The repository is the official implementation for the paper: Weiping Zheng, Zongxiao Chen, Kaiyuan Zheng, Weijian Zheng, Yiqi Chen, Xiaomao Fan, "WorkloadDiff: Conditional Denoising Diffusion Probabilistic Models for Cloud Workload Prediction", IEEE Transactions on Cloud Computing.
We have devised a two-path neural network tailored for cloud workload prediction. This network consists of four residual layers, with bidirectional dilated convolution (Bi-DilConv) serving as the fundamental building block.
Dependencies can be installed using the following command:
pip install -r requirements
Training and forecasting for the workload dataset.
python exe_main.py --nsample [number of samples]
Execute multiple times and calculate the average.
sh exe_WorkloadDiff.sh
Make predictions using a pre-trained model.
python exe_main.py --modelfolder [model file path]
We are grateful for the valuable codes in the following GitHub repositories, which have been instrumental in achieving our baseline results.
- DeepAR: https://github.com/arrigonialberto86/deepar
- CSDI: https://github.com/ermongroup/CSDI/tree/main
- N-BEATS: https://github.com/philipperemy/n-beats/tree/master
- SARIMA: https://github.com/krishnaik06/ARIMA-And-Seasonal-ARIMA/tree/master
- LSTNet: https://github.com/laiguokun/LSTNet/tree/master
- WaveNet: https://github.com/ZhouYuxuanYX/Wavenet-in-Keras-for-Kaggle-Competition-Web-Traffic-Time-Series-Forecasting/tree/master
- LSTM: We implement an LSTM model by ourselves. The LSTM model comprises a single LSTM layer with 64 units, a Fully-Connected layer with ReLU activation function, and a dropout ratio of 0.3.
- CNN-LSTM: We implement the model by ourselves. The model consists of a one-dimensional CNN with 32 filters and a kernel size of 3, followed by max-pooling with a size of 2. The LSTM module includes a layer with 64 units, a Fully-Connected layer, a dropout ratio of 0.3, and ReLU activation function.
- CEEMDAN-Informer: We utilized the authors' official implementation and requested the source code via email.
@article{zheng2024workloaddiff,
title={WorkloadDiff: Conditional Denoising Diffusion Probabilistic Models for Cloud Workload Prediction},
author={Zheng, Weiping and Chen, Zongxiao and Zheng, Kaiyuan and Zheng, Weijian and Chen, Yiqi and Fan, Xiaomao},
journal={IEEE Transactions on Cloud Computing},
year={2024},
publisher={IEEE}
}