一般而言，雲端資源是昂貴的，雲端平台資源的使用量是不固定的。當管理者啟用雲端資源時，其已啟用而未利用到的資源將會被浪費，或是當所啟用的資源少於雲端平台之需求時，雲端平台的穩定度因而受到影響。因此，準確的預估雲端資源的使用量以符合使用者之需求，對於一個應用系統而言是重要的。
本論文提出了一個混和預測模型機制來預測雲端平台的資源使用量。該模型是混和類神經網路與遺傳基因演算法。此模型主要分成三個步驟：第一，運用遺傳基因演算法去做歷史資料的分析，找出歷史資料中的隱藏樣式類型(pattern type)；其次，運用已找出的隱藏樣式類型將歷史資料進行分類，並利用正確樣式(current pattern)找出候選類別；最後，我們將找出的候選類別當成訓練資料並對預測模型進行訓練以獲得參數。當訓練完成獲得所需參數後即可進行預測。
為了驗證所提出的預測模型的有用性，我們利用多種可能具有樣式的資料來進行測試，並且與類神經網路、統計學上的子數平滑法(exponential smoothing)及 Yang et al. 所提出的樣式融合模型(pattern fusion model)這些方法做比較。實驗顯示，論文所提出的混和預測模型在準確度、適應性與穩定度都有較佳的表現。

In general, the public cloud resource is a pay-per-use model and the resource usage of the cloud computing platform is unstable. If the platform’s managers rent the inappropriate resources usage, some of the resources usage will be wasted. Therefore, predicting the correct resources usage of the cloud platform for an application is important.
In this thesis, neural network combined with genetic algorithm is proposed to predict the cloud platform’s resources usage. The hybrid neural network/genetic algorithm model consist of three steps. First, the historical data is analyzed and the potential characteristic pattern type will be found by a simple genetic algorithm. Next, the historical data is classified into several categories base on the found pattern type and one of the categories is selected to be the candidate category. Finally, the hybrid neural network/genetic algorithm model is trained by using the data from the candidate category and then is applied to predict the resources usage of the cloud platform.
The proposed method is tested with four different data sets and compared with neural network, exponential smoothing and pattern fusion model respectively. Each data set has diverse characteristics and is used to test various attribute. Experimental results indicate the proposed method has better predicting ability, adjustability and stability.

1. Armstrong, J. and T. Overton, Estimating nonresponse bias in mail surveys. Journal of marketing research, 14: p. 396-402. 1977.
2. Barlow, R.E. and F. Proschan, Statistical theory of reliability and life testing: probability models, DTIC Document, Editor^Editors. 1975.
3. Caminero, A.C., et al. Cloud-based e-learning infrastructures with load forecasting mechanism based on Exponential Smoothing: A use case. in Frontiers in Education Conference (FIE), 2011. 2011.
4. Hasançebi, O. and F. Erbatur, Evaluation of crossover techniques in genetic algorithm based optimum structural design. Computers & Structures, 78(1): p. 435-448. 2000.
5. Heckerling, P.S., et al., Use of genetic algorithms for neural networks to predict community-acquired pneumonia. Artificial Intelligence in Medicine, 30(1): p. 71-84. 2004.
6. Holland, J.H., Adaptation in Natural and Artificial Systems: The University of Michigan Press. 1975.
7. Hornik, K., M. Stinchcombe, and H. White, Multilayer feedforward networks are universal approximators. Neural networks, 2(5): p. 359-366. 1989.
8. Irani, R. and R. Nasimi, Evolving neural network using real coded genetic algorithm for permeability estimation of the reservoir. Expert Systems with Applications, 38(8): p. 9862-9866. 2011.
9. Islam, S., et al., Empirical prediction models for adaptive resource provisioning in the cloud. Future Generation Computer Systems, 28(1): p. 155-162. 2012.
10. Kung, S. and J. Hwang. An algebraic projection analysis for optimal hidden units size and learning rates in back-propagation learning. in Neural Networks, 1988., IEEE International Conference on. IEEE. 1988.
11. Li, B. and W. Jiang, A novel stochastic optimization algorithm. Systems, Man, and Cybernetics, Part B: Cybernetics, IEEE Transactions on, 30(1): p. 193-198. 2000.
12. Lora, A.T., et al., Electricity market price forecasting based on weighted nearest neighbors techniques. Power Systems, IEEE Transactions on, 22(3): p. 1294-1301. 2007.
13. Malcolm, D.G., et al., Application of a technique for research and development program evaluation. Operations research, 7(5): p. 646-669. 1959.
14. Martinez Alvarez, F., et al., Energy time series forecasting based on pattern sequence similarity. Knowledge and Data Engineering, IEEE Transactions on, 23(8): p. 1230-1243. 2011.
15. Mell, P. and T. Grance, The NIST definition of cloud computing (draft). NIST special publication, 800: p. 145. 2011.
16. Ming, M. and M. Humphrey. A Performance Study on the VM Startup Time in the Cloud. in Cloud Computing (CLOUD), 2012 IEEE 5th International Conference on. 2012.
17. Nan-sheng, P. and S. Ying-ling. Research on short-term load forecasting based on adaptive hybrid genetic optimization BP neural network algorithm. in Management Science and Engineering, 2008. ICMSE 2008. 15th Annual Conference Proceedings., International Conference on. IEEE. 2008.
18. Saripalli, P., et al. Load Prediction and Hot Spot Detection Models for Autonomic Cloud Computing. in Utility and Cloud Computing (UCC), 2011 Fourth IEEE International Conference on. 2011.
19. Wei, F., et al. RPPS: A Novel Resource Prediction and Provisioning Scheme in Cloud Data Center. in Services Computing (SCC), 2012 IEEE Ninth International Conference on. 2012.
20. Yang, D., et al., A pattern fusion model for multi-step-ahead CPU load prediction. Journal of Systems and Software. 2012.
21. Zurada, J.M., Introduction to artificial neural systems: West St. Paul. Vol. 408. 1992.