基于时间序列算法的资金流入流出预测模型比较分析

Comparative Analysis of Capital Inflow and Outflow Prediction Models Based on Time Series Algorithm

摘要: 商业公司的金融服务平台通常有着上亿的客户，庞大的用户群体使得其服务场景中每天都必然会触及大量资金的流入和流出，对资金的管控压力也越来越大。在当前的宏观经济环境和现实状况下，要确保资本管理在一个可控的范围内，对未来的资本流量规模做出预估十分必要。现有文献表明，在实际操作中有多种模型可以进行相关的预测。使用比较多的包括基于统计学方法来完成的ARIMA模型、周期因子时间序列预测模型以及基于神经网络的LSTM时间序列预测模型等。本文以余额宝平台用户资金流量为背景，通过比较以上三种不同资金流量预测模型的适用条件、优点及局限性，尝试对比分析出针对不同条件下更加准确有效的资金流入流出预测模型，最大程度贴近未来资金流量真实值，从而更好地进行企业的经营管理。

Abstract: The financial service platform of commercial companies usually has hundreds of millions of cus-tomers. The huge user group makes it inevitable to touch a large amount of capital inflow and out-flow every day in its service scenario, and the pressure on capital control is also increasing. In the current macroeconomic environment and current situation, it is very necessary to predict the scale of future capital flow in order to ensure that capital management is within a controllable range. The existing literature shows that there are many models that can make relevant predictions in practi-cal operation. ARIMA model based on statistical method, periodic factor time series prediction model and LSTM time series prediction model based on neural network are widely used. Taking the user capital flow of Yu’e Bao platform as the background, by comparing the applicable conditions, advantages and limitations of the above three different capital flow prediction models, this paper attempts to compare and analyze a more accurate and effective capital inflow and outflow predic-tion model under different conditions, which is close to the real value of future capital flow to the greatest extent, so as to better carry out the operation and management of enterprises.