摘要
提出一种基于Prophet框架的银行网点备付金预测方法,即HC方法(holiday changepoints method)。首先以银行网点交易流水数据为基础,统计每个现金备付周期内的交易存取款额指标,并进行标准化得到备付金时间序列;然后,构建非周期性的节假日列表和趋势转折点列表,利用Prophet框架完成对这2类列表中特殊点的特征计算,有效解决"异常值"和"拐点"的预测问题;最后,结合可视化技术实时观测算法效果调节参数,得到预测模型。以平均绝对误差、均方根误差、平均绝对百分比误差和绝对误差这4个性能度量指标来评估HC模型对银行网点备付金时间序列的预测效果。研究结果表明:该算法在银行网点备付金预测问题上相较于ARMA算法和LSTM算法具有更高的准确率。
A method called HC(holiday changepoints method)was presented based on the Prophet framework for the forecast of bank payment.Firstly,for each cash reserve period,the transaction deposit amount index based on trading data of bank outlets was calculated and standardized to obtain reserve time series.Then,the non-periodic holiday list and trend turning point list were constructed.Based on the Prophet framework,the characteristics of the special case in the two types of lists were further calculated,and the"abnormal value"and"inflection point"prediction problem was effectively solved.Finally,by using visualization technology,the real-time observation algorithm effect was adjusted to build a prediction model.The predicted results on the payment time series of bank outlets were evaluated by four performance metrics including average absolute error,root mean square error,average absolute percentage error and absolute error.The results show that the HC method is more accurate than the ARMA algorithm and the LSTM algorithm in reserve prediction of bank outlets.
作者
李丽萍
段桂华
王建新
LI Liping;DUAN Guihua;WANG Jianxin(School of Information Science and Engineering,Central South University,Changsha 410083,China)
出处
《中南大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2019年第1期75-82,共8页
Journal of Central South University:Science and Technology
基金
国家自然科学基金资助项目(61572530
61602171)~~