| abstract |
The invention discloses a cloud computing energy consumption prediction method based on time series clustering, which performs feature extraction and vectorization mapping, applies clustering algorithm to classify subsequences, generates homogeneous clusters, and builds models and trains according to clusters respectively. and prediction, based on deep learning, autoencoder, GAF method, convolutional autoencoder, clustering algorithm based on input subsequence and other methods to mine the dynamic characteristics of energy consumption data and the high-order nonlinear relationship between related multivariables , reducing the time complexity, the selection of the time series prediction model has greater flexibility, and the performance and accuracy of the prediction have been greatly improved. |