基于BP神经网络的保险欺诈识别研究——以中国机动车保险索赔为例

当前保险欺诈在国内外呈现蔓延态势,尤其体现在机动车保险领域,欺诈识别已成为保险欺诈研究的核心内容.目前保险欺诈识别有统计回归和神经网络两大类方法,这两种方法在指导思想和识别流程上各有优缺.本文基于我国财产保险公司车险索赔样本数据,检验BP神经网络在我国保险欺诈识别中的有效性;同时为了尝试统计回归和神经网络的有效融合,本...     

财务报表舞弊识别分析

一、国内外研究综述综观国内外关于会计舞弊问题研究的文献可以发现,关于会计舞弊问题研究,既有规范研究又有实证研究,其研究内容涉及会计舞弊的动因、识别、防范与治理等多方面,识别方法既有传统识别方法又有统计学识别技术和神经网络模型以及其他数据     

基于深度卷积神经网络的相机模式识别

人工智能开始走进生活方方面面,并已经开始应用于各行各业。随着深度卷积神经网络的快速发展,深度卷积网络结构的优化、完善,深度卷积神经网络在图像识别、检测领域发挥着重要的作用。目前,给定相机图像来识别手机品牌的相机模式识别在图像溯源领域引起了很大的关注。本文首先应用当前流行的深度卷积神经网络(Goog Le Net、Res Net、Dense Net)在多种手机相机模式上进行识别,同时针对三种深度卷积网络在网络结构上、识别效果上进行分析比较,我们的实验结果证明了深度卷积神经网络在相机模式识别上的高效性。     

基于图卷积神经网络的客户识别模型

作为近几年发展最快的数据挖掘新技术,图神经网络的强大之处在于其能学习到图的嵌入表示,从而提取图结构的语义特征,是一项极具发展潜力的技术.从近几年的业界应用来看,图神经网络在分子制药、交通规划、线上社交、电子商务、自然语言处理、图像影像处理方面都取得了令人瞩目的成绩.光大银行重庆分行建设了基于客户金融数据及其关系的图神经网络,实现了基于图卷积神经网络的客户识别模型.该行在实际应用中发现,图神经网络的数据表征能力比传统机器学习模型更具优势,模型对未知样本的泛化认知能力更强,能胜任更复杂的数据挖掘任务,并且已经产生显著的业务价值.     

企业内部控制缺陷识别与诊断研究——基于神经网络的模型构建

内部控制缺陷认定是判断企业内部控制是否有效的重要依据,是内部控制评价的重要部分。本研究将BP神经网络技术运用于内部控制缺陷认定,探讨通过建立BP内控缺陷识别模型实现对企业内部控制缺陷的识别诊断,解决目前缺陷认定过度依赖评价人员职业素养且主要集中在事后判断,缺少事中识别和预警的问题。研究构建了BP缺陷识别模型,分析了利用模型识别缺陷程度的基本原理,并对模型的应用效果进行了识别检验。本文为研究和开展企业内部控制缺陷自动化识别跟踪提供了一种可行路径,也为人工智能与内部控制结合探讨提供了一种思路和范例。     

碳中和背景下气温衍生品定价研究——基于ELM神经网络方法

气温衍生品是一种用来规避天气风险的新型金融工具,它对能源、农业和旅游业等行业的稳健运行、绿色金融的发展、碳中和目标的实现都具有十分重要的价值.选取我国六个典型城市2009—2018年的日平均气温作为样本数据,利用ELM神经网络模型对气温时间序列进行预测与误差分析,借助蒙特卡洛模拟方法对气温衍生品定价.研究结果表明,ELM神经网络较ARMA模型和BP神经网络气温预测精度有显著提高,可为气温衍生品的定价奠定基础.     

机器学习对上市公司年报错报的识别研究--财务重述预测的视角

上市公司年报披露的会计信息是利益相关者决策的重要依据,错报及之后的财务重述降低了会计信息质量,增加了注册会计师的审计风险。随着盈余管理手段愈发隐蔽,单纯依赖传统的风险评估方法可能难以有效识别错报风险。机器学习具有独特的学习模式和甄别方法,能对错报做出有效识别,有助于降低审计风险。实证结果表明,与logistic回归相比,多层感知机神经网络等五类机器学习方法对上市公司调减盈余的年报财务重述的预测准确率、F1分值和第Ⅰ类错误率等方面都表现更优;研究还发现,营运能力等四类特征在预测过程中起到重要作用。     

基于神经网络的期权定价研究综述

本文介绍了期权定价理论,详细描述了期权定价研究的现状,并总结了当前几种期权定价主要方法及其基本指导思想。然后,结合当前期权定价方法的热点,重点阐述了神经网络在期权定价中的应用。最后,对基于神经网络预测的期权定价研究进行了总结,并指出了神经网络的期权定价的不足及研究方向。     

资产风险预警的神经网络技术实现

本文采用美元指数自1973年的日数据进行研究.首先,构造了一个衡量价格波动的函数,在此数异常波动时,往往就预示着价格将要发生变动;然后,利用基于遗传算法的神经网络构建资产价格异常波动预警系统.在神经网络的训练过程中,利用粗糙集对训练样本进行了约简,提高了训练效果.结果表明,该系统可以有效对资产价格的异常波动做出预警,从而为风险管理提供了一种新的思路.     

基于卷积神经网络在征信方面的研究

<正>卷积神经网络(CNN)最早用于手写数字识别并一直保持了其在该问题的霸主地位。近年来卷积神经网络在多个方向持续发力,语音识别、人脸识别等方面均有突破。一、项目简介基于卷积神经网络(CNN)的征信系统,作为个人征信系统第三方机构,利用合作单位已有大数据对卷积神经系统进行训练,使系统具有高准确率对未来借贷人交易数据进行分析,快速得出是否可提供贷款及借贷金额和时间。     

Short-sale refinancing and price adjustment speed to bad news: Evidence from a quasi-natural experiment in China

Short selling may accelerate stock price adjustment to negative news. However, the literature provides mixed evidence for this prediction. Using short-sale refinancing and a staggered difference-in-differences (DID) model, this paper explores the effect of short selling on stock price adjustment. Our results show that (1) short-sale refinancing improves the speed of stock price adjustment to negative news. This result holds after we control for endogeneity. (2) The positive relationship between short-sale refinancing and stock price adjustment speed is significant in subsamples of stocks with higher earnings management or lower accuracy of analyst forecasts, indicating that firms with more opaque information are more likely to be targeted by short sellers. In subsamples of stocks with a higher ownership concentration or lower ownership by institutional investors, short selling is more likely to increase the speed of stock price adjustment, indicating that ownership structure may influence negative news mining. (3) As short-sale refinancing exacerbates the absorption of bad news by stock prices, it increases crash risk. This study enriches the research on the economic consequences of short selling and provides empirical evidence supporting regulations on short selling in China.     

Cost accounting standard and cost accounting systems in Japan. Lessons from the past - recovering lost traditions

This paper aims to show two things. The first is how Japanese culture has contributed to the development of Japanese cost accounting history. The second is to reveal the research possibilities of cost accounting history. This paper also reviews the salient features of several important examples of these aspects of cost accounting practice in Japan. It therefore explores, through some practical illustrations, how and why Japanese cost accounting differs from that found in the West.     

Guidelines for Manuscripts Submitted to The China Journal of Accounting Research

【正】The China Journal of Accounting Research‘‘CJAR’’(ISSN 1755-3091)publishes quarterly.It contains peer-reviewed articles and commentaries on accounting,auditin...     

Shared Standard Setting in the Public Interest:A Strong Model

正The last two decades have been a definitive era in the evolution of the accountancy profession.In the wake of major corporate scandals at the turn of the century,an international public debate arose on the need for more effective and well-considered regulation;this debate then reignited during the global financial and sovereign debt crises.     

HOW TO PUT TOGETHER A COMPLETE PACKAGE

正Many Institute members know they cannot rely purely on technical knowledge and business acumen to remain competitive.As soft skills are increasingly important,Jemelyn Yadao finds out how CPAs can maintain     

Guidelines for Manuscripts Submitted to The China Journal of Accounting Research

正The China Journal of Accounting Research"CJAR"(ISSN 1755-3091)publishes quarterly.It contains peer-reviewed articles and commentaries on accounting,auditing and corporate governance issues that relate to the greater China region.We welcome the submission of both theoretical and empirical research papers pertinent to researchers,regulators and practitioners.Authors should note:     

The China puzzle: Opportunities for accounting research

How can China achieve phenomenal economic growth despite what is considered as 'weak' institutions in market-based economies? Xu(2011) provides a framework to u...     

Aims and Scope

<正>The focus of the China Journal of Accounting Research is to publish theoretical and empirical research papers that use contemporary research methodologies to...     

Income smoothing and the cost of debt

The literature on income smoothing focuses on the effect of earnings smoothing on the equity market.This paper investigates the effect of income smoothing on th...     

The case for management history

This article defines the case for treating management history as a discrete subject area, although one closely related to business, accounting, labour, and industrial history. It seeks to concentrate on the nature, process, and practice of management as an activity within the organization, and to argue that it is important that modern British managers understand the intellectual and cultural roots of their profession, since without these they will tend to be swayed by short-term management 'fads'. Within the broad topic there are several identifiable sub-areas for potential exploration, including the comparative historical development of management between different countries, management institutions, managerial careers and labour markets, management structures, management skills, and management thought and theory.