文章目录
| Tittle | Algorithm | Idea |
|---|---|---|
| classical methods | ||
| An adaptive prediction approach based on workload pattern discrimination in the cloud | An adaptive approach | categorizes the workloads into different classes which are automatically assigned for different models according to workload features |
| TODO | Exponential Smoothing(ES) | |
| A Hierarchical Framework for Modeling and Forecasting Web Server Workload | Auto Regression(AR) | A linear combination of past values of the variable under consideration is used to forecast the value for upcoming time instances. |
| Dual time-scale distributed capacity allocation and load redirect algorithms for cloud systems | Moving Average(MA) | non-linear optimization The model is appropriate for time series exhbiting seasonal behavior only |
| Efficient autoscaling in the cloud using predictive models for workload forecasting | Autoregressive Integrated Moving Average(ARIMA) | also discussed the challenges involved in auto scaling in a cloud environment |
| Workload Prediction Using ARIMA Model and Its Impact on Cloud Applications’ QoS | Autoregressive Integrated Moving Average(ARIMA) | ARIMA on different confidence interval to predict web server workload |
| Workload characterization and prediction in the cloud: A multiple time series approach | Hidden Markov Model | distinguish the temporal correlations in obtained clusters of VMs |
| A workload analysis of live event broadcast service in cloud | Regression techniques | the approach is based on simple statistical models that might not capture the patterns in more complex data |
| Workload characterization and prediction in the cloud: A multiple time series approach | multiple time series approch | The model does a grouping of similar applications's need in order to improve the accuracy of predictions |
| machine learning methods | ||
| Support vector machines experts for time series forecasting | Self organizing map(SOM) and support vector machines(SVMs) | Self organizing map was used to cluster the data in different regions while SVMs were used to predict the future data |
| Referential kNN Regression for Financial Time Series Forecasting | \(k\) Nearest Neighbors(kNN) | For financial time series prediction. kNNs are lazy learners and need high computational cost |
| Hierarchical neural networks based prediction and control of dynamic reconfiguration for multilevel embedded systems | the Neural network | used to model workload variations in multimidia designs |
| A cost-aware auto-scaling approach using the workload prediction in service clouds | Linear regression | the predicted workload was used to decide the type of scaling |
| Efficient resources provisioning based on load forecasting in cloud | Support Vector Regression(SVR) and Kalman smoother | It achieved high prediction accuracy |
| Combining time series prediction models using genetic algorithm to autoscaling Web applications hosted in the cloud infrastructure | Ensemble based model | It uses five different base prediction models. Each model is assigned a weight and contributes accordingly in predictions. The weight are assigned and optimized using genetic algorithm |
| Empirical prediction models for adaptive resource provisioning in the cloud | Neural network and Linear regression | |
| RVLBPNN: A Workload Forecasting Model for Smart Cloud Computing | back propagation learning algorithm | It adjusts the weights of model according to error trend. |
| A Predictive Method for Workload Forecasting in the Cloud Environment | neural network and steepest descent learning algorithm | suffers from high prediction errors. |
| Workload prediction in cloud using artificial neural network and adaptive differential evolution | neural network and adaptive differential evolution |