Industrial Time Series Prediction Based on Incremental DBSCAN-KNN with Self-learning Scheme

Industrial time series data are usually time-varying due to multiple factors such as environmental and human disturbances. As traditional time series predicting methods are often based on offline training ignoring the changes in working conditions, the prediction results may be inaccurate. In this paper, a time series prediction model based on incremental DBSCAN and KNN with self-learning scheme is proposed to address the problem of time-varying working conditions. The proposed model uses the incremental DBSCAN to automatically identify and expand working conditions with adjusting the number of clusters automatically, and then employs the KNN model to make predictions under different working conditions. Compared with the existing methods, the proposed method is more stable and improves the prediction accuracies of the model under different working conditions.