基于最鄰近算法的財政數(shù)據(jù)異常值實時監(jiān)測方法研究

打開文本圖片集

中圖分類號：TP393 文獻標志碼：A

文章編碼：1672-7274（2025）04-0013-04

Abstract： The existing real-time monitoring methods for abnormal values in financial data have low accuracy and high false alarmrates，which leads to inaccurate monitoring of abnormal values in financial data and has certain limitations.This article proposes a real-time monitoring method for financial data outliers basedon nearest neighbor algorithm.Firstly，bycalculatingthelocaldensityandminimumdistance，selectingtheRBFkernelfunction，areal-time monitoring modelfor financial dataoutliers is established basedontheclustering results.Secondly，the intensityratio ofthesetreference windowand investigation window iscalculated to extract abnormal paterns in fiscal data.Finally， based on the elbow rule curve and follwing acertain process，the monitoring task of abnormal values in financial data is completed.The experimentalresultsshowedthatusing thereal-time monitoring methodforfinancialdataoutliers basedon the nearest neighbor algorithm，the monitoring accuracy was over 95% ，and its average falsealarm rate was 3.21%

Keywords： local density; nearest neighbor algorithm; RBF kernel function; intensity ratio; elbow rule

在財政數(shù)據(jù)管理中如何準確、高效地識別和處理異常值成為保障數(shù)據(jù)質(zhì)量[1]、預防財務風險中亟待解決的問題。（剩余3110字）