Time series forecasting based on weighted clustering

This paper proposes a novel time series forecasting method based on a weighted self-constructing clustering technique. The training data patterns are processed incrementally. If a data pattern is not similar enough to an existing cluster, it forms a new cluster of its own. However, if a data pattern is similar enough to an existing cluster, it is added to the most similar cluster. During the clustering process, weights are learned for each cluster. For a given series of time-stamped data up to time t, we divide it into a set of training patterns. By using the weighted self-constructing clustering, the training patterns are grouped into a set of clusters. A radial basis function (RBF) network is then constructed, for which the obtained clusters are served as the basis functions of the hidden neurons. To forecast the value at time t + 1, the input pattern is fed into the resulting RBF network and the corresponding network output is taken as the estimation. Experimental results are shown to demonstrate the effectiveness of the proposed approach.