CN-121563701-B - Enterprise financial management auxiliary method and system based on artificial intelligence

CN121563701BCN 121563701 BCN121563701 BCN 121563701BCN-121563701-B

Abstract

The invention belongs to the technical field of information technology, and particularly discloses an artificial intelligence-based enterprise financial management auxiliary method and system, wherein the method comprises the steps of obtaining multi-layer data from transaction records, processing characteristic differences according to hierarchical classification to obtain a hierarchical financial data set, adopting the hierarchical financial data set to analyze distribution rules, adjusting a unified threshold range if the distribution shows loose and missed report signs, determining an optimized threshold group, monitoring strict false report conditions according to the optimized threshold group, fusing time evolution factors through a self-adaptive threshold algorithm to obtain a dynamic threshold model, and obtaining risk degree indexes in the dynamic threshold model.

Inventors

CHEN JIANCAI
CHEN YUANZHAO
HUANG SUYE
Lai Maotao
HUANG SHUNYANG
LIN FU
LIN XIANG

Assignees

闽西职业技术学院

Dates

Publication Date: 20260508
Application Date: 20260126

Claims (8)

1. An artificial intelligence-based enterprise financial management assistance method, comprising: acquiring multi-layer data from the transaction records, and processing characteristic differences according to the hierarchical classification to obtain a hierarchical financial data set; adopting a layered financial data set to analyze a distribution rule, and if the distribution shows loose report signs, adjusting a unified threshold range to determine an optimized threshold group; The adoption of the hierarchical financial data set to analyze the distribution rule, if the distribution shows loose report signs, the unified threshold range is adjusted, and the optimized threshold group is determined, including: Acquiring the distribution characteristics of data of each layer through a layered financial data set, and primarily screening the distribution characteristics to obtain a distribution abnormal point set; Analyzing whether loose signs exist in the distribution abnormal point sets according to the distribution abnormal point sets, adopting a preset judging rule, and judging that the loose signs exist if the abnormal point occupation ratio exceeds a preset threshold value to obtain loose sign marks; aiming at the loose sign identification, acquiring corresponding missing report risk data, extracting key fields from the missing report risk data, and determining a potential missing report interval; Analyzing the relevance between the potential missing report interval and the financial distribution, and processing the mapping between the distribution and the missing report interval by adopting a logistic regression model to obtain a relevance evaluation result; According to the relevance evaluation result, adjusting a threshold interval, executing a unified adjustment strategy aiming at a high relevance area in the evaluation result, and determining an adjusted threshold range; obtaining an optimal combination scheme from the adjusted threshold range, and matching corresponding threshold combinations aiming at different layering financial data to obtain a final optimal threshold group; Updating the data layering rule through the final optimized threshold group, re-analyzing the financial distribution aiming at the updated rule, and determining a final distribution adjustment scheme; According to the strict false alarm condition of the optimized threshold group monitoring, the time evolution factors are fused through the self-adaptive threshold algorithm, and a dynamic threshold model is obtained; acquiring a risk degree index in the dynamic threshold model, triggering an anomaly detection algorithm if the index exceeds a service environment limit, and judging a potential anomaly point; extracting mass false alarm features from potential abnormal points, filtering irrelevant signals by adopting an abnormal detection algorithm, and determining a real risk sequence; tracking a dynamic change track through a real risk sequence, and updating a judgment standard according to a service evolution rule to obtain self-adaptive early warning parameters; and integrating multi-layer data feedback by adopting self-adaptive early warning parameters, and correcting an early warning system structure if the feedback shows time evolution deviation, so as to obtain final risk control output.
2. An artificial intelligence based enterprise financial management assistance method as claimed in claim 1, wherein said deriving a hierarchical financial data set from the hierarchical classification of process characteristic differences by obtaining multi-layered data from transaction records comprises: acquiring original multi-layer data through transaction records; Splitting the multi-layer data layer by adopting a hierarchical structure dividing method to obtain independent data sets of each layer; performing classification processing on each layer of independent data set to obtain a hierarchical classification result; Judging the interlayer difference type according to the hierarchical classification result to obtain a difference classification label; Extracting key difference fields from the difference classification labels to obtain a layered feature set; Matching the hierarchical feature set with each layer of independent data set to obtain a preliminary hierarchical financial table; performing a data integrity check on the preliminary hierarchical financial table results in a final hierarchical financial data set.
3. The artificial intelligence-based enterprise financial management assistance method according to claim 1, wherein the monitoring of strict false alarm conditions according to the optimized threshold group, the fusion of time evolution factors by the adaptive threshold algorithm, the obtaining of the dynamic threshold model, includes: continuously tracking the monitoring condition of the strict false alarm by optimizing a threshold value group, acquiring related false alarm data from a history record by adopting a data acquisition tool, and determining a preliminary false alarm distribution range; according to the preliminary false alarm distribution range and the time evolution factor, extracting key change points in the time sequence to obtain time-related false alarm fluctuation characteristics; Aiming at time-related false alarm fluctuation characteristics, a preset logistic regression model is applied to process the relation between fluctuation and a threshold value, and a corresponding threshold value adaptability adjustment direction is obtained; screening out an adjustment basis related to the strict false alarm height from the threshold adaptive adjustment direction, and if the display false alarm frequency of the adjustment basis exceeds a preset threshold, generating a corresponding dynamic threshold updating scheme; through a dynamic threshold updating scheme and combining with adaptive requirements, parameter optimization is carried out on the existing threshold algorithm, and updated threshold configuration combination is determined; According to the updated threshold configuration combination, aiming at false alarm analysis results of different time periods, constructing an operation rule of a dynamic model to obtain a final threshold application frame; And acquiring a final threshold application frame, comparing abnormal points in the monitoring condition in real time, and triggering a threshold adjustment mechanism if the abnormal points deviate from a preset range, so as to judge the adjusted monitoring stability.
4. The artificial intelligence based enterprise financial management assistance method of claim 1, wherein the acquiring the risk level indicator in the dynamic threshold model, if the indicator exceeds the service environment limit, triggers an anomaly detection algorithm, and determines a potential anomaly point, includes: Obtaining a risk degree index from a dynamic threshold model, and primarily screening index data to obtain an abnormal candidate set exceeding a preset limit; according to the abnormal candidate set, a pre-established classification model is adopted to classify the data points in the candidate set, and the important attention objects belonging to potential abnormalities are determined; For the important attention objects, extracting corresponding environment limit data from a service environment, acquiring time sequence information related to an abnormal position, and judging whether a continuous deviation phenomenon exists or not; if the time sequence information shows the continuous deviation phenomenon, starting an abnormality detection flow, acquiring a specific distribution range of abnormal positions, and determining a core area of abnormal points; extracting frequency characteristics of abnormal occurrence by combining the data of the core area with the history record in the detection flow to obtain a periodic mode of abnormal behavior; After the periodic mode is acquired, aiming at a key time point in the mode, applying a logic judging tool, if the index value of the key time point exceeds the environment limit, generating a corresponding abnormal mark, and determining a final abnormal confirmation result; And updating a risk degree evaluation rule in the threshold model through an abnormal confirmation result to obtain an adjusted index monitoring frame, and finishing continuous tracking of potential abnormalities.
5. The artificial intelligence based enterprise financial management assistance method of claim 1, wherein the extracting mass false positive features from potential outliers, filtering extraneous signals using an anomaly detection algorithm, determining a real risk sequence, comprises: acquiring abnormal signals from a service system, primarily classifying the signals through a pre-established classification model, and filtering out obvious interference information to obtain a primarily cleaned signal set; aiming at the initially cleaned signal set, carrying out deep analysis on the signals by adopting an anomaly detection algorithm, and identifying false alarm data contained in the signals to obtain a refined anomaly signal group; extracting key information associated with the service from the refined abnormal signal group, and carrying out structural arrangement on the signal group through an information processing tool to determine a core signal of the potential problem; Acquiring corresponding service environment data according to the core signal, and marking the core signal as a key attention object in a risk sequence if the core signal has significant deviation from the environment data; Aiming at the important attention object, acquiring historical time sequence information of the important attention object, judging whether a continuous deviation phenomenon exists or not through a comparison analysis tool, and obtaining a judgment result of the deviation state; according to the judging result of the deviation state, if the judging result shows continuous deviation, the priority ordering is carried out on the continuous deviation, and a high-priority risk sequence is determined; and aiming at the risk sequence with high priority, adopting an information processing tool to generate a corresponding abnormal mark, and finishing tracking processing of the potential problem through mark record.
6. The artificial intelligence-based enterprise financial management assistance method according to claim 1, wherein the tracking of the dynamic change track through the real risk sequence, updating the judgment standard according to the business evolution rule, and obtaining the adaptive early warning parameters comprises: Acquiring dynamic change data of a risk sequence from a service system, and carrying out structural arrangement on a change path by adopting an information extraction tool to obtain path information after preliminary arrangement; aiming at the path information after preliminary arrangement, comparing the evolution trend of the service through a pre-established service rule base, judging whether the path information accords with the expected direction of the evolution trend, and if so, determining the path information as a key change path; acquiring a corresponding business adjustment record according to the key change path, and analyzing the relevance between the change path and business adjustment by adopting a data matching tool to obtain a relevance analysis result; Aiming at the correlation analysis result, if the correlation analysis result shows that the change path is consistent with the business adjustment, updating the evaluation basis through the information processing module to obtain an updated evaluation standard; According to the updated evaluation standard and the adaptive requirements, an information mapping tool is adopted to adjust the early warning parameters, and an adjusted parameter set is determined; binding the adjusted parameter set with a business evolution trend through a data storage tool, acquiring a bound parameter application range, and judging whether all key change paths are covered or not; And according to the bound parameter application range, if the coverage meets the requirement of service adjustment, applying parameter updating to a service system through a data synchronization tool to obtain a final self-adaptive early warning parameter.
7. The artificial intelligence based enterprise financial management assistance method of claim 1, wherein the integrating the multi-layer data feedback using adaptive early warning parameters, if the feedback shows a time evolution deviation, modifying the early warning architecture to obtain a final risk control output, comprises: acquiring multi-layer time sequence change information from a service real-time data stream to obtain an original sequence set; Extracting sequence change characteristics of each layer by adopting a time window dividing method aiming at an original sequence set to obtain a layered change characteristic set; according to the layered change feature set, calculating time deviation values between each layer sequence and the reference sequence to obtain a multi-layer deviation index set; If at least one deviation value in the multi-layer deviation index set exceeds a preset threshold value, marking the corresponding layer as an abnormal feedback layer to obtain an abnormal feedback layer set; aiming at the abnormal feedback layer set, adjusting the weight ratio of each layer in the early warning calculation structure by a proportion weighting method to obtain an adjusted calculation structure; according to the adjusted calculation structure, re-calculating the early warning parameter value by combining the current real-time data stream to obtain updated self-adaptive early warning parameters; and carrying out risk judgment on the current business real-time data stream by adopting the updated self-adaptive early warning parameters to obtain final risk control output.
8. An artificial intelligence based enterprise financial management assistance system, the system comprising: the data acquisition and layering processing module is used for acquiring multi-layer data from the transaction records and processing characteristic differences according to the level classification to obtain a layering financial data set; the threshold optimization module is used for analyzing a distribution rule by adopting a layered financial data set, and if the distribution shows loose report signs, adjusting a unified threshold range and determining an optimized threshold group; The adoption of the hierarchical financial data set to analyze the distribution rule, if the distribution shows loose report signs, the unified threshold range is adjusted, and the optimized threshold group is determined, including: Acquiring the distribution characteristics of data of each layer through a layered financial data set, and primarily screening the distribution characteristics to obtain a distribution abnormal point set; Analyzing whether loose signs exist in the distribution abnormal point sets according to the distribution abnormal point sets, adopting a preset judging rule, and judging that the loose signs exist if the abnormal point occupation ratio exceeds a preset threshold value to obtain loose sign marks; aiming at the loose sign identification, acquiring corresponding missing report risk data, extracting key fields from the missing report risk data, and determining a potential missing report interval; Analyzing the relevance between the potential missing report interval and the financial distribution, and processing the mapping between the distribution and the missing report interval by adopting a logistic regression model to obtain a relevance evaluation result; According to the relevance evaluation result, adjusting a threshold interval, executing a unified adjustment strategy aiming at a high relevance area in the evaluation result, and determining an adjusted threshold range; obtaining an optimal combination scheme from the adjusted threshold range, and matching corresponding threshold combinations aiming at different layering financial data to obtain a final optimal threshold group; Updating the data layering rule through the final optimized threshold group, re-analyzing the financial distribution aiming at the updated rule, and determining a final distribution adjustment scheme; the dynamic threshold modeling module is used for monitoring strict false alarm conditions according to the optimized threshold group and obtaining a dynamic threshold model by fusing time evolution factors through a self-adaptive threshold algorithm; the anomaly detection triggering module is used for acquiring the risk degree index in the dynamic threshold model, triggering an anomaly detection algorithm if the index exceeds the service environment limit, and judging a potential anomaly point; The real risk sequence determining module is used for extracting mass false alarm features from potential abnormal points, filtering irrelevant signals by adopting an abnormal detection algorithm, and determining a real risk sequence; The self-adaptive early warning parameter generation module is used for tracking a dynamic change track through a real risk sequence, updating a judgment standard according to a service evolution rule and obtaining self-adaptive early warning parameters; and the risk control output module is used for integrating multi-layer data feedback by adopting the self-adaptive early warning parameters, and correcting the early warning system structure if the feedback shows time evolution deviation so as to obtain final risk control output.

Description

Enterprise financial management auxiliary method and system based on artificial intelligence Technical Field The invention relates to the technical field of information technology, in particular to an enterprise financial management auxiliary method and system based on artificial intelligence. Background Enterprise financial management is a core support for modern corporate operations, directly related to fund security, operational decisions, and long-term survival and development. As enterprise scale increases and business complexity continues to rise, the concealment, linkage, and cross-level propagation characteristics of financial anomalies become increasingly significant, making timely discovery and effective control of risk a significant challenge that all enterprises must face. When facing multi-level financial data, most of the current financial monitoring means commonly have a contradiction which is difficult to avoid, if the same judgment standard is uniformly used for all anomalies, a large amount of real risks are missed due to too loose judgment, or a large amount of false alarm flooding is caused due to too strict judgment, so that no effective information is available. The threshold setting mode of the 'one-cut' is difficult to simultaneously adapt to the inherent rules and actual business demand differences of different levels of data. The transaction records, accounting subjects, department reports and enterprise overall reports are in four levels, the granularity of information carried by the transaction records, the fluctuation characteristics and the risk meanings are completely different, but the transaction records, the accounting subjects, the department reports and the enterprise overall reports are often processed under the same monitoring frame, and the result is that the system is either insensitive or frequently warned, so that the judging efficiency of financial staff on real risks is seriously affected. A more critical issue is that the true risk level of financial anomalies dynamically changes over time, business environment and historical performance. In the past, the normal transaction amount range may become a high risk signal in a specific period, the cost fluctuation characteristic of a certain department for several months may indicate that the management vulnerability is gradually expanding, and certain key indexes of the whole enterprise may have early hidden danger signals even if the key indexes are still in a so-called safe interval. These changes are not occasional, but are necessarily driven by multiple factors such as business pace, seasonal laws, policy adjustments, market fluctuations, etc. A fixed threshold cannot keep pace with this constant evolving reality, resulting in a system that often fails at critical times. Therefore, how to establish different abnormal judgment standards capable of continuously self-adjusting along with time and detection effect according to unique distribution characteristics and specific business scenes of different levels of financial data becomes a key problem for constructing a truly intelligent and efficient enterprise financial risk early warning system. Disclosure of Invention The invention provides an artificial intelligence-based enterprise financial management auxiliary method and system, and aims to solve the problem of how to respectively establish different abnormal judgment standards capable of being continuously and self-adjusted along with time and detection effect according to unique distribution characteristics and specific business scenes of financial data of different levels in the prior art, and the method and system become a real intelligent and efficient enterprise financial risk early warning system. In order to solve the technical problems, the invention adopts the following technical scheme: An enterprise financial management auxiliary method based on artificial intelligence comprises the steps of obtaining multi-layer data from transaction records, processing characteristic differences according to level classification to obtain a layered financial data set, adopting the layered financial data set to analyze distribution rules, adjusting a unified threshold range if the distribution shows loose and missed report signs, determining an optimized threshold group, monitoring strict and wrong report conditions according to the optimized threshold group, fusing time evolution factors through an adaptive threshold algorithm to obtain a dynamic threshold model, obtaining a risk degree index in the dynamic threshold model, triggering an anomaly detection algorithm to judge potential abnormal points if the index exceeds a service environment limit, extracting mass misinformation features from the potential abnormal points, filtering irrelevant signals by adopting the anomaly detection algorithm to determine a real risk sequence, tracking dynamic change tracks through the real risk sequence, updating judgment standa