Method for fund account identification and fund transaction relation network analysis of fund collator

Abstract

The invention provides a fund account identification and fund transaction relation network analysis method for a fund allocation person, and relates to the field of research and judgment of securities criminal suspicion clues. The fund account identification and analysis method for the investor comprises the following steps: S1, collecting and marking fund transaction data; s2, fund data cleaning, complementing and duplicate removal; s3, fund feature selection of a fund allocation person; s4, performing supervised learning, and determining an optimal classification model; s5, identifying and predicting the fund account of the investor, and outputting a result; s6, building a fund association model, and analyzing the fund transaction network relationship of the investor; and S7, performing intelligent comparative analysis on output results of the processes S5 and S6, and outputting a final result. The discriminating process of fund account identification of the investor comprises the following steps of: 1) selecting a proper data characteristic variable according to scene experience; 2) performing logistic regression algorithm fitting on the selected data features; and 3) calculating whether the classification result is true or false according to the fitting parameters of the predicted user data.

Description

technical field [0001] The invention relates to the field of research and judgment of clues of securities crime suspects, specifically a method for capital account identification of a distributor and a network analysis method for capital transaction relationships. Background technique [0002] The traditional data analysis of funds suspected of securities crimes mainly uses stand-alone analysis tools and excel, which is mainly divided into four steps. First, use Excel to calculate the transaction times, amount, time, etc. The capital account of the fund is selected, and then each capital account is manually analyzed one by one to judge the possibility of being a fund account of the distributor, and finally, these suspicious capital accounts of the distributor are manually analyzed layer by layer to find suspicious fund network relationships. There are five shortcomings in this identification and analysis method: first, the screening criteria are relatively simple, the capita...

AI Technical Summary

Problems solved by technology

There are five shortcomings in this identification and analysis method: first, the screening criteria are relatively simple, the capital feature dimension is single, the features are isolated from each other, and there is no good integration method; Relying on manual analysis of each fund account one by one, there is no quantitative standard, the accuracy rate is low, and the promotion is difficult. Third, it requires high resource investment. Due to the single statistical characteristics of the first step, there is no good integration method, which is harmful to the allocation of funds. Accounts lack specific quantitative standards, so a large amount of screening, identification and analysis work has been left to the stage of manual investigation, which requires a lot of manpower, material and financial resources. The speed is rare in history, the source of capital transaction data is diverse, the structure is diverse, there is no unified standard, the data quality is uneven, and many core fields have problems such as missing items, missing items, and special characters. As a result, the stand-alone version of the tool cannot support actual business needs. Fifth, securities cases are highly specialized, and there is a serious lack of professionals with excellent business capabilities. It is particularly difficult to quickly and accurately identify the funder's account in securities cases and analyze the network relationship of the funder's capital transaction, which seriously restricts securities companies. Efficiency and quality of criminal suspect research and judgment

Images