SAS软件：欺诈分析实战指南——实时监控与金融犯罪防控行动

需积分: 10 50 浏览量更新于2024-07-18 收藏 8.4MB PDF 举报

《欺诈分析与SAS®：特殊收藏》是一本由Bart Baesens撰写的专业书籍，于2017年由SAS Institute Inc.在北卡罗来纳州的Cary出版。该书专为欺诈分析师设计，展示了如何利用SAS软件的强大功能进行实时的数据监控和深度调查。书中涵盖了广泛的主题，特别关注了欺诈分析在金融犯罪、洗钱和恐怖主义等领域的应用。书中强调了SAS软件在欺诈检测中的关键作用，通过实际案例和理论相结合，读者可以学习到一系列技术，如异常检测、模式识别、预测建模以及机器学习方法。这些技术旨在帮助金融机构和其他组织提高对潜在欺诈行为的敏感度和响应速度，减少经济损失并维护市场秩序。作者通过严谨的学术研究和实践案例，深入剖析了如何有效地将SAS工具集成到欺诈风险管理框架中，确保数据安全性和合规性。此外，书中还包含了版权声明，强调了电子版本的获取和使用须遵守相关规定，以尊重知识产权。对于纸质版，未经SAS Institute Inc.事先书面许可，禁止任何形式的复制、存储或传输。而对于电子版，用户应受供应商当时设定的条款约束，同时明确指出非法扫描、上传和通过互联网未经授权分发本书是违法的，并鼓励读者支持正版著作。《欺诈分析与SAS®：特殊收藏》是一本既实用又具有深度的参考资料，适合那些希望提升欺诈检测技能的IT专业人士和数据分析人员，同时也是企业进行风险管理和合规审计的重要参考资料。通过阅读这本书，读者可以了解到如何利用SAS的强大功能来构建一套完整的欺诈预防和应对策略。

Display 6. Scenario Conditions – SAS10019

When configuring the output of the second window, select Group by NUM_SRC_KEY as shown in the

screen shot below. This causes the output of the second window to be the aggregated data from the

output of the first window. The aggregation is calculated at the scenario-entity number level stored in the

NUM_SRC_KEY column, in this case SAS10016 || Account and SAS10019 || Account described earlier.

Display 7. New Window #2

Since the input data from the first window contains transactions that match the conditions dictated by

more than one scenario, we must devise a way to funnel the transactions to the second window’s

scenarios appropriately. We can accomplish this by creating a new column for each rule and applying a

filter. By doing so, that column receives only the transactions relevant for the corresponding scenario as

rule gets applied and the column gets aggregated. For scenario SAS10016, we need the aggregated

amount of the transactions, the number of branches where the deposits were made, and the number of

deposits. To satisfy the rules, we create three columns for this scenario. Each column applies one of the

scenario conditions along with a filter where AML_SCENARIO = SAS10016.

As you can see in the screen shot below, filtering on the scenario prior to the aggregation process can be

accomplished by adding a condition to the relevant columns. In this case, Scenario SAS10016 has three

columns that require aggregation: currency_amount_sum, deposit_cnt, and branch_number_dcnt. By

applying this filter, their corresponding results include only the transactions that met the criteria for

SAS10016 from the previous window.

Display 8. Column Properties – SAS10016

For scenario SAS10019, we want to aggregate the currency amount of type Wire in and Wire out. We

have already filtered data from the first window such that it contains only wire based transactions for

SAS10019. To calculate the total Wire in amount, we create a column total_in_sum that aggregates the

currency_amount only when the transaction type is “C”, indicating credit activities. To calculate the total

Wire out amount, we create a column total_out_sum, where the transaction type is “D”, indicating debit

activities. As you can see from the screen shot below, two new columns have been created with the

corresponding conditions applied. We can see that we are only aggregating the transactions for

SAS10019 that met the criteria for SAS10019 in the previous window.

剩余104页未读，继续阅读

sinat_41581062

粉丝: 21
资源: 174

SAS软件：欺诈分析实战指南——实时监控与金融犯罪防控行动

IEEE-CIS-Fraud-Detection

IEEE-CIS Fraud Detection-数据集

SAS for Data Analysis

SAS编程1：基础

#第 2步：计算 表NotFraud 的相关性 # 这里计算 除 fraud列 的剩余7列的相关性 corr_Fraud = Fraud.loc[:, Fraud.columns != 'fraud'].corr() # 查看计算结果 corr_Fraud

# 统计是否被诈骗数量分布 num_fraud = data[data['fraud'] == 1.0].shape[0] # fraud = 1.0 表示 被诈骗数据 num_no_fraud = ??? # fraud = 0.0 表示 未被诈骗数据

df1 = data.query('fraud == 1') df2 = data.query('fraud == 0').sample(len(df1)) data = pd.concat([df1,df2],axis=0).reset_index(drop=True)

trafigura nickel fraud news

ieee-cis fraud detection knn

# 这里计算 除 fraud列 的剩余7列的相关性，因为 fraud是标签不算特征 corr_NotFraud = ??? # 查看计算结果 corr_NotFraud

最新资源

#第 2步：计算表NotFraud 的相关性 # 这里计算除 fraud列的剩余7列的相关性 corr_Fraud = Fraud.loc[:, Fraud.columns != 'fraud'].corr() # 查看计算结果 corr_Fraud

# 统计是否被诈骗数量分布 num_fraud = data[data['fraud'] == 1.0].shape[0] # fraud = 1.0 表示被诈骗数据 num_no_fraud = ??? # fraud = 0.0 表示未被诈骗数据

# 这里计算除 fraud列的剩余7列的相关性，因为 fraud是标签不算特征 corr_NotFraud = ??? # 查看计算结果 corr_NotFraud