Abnormal data identification method and device based on artificial intelligence, equipment and medium

Abstract

The invention relates to an artificial intelligence technology, can be applied to the technical field of finance, and discloses an artificial intelligence-based abnormal data identification method, device, equipment and medium, and the method comprises the steps: collecting investment data in real time, and inputting the investment data into an investment data semantic large model; performing anomaly identification on the investment data based on the investment data semantic large model to generate an abnormal data identification result; obtaining abnormal investment data in the abnormal data identification result, analyzing an abnormal condition of the abnormal investment data based on a graph neural network algorithm, and generating an abnormal reason and abnormal positioning information; and obtaining a repair suggestion based on the exception reason, and pushing the exception reason, the exception positioning information and the repair suggestion to a visual interface. According to the invention, investment data is scanned in real time through the AI algorithm, abnormal values are rapidly identified, and human errors are reduced; abnormal reasons can be automatically positioned, repair suggestions can be provided, real-time monitoring is achieved, and abnormal data recognition efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to an abnormal data identification method, apparatus, device, and medium based on artificial intelligence. Background Technology

[0002] In the financial sector, investment performance evaluation, risk control and early warning, and asset allocation all rely on the latest data on holdings, returns, and transactions for investment analysis, decision-making, and performance measurement. These analyses and decisions demand consistent data standards and high accuracy. Currently, most financial institutions still rely on relatively traditional technologies in this area, primarily using manual methods and fixed rule-based verification, which has several shortcomings.

[0003] Traditional data verification relies heavily on daily item-by-item checks by IT operations and investment management personnel. Faced with different asset types and multiple dimensions of indicators (such as returns, holdings, and transaction volumes), manual review is inefficient and prone to errors. This is especially true for large financial companies, where the sheer size of assets, the breadth of investments, the volume of transactions, and the complexity of data sources lead to a dramatic increase in data volume and complexity, making it difficult for manual verification to meet the requirements of real-time performance and accuracy. Most systems lack intelligent anomaly detection mechanisms, relying primarily on static rules (such as threshold judgments) to identify anomalies. This method has a low hit rate, high false positive rate, and is ill-suited to dynamic scenarios. Furthermore, inconsistent data structures prevent the formation of an integrated, intelligent analysis pipeline.

[0004] Therefore, existing methods for identifying abnormal data generally rely on manual review. Manual review is inefficient, prone to errors, and cannot meet the requirements of real-time performance and accuracy. Existing anomaly identification mechanisms use static rules for capture, resulting in low hit rates, high false alarms, and difficulty in adapting to dynamic scenarios, thus leading to low accuracy in identifying abnormal data. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the present invention provides an artificial intelligence-based method, apparatus, device and medium for anomaly data identification, aiming to solve the problems of existing anomaly data identification methods, which generally adopt manual review methods. Manual review is inefficient, prone to errors, and cannot meet the requirements of real-time and accuracy. Existing anomaly identification mechanisms use static rules for capture, which have low hit rate, high false alarm rate and are difficult to adapt to dynamic scenarios, thus resulting in low accuracy of anomaly data identification.

[0006] The technical solution of the present invention is as follows:

[0007] The first embodiment of the present invention provides an artificial intelligence-based method for identifying abnormal data, the method comprising:

[0008] Real-time acquisition of investment data, and input of the investment data into the investment data semantic big data model;

[0009] Based on the semantic big data model of the investment data, anomaly identification is performed on the investment data to generate anomaly identification results;

[0010] Obtain abnormal investment data from the abnormal data identification results, analyze the abnormal situation of the abnormal investment data based on the graph neural network algorithm, and generate abnormal cause and abnormal location information;

[0011] Based on the cause of the anomaly, repair suggestions are obtained, and the cause of the anomaly, anomaly location information, and repair suggestions are pushed to the visualization interface.

[0012] Another embodiment of the present invention provides an artificial intelligence-based anomaly data identification device, the device comprising:

[0013] The data acquisition module is used to collect investment data and input the investment data into the investment data semantic big data model;

[0014] Anomaly identification module is used to identify anomalies in the investment data based on the semantic big data model of the investment data and generate anomaly identification results;

[0015] An anomaly information location module is used to acquire abnormal investment data from the abnormal data identification results, analyze the anomalies in the abnormal investment data based on a graph neural network algorithm, and generate anomaly cause and anomaly location information.

[0016] The data push module is used to obtain repair suggestions based on the cause of the anomaly, and push the cause of the anomaly, the anomaly location information, and the repair suggestions to the visualization interface.

[0017] Another embodiment of the present invention provides a computer device, the computer device including at least one processor; and,

[0018] A memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the steps of the artificial intelligence-based anomaly data identification method described above.

[0020] Another embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the above-described artificial intelligence-based abnormal data identification method.

[0021] Beneficial Effects: The artificial intelligence-based abnormal data identification method, apparatus, device, and medium of this invention include: real-time acquisition of investment data; inputting the investment data into an investment data semantic big data model; performing anomaly identification on the investment data based on the investment data semantic big data model to generate anomaly identification results; acquiring abnormal investment data from the anomaly identification results; analyzing the anomalies of the abnormal investment data based on a graph neural network algorithm to generate anomaly causes and anomaly location information; obtaining repair suggestions based on the anomaly causes; and pushing the anomaly causes, anomaly location information, and repair suggestions to a visualization interface. This invention uses an AI algorithm to scan investment data in real time, quickly identifying outliers (such as data delays, outliers, missing values, etc.), reducing human error, and avoiding investment losses due to data errors. Combining the big data model and graph neural network, the system can automatically locate the causes of anomalies (such as data source problems, calculation logic errors, etc.) and provide repair suggestions. It runs automatically daily without manual intervention, significantly reducing the time spent on manual verification and repair. Through real-time data monitoring and anomaly alarms, managers can quickly respond to problems and avoid investment decision errors caused by data delays or errors. Through automated anomaly detection and repair, the system can continuously improve data quality, providing reliable support for long-term investment decisions. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram illustrating the application environment of an embodiment of an artificial intelligence-based abnormal data identification method of the present invention;

[0024] Figure 2 This is a flowchart of a preferred embodiment of an artificial intelligence-based abnormal data identification method of the present invention;

[0025] Figure 3 This is a schematic diagram of the functional modules of a preferred embodiment of an artificial intelligence-based abnormal data identification device of the present invention;

[0026] Figure 4 This is a schematic diagram of a preferred embodiment of a computer device according to the present invention;

[0027] Figure 5 This is another structural schematic diagram of a preferred embodiment of a computer device according to the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention is further described in detail below. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0029] The embodiments of the present invention will be described below with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0030] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Here, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0031] The method provided in this application can be applied to artificial intelligence (AI) scenarios. AI is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. In other words, artificial intelligence is a branch of computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine capable of reacting in a way similar to human intelligence. Artificial intelligence studies the design principles and implementation methods of various intelligent machines, enabling them to possess perception, reasoning, and decision-making functions. Research in the field of artificial intelligence includes robotics, natural language processing, computer vision, decision-making and reasoning, human-computer interaction, recommendation and search, and fundamental AI theories.

[0032] The artificial intelligence-based abnormal data identification method provided in this invention can be applied to, for example... Figure 1In this application environment, the client communicates with the server via a network. The client accesses the server's network or business platform, and the server can collect investment data in real time, inputting the investment data into an investment data semantic model; based on the investment data semantic model, it performs anomaly identification on the investment data, generating anomaly identification results; it obtains abnormal investment data from the anomaly identification results, analyzes the anomalies of the abnormal investment data based on a graph neural network algorithm, and generates anomaly causes and anomaly location information; based on the anomaly causes, it obtains repair suggestions, and pushes the anomaly causes, anomaly location information, and repair suggestions to a visual interface. In this invention, by using AI algorithms to scan investment data in real time, anomalies are quickly identified, reducing human error; it can automatically locate the causes of anomalies and provide repair suggestions, achieving real-time monitoring and improving the efficiency of anomaly data identification. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0033] To address the aforementioned problems, embodiments of the present invention provide an artificial intelligence-based method for identifying abnormal data. Please refer to [link to relevant documentation]. Figure 2 , Figure 2 This is a flowchart of a preferred embodiment of an artificial intelligence-based abnormal data identification method of the present invention. Figure 2 As shown, it includes:

[0034] Step S100: Collect investment data in real time and input the investment data into the investment data semantic big model.

[0035] This invention primarily aims to replace traditional rule-driven and manual review mechanisms for auditing investment data. This investment data includes, but is not limited to, asset tag data, valuation data, transaction data, and return data. It also includes, but is not limited to, market data, transaction data, valuation data, and accounting data. The extracted data is input into a pre-trained semantic model of investment data. The semantic model of investment data is constructed based on this semantic model.

[0036] The Investment Data Semantic Big Model is an intelligent model that combines Natural Language Processing (NLP) and deep learning technologies, specifically designed for parsing, understanding, and generating investment-related text data. The Semantic Big Model is a high-order model built on massive amounts of text data using deep learning algorithms (especially NLP techniques), capable of understanding, generating, and reasoning about human language. It employs a Transformer architecture and uses a self-attention mechanism to dynamically calculate the correlation strength between words in the text, achieving global semantic understanding. For example, when analyzing "equipment aging leads to increased energy consumption," the model can automatically identify the causal relationship between "equipment aging" and "increased energy consumption." The training mechanism includes: Pre-training: Unsupervised learning is performed on unlabeled text data (such as books, web pages, and papers) to master language rules and common sense by predicting occluded words (e.g., "The capital of country X is [MASK]"). Fine-tuning: Secondary training is conducted using domain-specific data (such as legal documents and medical literature) to give the model professional capabilities, such as transforming a general-purpose model into a financial analysis tool. It also incorporates visual and auditory information to form cross-modal understanding capabilities. For example, the model can process image descriptions and text dialogues simultaneously, enabling richer interactions.

[0037] Step S100, namely, real-time acquisition of investment data and inputting the investment data into the investment data semantic big data model, includes:

[0038] Step S101: Collect investment data from several investment data systems in real time;

[0039] Step S102: Obtain the anomaly detection threshold corresponding to the financial business type;

[0040] Step S103: Construct an initial large semantic model based on long short-term memory networks, attention mechanisms, and the isolated forest algorithm;

[0041] Step S104: Obtain the time series dataset of investment data, and train the initial semantic big model based on the time series dataset and the anomaly detection threshold. When the training is completed, the investment data semantic big model is generated.

[0042] In this embodiment of the invention, data is extracted in real time from multiple systems such as market data, trading data, valuation data, and accounting data. Anomaly detection thresholds corresponding to different financial business types are obtained; these thresholds are dynamic and can be set differently for different business types.

[0043] A large-scale semantic model for investment data (Finance Data LLM) is pre-built. This model helps understand the logical relationships between different asset types and field structures; it also analyzes potential anomalies using upstream and downstream logical chains. During the construction of the investment data semantic model, various algorithms, including self-supervised learning, Isolation Forest, and LSTM+attention mechanisms, are introduced to model the temporal and lateral structures of the data; detection thresholds are dynamically adjusted by integrating financial business rules. During the detection process, machine learning algorithms such as Isolation Forest and AutoEncoder are used for identification, supplemented by a rule engine for business constraint judgment.

[0044] Self-supervised learning (SSL) is a machine learning method that generates supervisory signals from the data itself by designing pre-training tasks, without requiring manual data labeling. Its core idea is to utilize the inherent structure of the data (such as temporal, spatial, and semantic relationships) to construct surrogate targets, enabling the model to learn transferable, general feature representations that can then serve downstream tasks (such as classification, detection, and generation).

[0045] The Isolation Forest algorithm is a highly efficient unsupervised anomaly detection algorithm that isolates outliers by constructing random trees, making it particularly suitable for handling high-dimensional and large-scale datasets. The isolation algorithm assumes that outliers have two key characteristics: scarcity (typically accounting for no more than 10% of the dataset) and dissimilarity (attribute values ​​significantly different from normal points, often located in the marginal regions of the data distribution). Based on these two points, the algorithm assumes that outliers are easier to quickly "isolate" (i.e., separated from the data through a small number of splitting operations). The construction method of the Isolation Tree (iTree) is as follows: Random splitting: Randomly sample a subset from the dataset, recursively and randomly select features and split values ​​(such as the median value of numerical features) to divide the data space, forming a binary tree structure. Termination condition: Segmentation stops when a data point is completely isolated (i.e., falls into a leaf node) or the tree reaches a preset height. Path length: The number of edges from the root node to a leaf node (i.e., the number of splits) reflects its degree of "isolation." Outliers, being far from dense areas, usually have shorter path lengths. Anomaly score calculation: Randomness is reduced by ensemble of multiple isolated trees, the average path length of data points is calculated and normalized to anomaly score.

[0046] LSTM (Long Short-Term Memory) combined with attention mechanism is a powerful sequence modeling method. By capturing long-distance dependencies through LSTM and then using attention mechanism to dynamically focus on key information, it significantly improves the performance of the model in tasks such as natural language processing (NLP) and time series prediction.

[0047] Step S200: Based on the semantic big data model of the investment data, perform anomaly identification on the investment data and generate anomaly identification results.

[0048] The investment data is analyzed within the constructed semantic model to determine if any anomalies exist. If anomalies are found, the system outputs an anomaly detection result. If no anomalies are found, the system outputs a normal detection result.

[0049] Step S200, which involves identifying anomalies in the investment data based on the semantic big data model and generating anomaly identification results, includes:

[0050] Step S201: Based on the semantic big data model of investment data, obtain the logical relationships between different asset types and field structures in the investment data;

[0051] Step S202: Obtain the corresponding dynamic detection threshold based on the asset type of the investment data;

[0052] Step S203: Generate upstream and downstream logical chains based on the logical relationships, identify the investment data according to the upstream and downstream logical chains and the dynamic detection threshold, and generate abnormal data identification results.

[0053] The asset types in the investment data can be divided into five categories based on liquidity and risk-return characteristics. The core fields and logical relationships for each asset category are as follows:

[0054] Core data points for cash-like assets include: currency unit (CNY / USD), holdings, yield, and liquidity rating. The logical relationship is that the yield is negatively correlated with GDP growth rate and inflation rate (e.g., if the yield of money market funds is lower than the GDP growth rate, the allocation ratio needs to be adjusted). The liquidity rating directly affects the proportion of emergency funds reserves.

[0055] Core fields for fixed-income assets: bond type (government bonds / corporate bonds), coupon rate, yield to maturity, duration, and credit rating. Logical relationships: duration is positively correlated with interest rate risk. Credit rating is negatively correlated with default risk. The difference between the coupon rate and the market interest rate determines the premium / discount.

[0056] Core fields for equity assets include: stock code, industry classification, market capitalization, price-to-earnings ratio (PE), price-to-book ratio (PB), return on equity (ROE), and dividend payout ratio. Logical relationships: PE is positively correlated with industry growth (e.g., technology stocks generally have higher PE ratios than consumer stocks). ROE is positively correlated with shareholder return capability (e.g., companies with ROE > 15% have long-term investment value). Dividend payout ratio is directly related to dividend yield.

[0057] Key fields for alternative assets: Asset type (gold / crude oil / agricultural products), contract unit, margin ratio, volatility, and basis. Logical relationships: Margin ratio is negatively correlated with leverage. Basis is related to arbitrage opportunities. Volatility is positively correlated with risk hedging needs (e.g., when crude oil volatility rises, the option hedging ratio needs to be increased).

[0058] Core fields of data assets: data type (finance / e-commerce / healthcare), data volume, update frequency, data quality score, and application scenario. Logical relationships: Data quality score is positively correlated with monetization value (e.g., datasets with a score >80 can be sold at a premium). Update frequency matches real-time decision-making needs (e.g., high-frequency trading requires millisecond-level data updates). Application scenario determines the pricing model (e.g., milestone payments can be used when medical data is used for drug development).

[0059] The semantic big data model can perform field association rule mining during identification. By analyzing historical data, the big data model can identify implicit relationships between fields (e.g., "high volatility assets" and "low liquidity assets" often co-occur in alternative investment portfolios). Example: Input "gold price and US dollar index", the model outputs a negative correlation coefficient of -0.8 and generates a trend chart. It can also verify asset allocation logic. Input an allocation plan (e.g., "60% stocks + 30% bonds + 10% cash"), and the model can verify whether the risk-reward ratio meets the target (e.g., whether the Sharpe ratio is >1.2). Example: The model suggests, "The current plan has too high a proportion of technology stocks; it is recommended to increase consumer stocks to hedge industry risk." It can also detect abnormal data. By comparing fields within reasonable ranges, the model can identify data errors (e.g., "a bond with a credit rating of AAA but a yield of 10%" may be an input error). Example: The model marks "a stock's price-to-earnings ratio is negative," requiring verification of financial statements.

[0060] Anomaly source analysis based on the upstream and downstream logical chain of investment data requires starting from the complete path of data generation from the source to the final application, and identifying risk points in each link that may disrupt logical consistency.

[0061] The investment data logic chain can be divided into five stages, each with specific dependencies:

[0062] The logical chain of the data generation stage: business event (such as a transaction occurring) → system record (such as the order system generating transaction data) → preliminary verification (such as fields not being empty and the format being valid); key dependency: the matching between business rules (such as "bond transactions need to record the coupon rate") and system functions (such as whether the API can capture real-time data).

[0063] The logical chain of data transmission is as follows: source system (e.g., exchange) → transmission channel (e.g., leased line / Internet) → target system (e.g., risk control platform); key dependencies are: network stability (e.g., latency <100ms) and data integrity (e.g., no packet loss).

[0064] The logical chain of the data processing stage: raw data → cleaning (such as deduplication and error correction) → transformation (such as unit unification) → calculation (such as Sharpe ratio) → storage; key dependencies: correctness of calculation rules (such as no errors in formulas) and accuracy of field mapping (such as "PE" corresponding to "price-earnings ratio" instead of "price-book ratio").

[0065] The logical chain of the data analysis stage: storing data → model input (such as machine learning features) → strategy generation (such as asset allocation recommendations) → decision output; key dependencies: the rationality of model assumptions (such as whether the "market efficient" assumption holds) and the compatibility of strategy constraints (such as whether "the proportion of high volatility assets ≤ 15%" conflicts with "tech stocks must be allocated 20%").

[0066] The logical chain of data application stage: decision output → execution system (such as trading terminal) → market feedback (such as transaction report) → effect evaluation (such as drawdown monitoring); key dependencies: execution efficiency (such as order delay <1 second) and market environment stability (such as no black swan events).

[0067] Step S300: Obtain abnormal investment data from the abnormal data identification results, analyze the abnormality of the abnormal investment data based on a graph neural network algorithm, and generate abnormality cause and abnormality location information. The abnormality location information is the distribution location of the abnormality.

[0068] Graph neural networks are used to analyze the dependency paths between data, and large models are employed to analyze common causes, generating anomaly causes and anomaly location information. Anomalies are caused by abnormalities in investment data, such as anomalies during the data generation stage. Type 1: Business rules not covered; Example: A private equity fund added a "digital currency" asset type, but the order system did not require recording the "wallet address," making it impossible to trace the flow of funds during subsequent audits. Impact: Increased compliance risk, potentially facing regulatory penalties.

[0069] Type 2: System Functionality Defect. Example: A bond trading system failed to capture the "early redemption clause" field, leading to an error in duration calculation. Impact: The interest rate risk of fixed-income portfolios was underestimated, resulting in actual drawdowns exceeding expectations.

[0070] Step S300 involves acquiring abnormal investment data from the abnormal data identification results, analyzing the abnormality of the abnormal investment data based on a graph neural network algorithm, and generating abnormality cause and abnormality location information, including:

[0071] Step S301: Obtain the abnormal data identification result and determine whether there is abnormal investment data in the abnormal data identification result. If there is abnormal investment data, proceed to step S302; if there is no abnormal investment data, proceed to step S303.

[0072] Step S302: Analyze the abnormal situation of the abnormal investment data based on the graph neural network algorithm, generate the cause of the abnormality, locate the distribution of the abnormal investment data, and generate abnormal location information;

[0073] Step S303: Push the abnormal data identification results to the visualization interface.

[0074] This invention, in its embodiment, determines whether there is abnormal investment data in the abnormal data identification results. If abnormal investment data is found, a graph neural network is used to analyze the causes of the abnormal investment data and generate anomaly reasons. Simultaneously, the distribution of abnormal investment data is located, generating anomaly location information. If no abnormal investment data is found, the abnormal data identification result is directly fed back and pushed to a visualization interface. Graph Neural Networks (GNNs) are a class of deep learning models specifically designed for processing graph-structured data. By iteratively aggregating information about the node itself and its neighbors, they learn low-dimensional representations (Embeddings) of nodes, thereby supporting tasks such as node classification, graph classification, and link prediction. Graph-structured data includes nodes (representing entities, such as users in a social network or atoms in a molecule), edges (representing relationships between nodes, such as user follows or chemical bonds between atoms), and attributes (features that nodes or edges can carry, such as user age or atom type). The challenge is that graph data has non-Euclidean properties (the number and arrangement of node neighbors are variable), making traditional CNNs / RNNs unsuitable for direct application. GNN's message passing mechanism: Each node updates its own representation by aggregating neighbor information, and reaches a stable state after multiple iterations.

[0075] Step S302 involves analyzing the anomalies in the abnormal investment data using a graph neural network algorithm, generating the causes of the anomalies, locating the distribution of the abnormal investment data, and generating anomaly location information, including:

[0076] Step S321: Establish a data dependency graph based on the investment data;

[0077] Step S322: Analyze the abnormal reasons for the abnormal investment data based on graph neural network algorithm and link analysis algorithm, and locate the abnormal data link corresponding to the abnormal data;

[0078] Step S323: Analyze the cause of the abnormal data link and generate the cause of the abnormality;

[0079] Step S324: Locate the distribution of the abnormal data links and generate abnormal location information.

[0080] Establish a data dependency graph and utilize Graph Neural Networks (GNNs) + link analysis to automatically locate error chains. A data dependency graph is a visualization tool that graphically displays the flow path, processing, and dependencies of data from source to destination. It helps enterprises track the origin and destination of data, ensure data quality and compliance, and optimize data governance processes. The data dependency graph constructs a directed graph through nodes (data assets) and edges (data flow relationships), clearly presenting the flow trajectory of data between systems, applications, and processes. This includes: Traceability: Quickly locate the root cause of data problems (e.g., if a report has incorrect data, it can be traced back to the original data source or processing logic). Compliance Audit: Meet relevant regulatory requirements for data transparency and prove the legality of data processing. Impact Analysis: Assess the impact of data changes on downstream systems (e.g., which reports might be affected by modifying database fields). Resource Optimization: Identify redundant data flows and reduce storage and computing costs. Node types in the data dependency graph include data sources: original data sources (e.g., database tables, API interfaces, file systems). Processing nodes: Data transformation, cleaning, aggregation, and other operations (e.g., SQL scripts, machine learning models). Data destination: The final storage or consumption location (e.g., data warehouse, application system). Metadata: Information describing data attributes (e.g., field type, update frequency, owner). Edge relationships include data flow direction: unidirectional or bidirectional relationships from source to destination. Dependencies: The dependencies of processing nodes on input data. Triggering conditions: The rules that trigger data flow (e.g., scheduled tasks, event-driven processes).

[0081] Based on graph neural network algorithm and link analysis algorithm, the abnormal reasons of this abnormal investment data are analyzed and the abnormal data link corresponding to the abnormal data is located; the abnormal reasons of the abnormal data link are analyzed and the abnormal reasons are generated; the distribution of abnormal data link is located and abnormal location information is generated.

[0082] Step S400: Obtain repair suggestions based on the cause of the anomaly, and push the cause of the anomaly, the anomaly location information, and the repair suggestions to the visualization interface.

[0083] Provide a visual interface. For example, visualization can be achieved using a data anomaly dashboard; for common error types, the system can provide repair suggestions (such as retrieving data from the source system or re-triggering the calculation process). The identified cause of the anomaly, anomaly location information, and repair suggestions are all pushed to the visual interface. Through real-time data monitoring and anomaly alerts, managers can quickly respond to problems and avoid investment decision errors caused by data delays or errors.

[0084] Step S400, which involves obtaining repair suggestions based on the cause of the anomaly and pushing the cause of the anomaly, the anomaly location information, and the repair suggestions to the visualization interface, includes:

[0085] Step S401: Pre-build a repair suggestion database, which is used to store repair suggestions corresponding to the causes of the anomalies;

[0086] Step S402: Match the cause of the anomaly with the repair suggestions in the repair suggestion database. If a repair suggestion is matched, proceed to step S403. If no repair suggestion is matched, proceed to step S404.

[0087] Step S403: Push the cause of the anomaly, the location information of the anomaly, and the repair suggestions to the visualization interface;

[0088] Step S404: Push the cause of the anomaly and the location of the anomaly to the visualization interface.

[0089] Common anomaly causes and corresponding repair suggestions are pre-stored to create a repair suggestion database. This database can be built using either MySQL or Neo4j. MySQL, as a relational database, excels at handling structured data and supports ACID transactions, making it suitable for scenarios involving orders and user information. Neo4j, as a graph database, stores data through nodes and relationships, supporting efficient path queries and complex network analysis, making it suitable for scenarios involving social networks and recommendation systems. By combining the transaction processing capabilities of MySQL with the graph analysis capabilities of Neo4j, efficient data storage and complex queries can be achieved.

[0090] The system matches the anomaly cause with repair suggestions from the repair suggestion database. If a repair suggestion is found, the anomaly cause, anomaly location information, and repair suggestion are pushed to the visualization interface. If no repair suggestion is found, the anomaly cause and anomaly location information are pushed to the visualization interface.

[0091] Step S403, which involves pushing the cause of the anomaly, the anomaly location information, and the repair suggestion to the visualization interface if a repair suggestion is found, includes:

[0092] Step S431: If a repair suggestion is matched, obtain the number of repair suggestions and determine whether the number of repair suggestions is less than or equal to a preset threshold. If the number of repair suggestions is less than or equal to the preset threshold, proceed to step S432. If the number of repair suggestions is greater than the preset threshold, proceed to step S433.

[0093] Step S432: Push the cause of the anomaly, the location information of the anomaly, and the repair suggestions to the visualization interface;

[0094] Step S433: Obtain the correlation between the repair suggestions and the cause of the anomaly, sort the repair suggestions according to the correlation, obtain several repair suggestions ranked before a preset threshold, and push the cause of the anomaly, the anomaly location information, and several repair suggestions ranked before the preset threshold to the visualization interface.

[0095] In this embodiment of the invention, if a repair suggestion is matched, the number of repair suggestions is obtained, and it is determined whether the number of repair suggestions is less than or equal to a preset threshold, where the preset threshold is N, and N is a positive integer greater than 1. If the number of repair suggestions is less than or equal to N, the cause of the anomaly, the anomaly location information, and all repair suggestions are displayed on the visualization interface. If the number of repair suggestions is greater than N, a "Top-N Possible Causes" prompt and suggestion are output.

[0096] Compared with the prior art, the embodiments of the present invention have the following technical advantages:

[0097] Improve data accuracy. Automated anomaly detection: The system uses AI algorithms to scan investment data in real time, quickly identifying outliers (such as data delays, outliers, missing values, etc.), reducing human error and avoiding investment losses due to data errors. Root cause analysis and repair: Combining large-scale models and graph neural networks, the system can automatically locate the causes of anomalies (such as data source problems, calculation logic errors, etc.) and provide repair suggestions.

[0098] Improve management efficiency. Automated processes: The system runs automatically daily without manual intervention, significantly reducing the time spent on manual verification and repair. Real-time monitoring and alerts: Through real-time data monitoring and anomaly alerts, managers can respond quickly to problems, avoiding investment decision errors caused by data delays or mistakes. Long-term data governance: Through automated anomaly detection and repair, the system continuously improves data quality, providing reliable support for long-term investment decisions.

[0099] Enhanced interpretability. Anomaly root cause analysis: The system provides explanations of possible causes, helping users quickly understand the source and impact of anomalies. Visualized interface: Through dashboards and data dependency graphs for data lineage tracing, users can intuitively see anomaly distribution and remediation paths, improving the user experience.

[0100] Optimize resource allocation. Data-driven decision-making: Through in-depth analysis of historical investment data, the system helps optimize asset allocation and risk control strategies, improves resource allocation efficiency, and contributes to higher investment returns.

[0101] It should be noted that there is no necessary order between the above steps. Those skilled in the art will understand from the description of the embodiments of the present invention that the above steps may have different execution orders in different embodiments, that is, they may be executed in parallel or in turn, etc.

[0102] Another embodiment of the present invention provides an artificial intelligence-based anomaly data identification device, which corresponds one-to-one with the artificial intelligence-based anomaly data identification method described in the above embodiments. For example... Figure 3 As shown, device 1 includes:

[0103] Data acquisition module 100 is used to collect investment data and input the investment data into the investment data semantic big model;

[0104] Anomaly identification module 200 is used to identify anomalies in the investment data based on the semantic big data model of the investment data and generate anomaly identification results;

[0105] The abnormal information location module 300 is used to acquire abnormal investment data from the abnormal data identification results, analyze the abnormal situation of the abnormal investment data based on the graph neural network algorithm, and generate abnormal cause and abnormal location information.

[0106] The data push module 400 is used to obtain repair suggestions based on the cause of the anomaly, and push the cause of the anomaly, the anomaly location information and the repair suggestions to the visualization interface.

[0107] For specific implementation details, please refer to the method embodiment; they will not be repeated here.

[0108] In one embodiment, the data acquisition module 100 is specifically used for:

[0109] Real-time collection of investment data from several investment data systems;

[0110] Obtain the anomaly detection threshold corresponding to the financial business type;

[0111] An initial large semantic model was constructed based on long short-term memory networks, attention mechanisms, and the isolated forest algorithm.

[0112] A time-series dataset of investment data is obtained, and the initial semantic big model is trained based on the time-series dataset and the anomaly detection threshold. When the training is completed, an investment data semantic big model is generated.

[0113] For specific implementation details, please refer to the method embodiment; they will not be repeated here.

[0114] In one embodiment, the abnormal data identification module 200 is specifically used for:

[0115] Based on the aforementioned semantic big data model, the logical relationships between different asset types and field structures in the investment data are obtained;

[0116] Obtain the corresponding dynamic detection threshold based on the asset type of the investment data;

[0117] Based on the aforementioned logical relationship, an upstream and downstream logical chain is generated. The investment data is then identified based on the upstream and downstream logical chain and the dynamic detection threshold, generating anomaly data identification results.

[0118] For specific implementation details, please refer to the method embodiment; they will not be repeated here.

[0119] In one embodiment, the anomaly information location module 300 is specifically used for:

[0120] Obtain the abnormal data identification results and determine whether there is abnormal investment data in the abnormal data identification results;

[0121] If abnormal investment data exists, the abnormal situation of the abnormal investment data is analyzed based on the graph neural network algorithm to generate the cause of the abnormality, and the distribution of the abnormal investment data is located to generate abnormal location information.

[0122] If no abnormal investment data is found, the abnormal data identification results will be pushed to the visualization interface.

[0123] For specific implementation details, please refer to the method embodiment; they will not be repeated here.

[0124] In one embodiment, the data push module 400 is specifically used for:

[0125] A repair suggestion database is pre-built, which is used to store repair suggestions corresponding to the causes of the anomalies;

[0126] The cause of the anomaly is matched with the repair suggestions in the repair suggestion database;

[0127] If a repair suggestion is found, the cause of the anomaly, the location of the anomaly, and the repair suggestion will be pushed to the visualization interface.

[0128] If no repair suggestion is found, the cause of the anomaly and the location of the anomaly will be pushed to the visualization interface.

[0129] For specific implementation details, please refer to the method embodiment; they will not be repeated here.

[0130] In one embodiment, the anomaly information location module 300 is further configured to:

[0131] A data dependency graph is established based on the aforementioned investment data;

[0132] The abnormal investment data was analyzed based on graph neural network algorithm and link analysis algorithm, and the abnormal data link corresponding to the abnormal data was located.

[0133] The causes of the abnormal data link are analyzed, and the causes of the abnormality are generated;

[0134] The distribution of the abnormal data links is located, and abnormal location information is generated.

[0135] For specific implementation details, please refer to the method embodiment; they will not be repeated here.

[0136] In one embodiment, the data push module 400 is further configured to:

[0137] If a repair suggestion is matched, the number of repair suggestions is obtained, and it is determined whether the number of repair suggestions is less than or equal to a preset threshold.

[0138] If the number of repair suggestions is less than or equal to a preset threshold, the cause of the anomaly, the location of the anomaly, and the repair suggestions will be pushed to the visualization interface.

[0139] If the number of repair suggestions exceeds a preset threshold, the relevance between the repair suggestions and the cause of the anomaly is obtained. The repair suggestions are then sorted according to their relevance. Several repair suggestions ranked before the preset threshold are obtained, and the cause of the anomaly, the anomaly location information, and several repair suggestions ranked before the preset threshold are pushed to the visualization interface.

[0140] For specific implementation details, please refer to the method embodiment; they will not be repeated here.

[0141] This invention provides an artificial intelligence-based anomaly data identification device that significantly improves the intelligence level of investment data management by combining a large-scale AI model with a personalized knowledge base. The benefits extend beyond data accuracy, management efficiency, and risk control; they also extend to user experience and long-term sustainable development. This system provides strong support for investment institutions in a complex and ever-changing market environment, helping them gain a competitive edge.

[0142] Another embodiment of the present invention provides a computer device, which may be a server, and its internal structure diagram may be as follows. Figure 4As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a server-side method for identifying abnormal data based on artificial intelligence.

[0143] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements client-side functions or steps of an artificial intelligence-based anomaly data identification method.

[0144] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0145] Real-time acquisition of investment data, and input of the investment data into the investment data semantic big data model;

[0146] Based on the semantic big data model of the investment data, anomaly identification is performed on the investment data to generate anomaly identification results;

[0147] Obtain abnormal investment data from the abnormal data identification results, analyze the abnormal situation of the abnormal investment data based on the graph neural network algorithm, and generate abnormal cause and abnormal location information;

[0148] Based on the cause of the anomaly, repair suggestions are obtained, and the cause of the anomaly, anomaly location information, and repair suggestions are pushed to the visualization interface.

[0149] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0150] Real-time acquisition of investment data, and input of the investment data into the investment data semantic big data model;

[0151] Based on the semantic big data model of the investment data, anomaly identification is performed on the investment data to generate anomaly identification results;

[0152] Obtain abnormal investment data from the abnormal data identification results, analyze the abnormal situation of the abnormal investment data based on the graph neural network algorithm, and generate abnormal cause and abnormal location information;

[0153] Based on the cause of the anomaly, repair suggestions are obtained, and the cause of the anomaly, anomaly location information, and repair suggestions are pushed to the visualization interface.

[0154] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0155] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0156] The embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0157] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a general-purpose hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can exist in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0158] It should be noted that if any software tools or components not belonging to our company appear in the embodiments of this application, they are merely for illustrative purposes and do not represent actual use.

[0159] Among other things, conditional language such as “can,” “may,” “may,” or “may,” unless otherwise specifically stated or otherwise understood as in the context in which they are used, is generally intended to convey that a particular implementation may include (but not others) certain features, elements, and / or operations. Therefore, such conditional language is also generally intended to imply that features, elements, and / or operations are necessary for one or more implementations in any way, or that one or more implementations must include logic for determining, with or without input or prompting, whether such features, elements, and / or operations are included or will be performed in any particular implementation.

[0160] The contents already described herein in this specification and accompanying drawings include examples of methods and apparatuses capable of providing artificial intelligence-based anomaly data identification. Of course, it is not possible to describe every conceivable combination of elements and / or methods for the purpose of describing the various features of this disclosure, but it will be appreciated that many other combinations and substitutions of the disclosed features are possible. Therefore, it will be apparent that various modifications can be made to this disclosure without departing from the scope or spirit of this disclosure. Furthermore, or in alternatives, other embodiments of this disclosure may become apparent from consideration of this specification and accompanying drawings and from practice of this disclosure as presented herein. It is intended that the examples presented in this specification and accompanying drawings be considered illustrative rather than restrictive in all respects. Although specific terminology is used herein, it is used in a general and descriptive sense and is not intended for limiting purposes.

Claims

1. An artificial intelligence-based method for identifying abnormal data, characterized in that... The method includes: Real-time acquisition of investment data, and input of the investment data into the investment data semantic big data model; Based on the semantic big data model of the investment data, anomaly identification is performed on the investment data to generate anomaly identification results; Obtain abnormal investment data from the abnormal data identification results, analyze the abnormal situation of the abnormal investment data based on the graph neural network algorithm, and generate abnormal cause and abnormal location information; Based on the cause of the anomaly, repair suggestions are obtained, and the cause of the anomaly, anomaly location information, and repair suggestions are pushed to the visualization interface.

2. The artificial intelligence-based abnormal data identification method according to claim 1, characterized in that, The real-time acquisition of investment data, and inputting the investment data into the investment data semantic big data model, includes: Real-time collection of investment data from several investment data systems; Obtain the anomaly detection threshold corresponding to the financial business type; An initial large semantic model was constructed based on long short-term memory networks, attention mechanisms, and the isolated forest algorithm. A time-series dataset of investment data is obtained, and the initial semantic big model is trained based on the time-series dataset and the anomaly detection threshold. When the training is completed, an investment data semantic big model is generated.

3. The artificial intelligence-based abnormal data identification method according to claim 1, characterized in that, The step of identifying anomalies in the investment data based on the semantic big data model and generating anomaly identification results includes: Based on the aforementioned semantic big data model, the logical relationships between different asset types and field structures in the investment data are obtained; Obtain the corresponding dynamic detection threshold based on the asset type of the investment data; Based on the aforementioned logical relationship, an upstream and downstream logical chain is generated. The investment data is then identified based on the upstream and downstream logical chain and the dynamic detection threshold, generating anomaly data identification results.

4. The artificial intelligence-based abnormal data identification method according to claim 1, characterized in that, The abnormal investment data in the abnormal data identification results are obtained, and the abnormal situation of the abnormal investment data is analyzed based on the graph neural network algorithm to generate abnormal cause and abnormal location information, including: Obtain the abnormal data identification results and determine whether there is abnormal investment data in the abnormal data identification results; If abnormal investment data exists, the abnormal situation of the abnormal investment data is analyzed based on the graph neural network algorithm to generate the cause of the abnormality, and the distribution of the abnormal investment data is located to generate abnormal location information. If no abnormal investment data is found, the abnormal data identification results will be pushed to the visualization interface.

5. The artificial intelligence-based abnormal data identification method according to claim 1, characterized in that, The step of obtaining repair suggestions based on the cause of the anomaly and pushing the cause of the anomaly, anomaly location information, and repair suggestions to the visualization interface includes: A repair suggestion database is pre-built, which is used to store repair suggestions corresponding to the causes of the anomalies; The cause of the anomaly is matched with the repair suggestions in the repair suggestion database; If a repair suggestion is found, the cause of the anomaly, the location of the anomaly, and the repair suggestion will be pushed to the visualization interface. If no repair suggestion is found, the cause of the anomaly and the location of the anomaly will be pushed to the visualization interface.

6. The artificial intelligence-based abnormal data identification method according to claim 4, characterized in that, The graph neural network algorithm is used to analyze the anomalies in the abnormal investment data, generate the causes of the anomalies, and locate the distribution of the abnormal investment data to generate anomaly location information, including: A data dependency graph is established based on the aforementioned investment data; The abnormal investment data is analyzed based on graph neural network algorithm and link analysis algorithm to locate the abnormal data link corresponding to the abnormal data. The causes of the abnormal data link are analyzed, and the causes of the abnormality are generated; The distribution of the abnormal data links is located, and abnormal location information is generated.

7. The artificial intelligence-based abnormal data identification method according to claim 5, characterized in that, If a repair suggestion is found, the cause of the anomaly, the anomaly location information, and the repair suggestion are pushed to the visualization interface, including: If a repair suggestion is matched, the number of repair suggestions is obtained, and it is determined whether the number of repair suggestions is less than or equal to a preset threshold. If the number of repair suggestions is less than or equal to a preset threshold, the cause of the anomaly, the location of the anomaly, and the repair suggestions will be pushed to the visualization interface. If the number of repair suggestions exceeds a preset threshold, the relevance between the repair suggestions and the cause of the anomaly is obtained. The repair suggestions are then sorted according to their relevance. Several repair suggestions ranked before the preset threshold are obtained, and the cause of the anomaly, the anomaly location information, and several repair suggestions ranked before the preset threshold are pushed to the visualization interface.

8. An artificial intelligence-based anomaly data identification device, characterized in that, The device includes: The data acquisition module is used to collect investment data and input the investment data into the investment data semantic big data model; An anomaly identification module is used to identify anomalies in the investment data based on the semantic big data model of the investment data and generate anomaly identification results. An anomaly information location module is used to acquire abnormal investment data from the abnormal data identification results, analyze the anomalies in the abnormal investment data based on a graph neural network algorithm, and generate anomaly cause and anomaly location information. The data push module is used to obtain repair suggestions based on the cause of the anomaly, and push the cause of the anomaly, the anomaly location information, and the repair suggestions to the visualization interface.

9. A computer device, characterized in that, The computer device includes at least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the steps of the artificial intelligence-based anomaly data identification method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the artificial intelligence-based abnormal data identification method according to any one of claims 1-7.

Images