Hierarchical decomposition method and combination forecasting scheme for access load on public map service platforms

The Internet and mobile networks have grown and expanded, and public map services have penetrated into people's daily lives, thereby generating high-intensity access loads on the platforms that service massive amounts of users. Cloud computing can be used to improve the service performance, but its pricing mode is demand-based as a commercial service. Estimating the scale of user requests in advance is important for meeting the demands of users in a timely manner and allocating appropriate resources in the cloud. In this study, we designed effective methods for describing and predicting the access load on public map service platforms (PMSPs). We propose a novel hierarchical decomposition model for describing the components and structure of access loads, where this model accommodates the non-stationary, variable intensity, and periodic nature of access loads. Furthermore, we propose a wavelet transform-time series decomposition-autoregressive integrated moving average (WT-TSD-ARIMA) prediction scheme to capture and predict data components with various different frequencies and characteristics based on a combination of one-level WT, TSD, and simple linear prediction models (such as ARIMA). Our experimental results show that this prediction scheme significantly improves the prediction performance for linear prediction models while maintaining their stable long-term prediction capacity. The proposed method may facilitate the reliable and adaptive scalable management of cloud computing resources for PMSPs in real time.