Forecasting of Short-term Tourism Demand Based on Multivariate Time Series Clustering and LSSVM

Abstract

A short-term tourism demand forecasting method based on multivariate time series clustering and LSSVM is proposed. Firstly, continuous time samples are intercepted into multivariate time series samples by using sliding time window; Then, it uses the multivariate time series clustering method based on principal component analysis to classify them and generate similar time segment subsets; Finally, the LSSVM model is used to forecast according to the subset data of similar time periods. The results show that compared with the comparison model, the model can effectively improve the forecasting accuracy.