A Novel Prediction Approach for Runoff Based On Hybrid HMM-SVM Model

This research demonstrates an application of Hidden Markov Model (HMM) and Support Vector Machine (SVM) for watershed-runoff forecasts. HMM is used for shape-based clustering by calculating log-likelihood values of each data to identify data in the data set with similar data pattern. Then we put these data into different classes based on their shapes and train their corresponding SVM model to predict the output of the system finally. The applications of daily runoff and monthly runoff are used for testing the competence of this method and experimental results demonstrate that this hybrid HMM-SVM algorithm can meet the prediction requirement and has high prediction accuracy.