A Self-Adapting and Efficient Dandelion Algorithm and Its Application to Feature Selection for Credit Card Fraud Detection

A dandelion algorithm(DA) is a recently developed intelligent optimization algorithm for function optimization problems. Many of its parameters need to be set by experience in DA,which might not be appropriate for all optimization problems. A self-adapting and efficient dandelion algorithm is proposed in this work to lower the number of DA's parameters and simplify DA's structure. Only the normal sowing operator is retained;while the other operators are discarded. An adaptive seeding radius strategy is designed for the core dandelion. The results show that the proposed algorithm achieves better performance on the standard test functions with less time consumption than its competitive peers. In addition, the proposed algorithm is applied to feature selection for credit card fraud detection(CCFD), and the results indicate that it can obtain higher classification and detection performance than the-state-of-the-art methods.

A Credit Card Fraud Detection Model Based on Multi-Feature Fusion and Generative Adversarial Network

Credit Card Fraud Detection(CCFD)is an essential technology for banking institutions to control fraud risks and safeguard their reputation.Class imbalance and insufficient representation of feature data relating to credit card transactions are two prevalent issues in the current study field of CCFD,which significantly impact classification models’performance.To address these issues,this research proposes a novel CCFD model based on Multifeature Fusion and Generative Adversarial Networks(MFGAN).The MFGAN model consists of two modules:a multi-feature fusion module for integrating static and dynamic behavior data of cardholders into a unified highdimensional feature space,and a balance module based on the generative adversarial network to decrease the class imbalance ratio.The effectiveness of theMFGAN model is validated on two actual credit card datasets.The impacts of different class balance ratios on the performance of the four resamplingmodels are analyzed,and the contribution of the two different modules to the performance of the MFGAN model is investigated via ablation experiments.Experimental results demonstrate that the proposed model does better than state-of-the-art models in terms of recall,F1,and Area Under the Curve(AUC)metrics,which means that the MFGAN model can help banks find more fraudulent transactions and reduce fraud losses.

Clinical features of adult spinal muscular atrophy:46 cases

BACKGROUND： Spinal muscular atrophy （SMA） is a kind of degenerative disease of nervous system. There are 4 types in clinic, especially types Ⅰ, Ⅱ and Ⅲ are common, and the researches on those 3 types are relative mature. Type IV is a kind of adult spinal muscular atrophy （ASMA）, which has low incidence rate and is often misdiagnosed as amyotrophic lateral sclerosis, muscular dystrophy, cervical syndrome, or others.OBJEETIVE： To observe the clinical features of 46 ASMA patients and analyze the relationship between course and activity of daily living. DESIGN ： Case analysis.SETTING： Departments of Neurology of the 81 Hospital of Chinese PLA, the Second Affiliated Hospital of Nanjing Medical College and General Hospital of Nanjing Military Area Command of Chinese PLA.PARTICIPANTS ： A total of 46 ASMA patients were selected from the Departments of Neurology of the 81 Hospital of Chinese PLA, the Second Affiliated Hospital of Nanjing Medical College and General Hospital of Nanjing Military Area Command of Chinese PLA between April 1998 and January 2002. All patients were consentient. Among 46 cases, there were 37 males and 9 females with the mean age of 42 years. The patients＇ courses in all ranged from 6 months to 23 years, concretely, courses of 37 cases were less than or equal to 5 years, and those of 9 cases were more than or equal to 6 years.METHODS： ① All the 46 ASMA patients were asked to check blood sedimentation, anti O, serum creatinine, creatine, blood creatine phosphokinase （CPK） and muscular biopsy as early as possible. ②X-ray was used to measure plain film of cervical vertebra borderline film of cranium and neck at proximal end of upper limb of 25 cases and plain film of abdominal vertebra at proximal end of lower limb of 17 cases. ③ Cerebrospinal fluid of lumbar puncture was checked on 42 cases, for routine examination, biochemical examination, and immunoglobulin examination. Electromyogram （EMG） was also examined to 42 cases. ④ Barthel index was used to evaluate activities of daily living （ADL） of patients with various courses. The index ranged from 1 to 100. The more the index of a ASMA was, the stronger his independence was. ⑤ The Barthel indexes of patients with courses ≤ 5 years and those ≥ 6 years were compared with univariate analysis of variance. MAIN OUTCOME MEASURES： ① Incidences of all patients at the first time; ② values of relative blood and blood biochemistry; ③results of muscular biopsy; ④ results of EMG and relative X-ray plain film of 42 cases; ⑤ results of cerebrospinal fluid of 42 cases; ⑥ comparisons of Barthel index of patients with various courses.RESULTS： A total of 46 ASMA patients were involved in the final analysis. ① Incidence on the first time： 25 patients had the disease at the proximal end of upper limb, 17 at the proximal end of lower limb, and 4 at the four limbs. ② Value of serum-blood CPK of one fourth patients was increased slightly （3.034-9.735 μkat/L; normal value： 0.400-3.001 μkat/L）, and other values of blood and blood biochemical indicator were normal. ③Results of muscle biopsy of all patients showed that a small group of muscular atrophy could be observed mostly, and muscle group in the same type and compensatory hypertrophy of muscle fibres were also observed with ATP enzyme staining. ④ Results of EMG of 42 cases suggested that 37 patients had mild and moderate nerve-derived injury and 3 had mild muscle-derived injury. Results of all the X-ray plain films in this study were normal. ⑤ Results of routine, biochemical and immunoglobulin examination in cerebrospinal fluid of lumbar puncture in 42 cases were all normal. ⑥The difference between Barthel indexes of patients with courses ≤ 5 years and those ≥ 6 years was not significant [（64.73±20.38） vs （68.89±21.76） points, P〉 0.05]. CONCLUSION ： ① Amyasthenia is mainly occurred at the proximal end of the four limbs of ASMA patients. A small group of muscular atrophy is its mostly pathological change, and the progression of the disease is slow. ② Most patients have mild and moderate nerve-derived injury under EMG examination.③ The duration of a patient suffered from the disease has no obvious effect on his ADL ability.

A Multi-Module Machine Learning Approach to Detect Tax Fraud

Tax fraud is one of the substantial issues affecting governments around the world.It is defined as the intentional alteration of information provided on a tax return to reduce someone’s tax liability.This is done by either reducing sales or increasing purchases.According to recent studies,governments lose over$500 billion annually due to tax fraud.A loss of this magnitude motivates tax authorities worldwide to implement efficient fraud detection strategies.Most of the work done in tax fraud using machine learning is centered on supervised models.A significant drawback of this approach is that it requires tax returns that have been previously audited,which constitutes a small percentage of the data.Other strategies focus on using unsupervised models that utilize the whole data when they search for patterns,though ignore whether the tax returns are fraudulent or not.Therefore,unsupervised models are limited in their usefulness if they are used independently to detect tax fraud.The work done in this paper focuses on addressing such limitations by proposing a fraud detection framework that utilizes supervised and unsupervised models to exploit the entire set of tax returns.The framework consists of four modules:A supervised module,which utilizes a tree-based model to extract knowledge from the data;an unsupervised module,which calculates anomaly scores;a behavioral module,which assigns a compliance score for each taxpayer;and a prediction module,which utilizes the output of the previous modules to output a probability of fraud for each tax return.We demonstrate the effectiveness of our framework by testing it on existent tax returns provided by the Saudi tax authority.

基于特征提取和集成学习的个人信用评分方法

在大数据蓬勃发展的今天,信息经济已经深入社会方方面面,个人信用体系建设的重要性越发突出。而传统的信用体系存在覆盖率不足、评价特征维度高、数据孤岛等问题,为了解决以上问题,提出一种基于特征提取和Stacking集成学习的个人信用评分方法(PSL-Stacking)。方法首先利用Pearson和Spearman系数对数据进行初始化分析剔除不相关数据,利用LightGBM算法进行特征选择,减少冗余特征对模型的影响;其次选取XGboost、LightGBM、Random Forest以及Huber回归等算法,利用Stacking集成学习技术构造个人信用评分模型。最后,以某电信数据为研究对象,对该上述模型的个人信用评分能力进行验证。实验结果得出上述模型具有很好的预测能力,能够准确的对用户信用进行评分,有效降低企业遭受金融欺诈、团伙套利等问题的风险。

基于深度学习和文本情感的上市公司财务舞弊识别方法

上市公司财务舞弊是指采用不当手段使会计信息失真的恶劣失信行为,对公司经营、经济发展和社会利益产生负面影响。目前,更多研究关注于金融数字数据,对文本信息和深度学习算法的研究较少,因此提出了一种基于深度学习并融合文本情感特征的上市公司财务舞弊识别方法。该方法首先对财务指标进行预处理,并使用BiLSTM提取股评文本的情感特征。然后使用RCC(residual-cross-convolutional)并行网络进行财务舞弊识别,其采取残差网络、交叉网络、卷积网络和长短期记忆网络并行的方式提取财务舞弊特征,并使用批标准化和全连接得到最终的识别结果。实验结果表明,该方法在识别上市公司财务舞弊的任务上较其他模型取得了更好的效果,召回率和AUC分别达到了88.46%和82.06%。

基于多种特征库的短信轰炸识别方法设计

该文设计一种基于多种特征库的短信轰炸识别拦截方案,可有效地识别轰炸端口和轰炸行为,为用户做好骚扰屏蔽,有效地减少骚扰投诉,做好通信保障,保护用户隐私及财产安全。

面向溯源的虚假危机信息传播主体识别与动机叙事方法研究

[目的/意义]虚假危机信息类型复杂,利益牵扯甚广,传播动机较为分散与隐蔽,有必要从溯源角度进行主体识别与动机叙事研究。[方法/过程]首先,在确定文本情感值、实体数、时序率、聚类簇与重叠量5个识别要素的基础上,设计深度隐晦语义理解、虚假危机信息判别、关联主体锁定与主体确证流程实现虚假危机信息传播主体识别功能,并通过画像对特征予以暴露。然后,以事理图谱为底座,在意图理解基础上构建动机叙事模型。最后,将两种方法糅合后开发自适应可视化操作界面以提升应用效率。[结果/结论]以“5·23卡塔尔立场声明”事件信息为例开展实证,结果表明此研究方法能够识别煽动型、造势型、转化型、侦查型与钳制型5种传播主体,动机分别集中在意见价值共鸣、宣扬自身力量、破除信息壁垒与渗透官方作业等方面。方法对比实验表明,文本设计较于其他方法具有更佳性能,然而仍有优化余地,如主体特征暴露与动机叙事粒度还有待提升。

基于知识图谱的银行反欺诈模型的研究与应用

针对传统的银行反欺诈模型已无法满足欺诈检测要求的及时性与准确性的问题,本文提出了一种基于知识图谱的反欺诈模型。该模型基于多源信息和高维衍生特征的大数据,构建知识图谱,对信贷个体进行全方位画像,分析关联关系,并抽取网络属性。从四大方面、两大维度挖掘风险特征,四大方面指个人基本信息、账户信息、征信和行为信息,两大维度指个人节点和网络结构。最后将风险特征代入LightGBM,判断是否为欺诈类型,并得到对应概率。实验表明,相比于仅使用个人自身特征的模型,使用个人特征加网络特征的模型效果更好,AUC和F1分数分别提升5.18%和5.71%。因此,该方案能够有效地为银行对个人信贷进行欺诈评估。

基于RFE-LGB算法的上市公司财务造假分析和预测

针对上市公司财务造假预测问题,采用结合了LightGBM与递归特征消除法(RFE)的方法进行数据建模。LightGBM以其超参数量少、强大的稳健性及对不平衡数据的高敏感性等特点著称。RFE作为一种封装式特征选择方法,能高度匹配所用预测模型,并通过设定特征子集评价函数作为停止条件,自动确定最优特征数量,这在特征选择领域具有较大优势。此外,选用平衡精度(BAcc)作为模型预测性能的评估指标,并通过调整LightGBM的分类权重参数来解决样本不平衡的问题。在5个不同行业财务数据集上的实验结果表明,所提出的RFE-LGB模型在上市公司财务造假预测任务中表现出良好的平衡性、稳健性和泛化性。该模型能有效识别与财务造假相关的关键指标,且仅使用较少的核心特征即可达到较高的预测精度。

基于多特征的SMS僵尸网络检测模型

SMS(短信息服务)作为移动终端最常用的服务之一也成为了恶意软件的主要目标,其中SMS僵尸网络是最具威胁性的攻击。针对SMS僵尸网络日益活跃,本文提出了一种基于多特征的SMS僵尸网络检测模型,该检测模型拥有两个核心模块,第一个模块利用短消息第一类特征对被检测信息做一次快速筛选来提高检测效率;第二个模块则是基于贝叶斯分类算法利用短信息的第二类特征将短信分类鉴别。实验结果表明,本文提出的方法能对SMS僵尸网络有96%的检测准确率。

Mobile SMS Spam Filtering for Nepali Text Using Naive Bayesian and Support Vector Machine

Spam is a universal problem with which everyone is familiar. A number of approaches are used for Spam filtering. The most common filtering technique is content-based filtering which uses the actual text of message to determine whether it is Spam or not. The content is very dynamic and it is very challenging to represent all information in a mathematical model of classification. For instance, in content-based Spam filtering, the characteristics used by the filter to identify Spam message are constantly changing over time. Na?ve Bayes method represents the changing nature of message using probability theory and support vector machine (SVM) represents those using different features. These two methods of classification are efficient in different domains and the case of Nepali SMS or Text classification has not yet been in consideration;these two methods do not consider the issue and it is interesting to find out the performance of both the methods in the problem of Nepali Text classification. In this paper, the Na?ve Bayes and SVM-based classification techniques are implemented to classify the Nepali SMS as Spam and non-Spam. An empirical analysis for various text cases has been done to evaluate accuracy measure of the classification methodologies used in this study. And, it is found to be 87.15% accurate in SVM and 92.74% accurate in the case of Na?ve Bayes.

基于三阶段集成学习的信用卡欺诈检测研究

信用卡欺诈是银行业面临的严峻问题,全球每年因信用卡欺诈造成的损失高达数十亿美元。然而信用卡交易数据存在特征冗余和样本不平衡的问题,这无疑增加了模型对少数欺诈交易的检测难度。针对以上问题,本文提出了三阶段集成学习模型“FS-IFKK-Stacking”:基于FS方法的特征选择、基于IFKK方法的不平衡数据处理和基于Stacking方法的异构模型集成。该模型同时解决了由于特征冗余和样本不均衡性导致的过拟合问题,能够更加准确地检测信用卡欺诈交易。基于Kaggle欧洲信用卡交易数据的实验表明:本文提出的“FS-IFKK-Stacking”模型对信用卡欺诈的检测效果明显优于基于原始样本训练的单类模型:在AUC提升0.44%的同时,对欺诈交易的召回率提升了3.27%。

基于多模型融合Stacking集成学习保险欺诈预测

本文结合人工智能的前沿理论研究,提出一种基于改进XGBoost与LightGBM模型融合的Stacking集成学习方式的保险欺诈行为预测模型.该模型对保险公司被保人保险欺诈行为的识别具有启示意义,有助于保险公司更好地识别被保人的骗保行为,强化自身风控体系.首先对XGBoost与LightGBM进行Stacking模型融合生成两个新特征,新生成的两个特征和原有的40个特征合并作为第二层Stacking训练模型的输入.其次,在Stacking的第二层中分别选择使用多种分类学习模型,包括Bagging、LightGBM、XGBoost以及传统机器分类模型,包括逻辑回归、高斯贝叶斯、决策树等,各模型的训练和参数均由K折交叉验证和遗传算法优化得到.算例数据来源于阿里云天池挑战赛公开的保险欺诈数据集,对构建多模型融合的Stacking模型预测性能进行了验证与测试.预测结果表明,与传统机器分类模型预测结果相比,基于XGBoost与LightGBM Stacking模型融合集成学习模型对保险欺诈行为具有较高的识别能力.最后,根据计算并可视化出最优分类模型不同特征的重要性结果,得出结论:被保人的职业、发生保险事故的城市、发生保险事故的地区、资本收益、资本亏损是识别保险欺诈行为的重要特征.

短信广告体裁研究——以淘宝促销短信为例

随着互联网和智能手机的普及,短信广告迅速发展,成为企业商家青睐的新型宣传体裁。本文基于Swales的语步分析理论和Hasan的体裁结构潜势理论,以500条淘宝促销短信为研究对象,深入研究短信广告的语步结构模型和语言特征。研究表明:为达到宣传和劝说的目的,短信广告有六个语步,其中语步一、三、五和六是必选语步,语步二和四为可选语步。在语言使用上,在语步二多用表示时间的名词和表示事件的动词,如“今晚”“点”“狂欢”等,在语步三和语步四多使用“减”“折”“送”等简单单音节动词、“券”“礼”等名词来突出说明活动优惠力度,句子简短。

舞弊治理：基于上市公司财务舞弊特征的分析

选取2003～2007年度5645家沪深两地上市公司为样本，首先对其中209家舞弊公司财务舞弊的影响因素做统计分析，明晰上市公司财务舞弊的特征；然后，根据舞弊公司不同特征进一步实证检验，得出大规模事务所的选择、股权集中度、公司规模与财务舞弊负相关，独立董事比例与财务舞弊成“U”型关系。但没有发现代表财务状况指标净资产收益率的显著性影响，究其原因，发现舞弊公司存在虚增利润以外的其他更多隐性盈余操纵手段。

基于内容的短信分类技术

研究了一种基于改进贝叶斯算法的短信分类方法。对中文文本短信,采用文档频度(DF)的特征项提取方法,借助自建短信语料库对改进的贝叶斯分类器进行了实验测试。实验结果表明:改进的分类器可以提高正常短信的通过率,并可以根据新的训练集训练出个性化的分类器,适应短信变化,满足用户的个性化需求,还结合黑白名单过滤机制实现对短信的过滤,减少了正常短信的误判率。

一种面向网络支付反欺诈的自动化特征工程方法

互联网金融欺诈正导致诸多社会经济问题.网络支付是互联网金融中的典型模式之一,此模式中的欺诈交易也是互联网金融欺诈的主要形式之一.通过构建基于机器学习的欺诈检测模型来识别欺诈交易的方法已成为网络支付反欺诈领域的主流思路.在构建欺诈检测模型的过程中,特征工程是最为关键的一步,特征的质量将直接影响模型的性能;通常,这也是最为耗时且对相关领域的专业知识要求最高的步骤.现有网络支付欺诈检测模型在特征工程上主要是领域专家基于业务知识以手动构造的形式来开展.而在网络支付模式下欺诈场景众多,不同场景下的特征构造流程不尽相同.人工特征构建方法已不能满足与日俱增的反欺诈需求.解决此问题的重要方法之一便是自动化特征工程.本文针对网络支付欺诈检测提出了一种轻量化、树结构、高效率、可扩展和可解释的自动化特征工程方法.该方法:(1)对计算条件的要求低且对数据集样本的依赖性小,这一优势是利用树结构模型进行特征构造得以实现;(2)可构造出深度层次的复杂特征和广度层次的各类型特征,这一优势是利用节点处特征构造的新型流程和转换函数权重向量的时效性更新机制得以实现;(3)在网络支付模式不同场景下可实现跨场景复用,这一优势是通过复用和扩展定制化转换函数得以实现;(4)构造出的特征具有可解释性,这一优势得益于基于结合转换函数与树模型的特征构造过程具备可表达性.本文在网络支付典型场景的业务数据集上验证了所设计的自动化特征工程方法的有效性.

基于特征字符串的GSM短信模块接口设计

短信接口程序是解决移动技术在企业网络中应用问题的有效途径,它能够把方便快捷的短信技术与企业各应用系统无缝对接;为实现计算机对短信的自动收/发,对基于GSM的短信接口进行了设计,分析了计算机利用GSM短信模块收发短信的工作原理,根据短信模块响应AT命令所返回的字符串,提出了一种基于"特征字符串"来控制接收/发送动作的设计方案,并在Windows XP平台下用Delphi 7.0加以实现;测试结果表明,该设计方案具有较高的发送速率和接收响应速度。

上市公司财务异常与舞弊疑点检测研究

对海量数据价值的挖掘趋势迎来了一个前所未有的大数据时代,将数据挖掘技术应用于财务舞弊疑点发现研究成为一个重要的新兴课题。选取2016—2020年我国深沪两市A股医药制造业上市公司为研究对象,以表征财务报表舞弊的关键指标作为检测变量,分别基于跨年数据和年度数据进行异常检测,从中挖掘出存在异常值的离群点上市公司,进而结合违规处罚信息对其进行疑点验证。研究结果表明:跨年数据样本量更大,模型学习更充分,异常检测的效果优于小样本的年度数据;分析导致异常的原因发现,离群点上市公司表现出资产负债率较高、总资产净利率偏低的特点,为识别可能的财务报表舞弊疑点提供了线索。本研究有助于科学界定重点监管的范围,精准锁定舞弊疑点对象,从而为监管部门甄别财务报表舞弊提供决策支持,对于构建现代化监管执法新模式具有重要的意义。