金融类罪分析模型

模型的研究报告，对模型的算法、架构、训练数据和测试数据及测试结果进行了详细说明。完成构建针对11个罪名的类罪资金审查分析模型35个，包括：非法吸收公众存款罪、集资诈骗罪的7个模型；针对地下钱庄犯罪（洗钱罪、非法经营罪）的7个模型；针对骗取贷款罪、贷款诈骗罪的4个模型。针对组织、领导传销活动罪的5个模型；针对保险诈骗罪的4个模型；针对操纵证券市场罪的3个模型；针对诈骗罪（电信网络犯罪）的3个模型；针对帮助信息网络犯罪活动罪的2个模型。各个模型的资金分析准确率最低值为90.35%，模型最长运行时间4.4秒。

This research report elaborates on the model's algorithm, architecture, training data, test data, and test results in detail. A total of 35 capital review and analysis models targeting 11 categories of criminal charges have been developed, including: 7 models for the Crimes of Illegal Absorption of Public Deposits and Fraudulent Fundraising; 7 models for underground banking-related Crimes, namely the Crime of Money Laundering and the Crime of Illegal Business Operation; 4 models for the Crimes of Obtaining Loans by Deception and Loan Fraud; 5 models for the Crime of Organizing and Leading Pyramid Schemes; 4 models for the Crime of Insurance Fraud; 3 models for the Crime of Manipulating the Securities Market; 3 models for the Crime of Fraud (Telecommunications and Network Crimes); and 2 models for the Crime of Assisting in Information Network Criminal Activities. The minimum capital analysis accuracy rate of each model is 90.35%, and the longest running time across all models is 4.4 seconds.