Forecasting Yu'e Bao's Inflows

Internet Finance is becoming a hot topic in recent years.In less than three years' time when domestic scholars begin to pay attention to this area,most of the research still stays at the qualitative level,lacking of quantitative research.In this paper,from the perspective of forecasting cash flow,I hope to complement domestic research in the field of Internet Finance.This paper will first review the domestic and overseas main research results and application in the field of Internet Finance.Then STL decomposition method will be introduced in detail,and some classical methods of time series prediction will be summed up.Next this paper will mainly introduce a combination forecasting model,which is the recommended model of this research.At last an empirical study will use three classical methods and the combination forecasting model on Yu'E Bao's daily purchase amount,in order to forecast its next month's data.After comparing the prediction effect of different models,we can get the following important conclusions:1)Combination model with STL decomposition will first separate each component,and then select the most appropriate prediction model for each component.This makes it easier to understand the structure of the data,and makes the whole forecasting process easier to explain and apply.2)Combination model takes residual component into the whole forecasting process,improving information utilization and reducing information loss in the process of forecasting.3)Because combination model is a composite of each component's forecasting results,its performance on short-term prediction is not only good,but also stable.4)Combination model performs best on long-term prediction.Long-term prediction's performance mainly depends on the prediction of long-term trends.Combination model first separates components and forecasts trend separately,which can better characterize the long-term trend and ensure the whole model's performance on the long-term forecasting.