 2308-9822
- Издатель
- L. N. Gumilyov Eurasian National University
- Аффилиации
- University of Mohaghegh Ardabili
Obuda University
Farabi University