Big Data Service Request Prediction Based on Historical Behavior Time Series

Abstract

Big data analysis service has been used widely in our society. For example, in financial field, users often use big data analysis services to analyze stocks, assets, and accounts in real-time investment decision-making. Therefore, real-time service response is very important from the perspective of user experience. Caching data and analysis results have been used widely in industrial practice. But these caches generally are passive, rigid and inefficient. Proactive caching approach for time-consuming data services is a worthwhile research problem. In addition, we have encountered this problem in a practical enterprise application. In this paper, we propose a data service request prediction approach based on historical user behavior time series analysis. Results show that this approach can improve the response speed of backend data services effectively.