基于同比的概率预测模型

摘要

个人外汇业务核查系统利用外汇交易大数据，通过对大数据的算法分析，将存在借用他人额度办理结售汇行为的个人，直接列入“关注名单”，能够及时地查找异常交易主体，预测分析可能成案的线索因素，为执法部门锁定目标、发现异常、甄别违规、快速执法提供依据，实现非现场数据分析与现场检查的有效结合，进而提高现场检查的实际效果。为了分析个人外汇业务中的分拆量、分拆金额总量的变化区间，掌握个人外汇业务数据的量变与质变，需要对个人外汇业务的分拆量、分拆金额量等进行预测分析。各种预测模型都有其应用范围，在对现有预测算法分析的基础上，我们基于同比预测模型，以概率为基础建立了同比概率预测模型。同比概率预测模型的外汇大数据的仿真验证表明：同比概率预测算法不仅能得到时间序列的数据变化趋势，同时可以使数据根据季节性特点呈现波动性变化。同比概率预测模型对年份间数据差距较大，且存在波动的数据进行预测时，精度高于以时间序列为主的灰色预测模型。 The verification system of personal foreign exchange business uses the big data of foreign exchange transactions to directly list individuals who borrow others’ quota to handle the settlement and sale of foreign exchange, through the algorithm analysis of big data. It can find out the subject of abnormal transaction in time, predict and analyze the possible clues, which provides the basis for legal authorities to lock in targets and find out abnormalities, screen violations and fast law enforcement. The effective combination of off-site data analysis and on-site inspection can be achieved, which can improve the actual effect of on-site inspection. It is necessary to predict and analyze the amount and total amount of the individual foreign exchange business in order to analyze the change range of split amount and total split amount, as well as master the quantitative and qualitative changes of the individual foreign exchange business data. Each forecasting model has its application scope. Based on the analysis of the existing prediction algorithms, we establish the forecasting of year-on-year probability based on probability. The simulation verification of the big foreign exchange data of the year-on-year probability forecasting model shows that it can not only obtain the data change trend of time series, but also make data fluctuate changes according to the seasonal characteristics. The forecasting model for year-on-year probability is more accurate than the gray forecasting model based on time series when predicting data with large data gaps between years and fluctuations.