Intelligent enterprise financing graph system and device based on distributed architecture

Abstract

The present invention provides an intelligent enterprise financing map system and device based on a distributed architecture, including: a distributed data acquisition module: collecting enterprise financing related data from a multi-source data interface; a real-time data processing module: based on a distributed message system and a stream processing framework, performing real-time processing and analysis on the collected data; a distributed storage module: based on a distributed system architecture and a distributed non-relational database, storing and managing large-scale structured and unstructured data; a machine learning prediction module: integrating an XGBoost regression model and a long short-term memory network LSTM algorithm to predict financing success rate and valuation; a 3D visualization module: based on 3D drawing technology, intuitively displaying an enterprise financing relationship map; a blockchain data recording module: based on distributed ledger technology, ensuring the non-tamperability of key data; a security management module: realizing data encryption and role-based access control.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to an intelligent enterprise financing graph system and device based on a distributed architecture. Background Art

[0002] With the rapid development of the entrepreneurial ecosystem, corporate financing activities are becoming increasingly frequent and complex. However, existing corporate financing analysis systems have the following problems:

[0003] 1) The data processing capability is limited and it is difficult to cope with massive real-time data;

[0004] 2) Lack of intelligent analysis and inability to provide deep insights;

[0005] 3) The visualization effect is poor and it is difficult to intuitively display complex relationships;

[0006] 4) Insufficient data security, and the risk of information leakage;

[0007] 5) The system has poor scalability and is difficult to adapt to rapidly changing market demands.

[0008] Therefore, there is an urgent need for an intelligent enterprise financing graph system that can solve the above problems.

[0009] Patent document CN109598619A discloses a community investment and financing service system based on blockchain technology, including a blockchain information system, a data processing center and a demand matching engine, wherein the blockchain information system includes: a third-party credit institution, a community investor basic information collection module, a community financier basic information collection module, an investment demand blockchain collection module; a financing demand blockchain collection module; the data processing center verifies and stores the blockchain generated by the blockchain information system, and then integrates the verification results into an investment list block and a financing list block, and includes them in the blockchain; the demand matching engine matches the investment needs of community investors with the financing needs of community financiers according to the blockchain provided by the data processing center, and reaches an investment and financing service transaction. However, this patent cannot completely solve the above-mentioned technical problems, nor can it meet the needs of the present invention. Summary of the invention

[0010] In view of the defects in the prior art, the purpose of the present invention is to provide an intelligent enterprise financing map system and device based on a distributed architecture.

[0011] The intelligent enterprise financing graph system based on distributed architecture provided by the present invention includes:

[0012] Distributed data collection module: collects enterprise financing related data from multi-source data interfaces;

[0013] Real-time data processing module: Based on the distributed messaging system and stream processing framework, it processes and analyzes the collected data in real time;

[0014] Distributed storage module: based on distributed system architecture and distributed non-relational database, it stores and manages large-scale structured and unstructured data;

[0015] Machine learning prediction module: Integrates XGBoost regression model and long short-term memory network LSTM algorithm to predict financing success rate and valuation;

[0016] 3D visualization module: Based on 3D drawing technology, it can intuitively display the corporate financing relationship map;

[0017] Blockchain data recording module: Based on distributed ledger technology, it ensures that key data cannot be tampered with;

[0018] Security management module: implements data encryption and role-based access control.

[0019] Preferably, the distributed data acquisition module includes: collecting multi-source data through API interface, web crawler and manual input, building a data processing pipeline using the big data processing and distribution system Apache NiFi, and performing data cleaning, conversion and standardization.

[0020] Preferably, the real-time data processing module includes: creating a topic of a distributed stream processing platform, storing information on financing rounds, investors and financing amounts, and performing real-time aggregation, anomaly detection and trend analysis;

[0021] Calculation of total financing amount:

[0022] Total_Funding = ∑(i=1 to n) Fi

[0023] Among them, Fi is the i-th financing amount, and n is the number of financings;

[0024] Anomaly detection using moving average and standard deviation:

[0025] MA(t) = (∑(i=t-k+1 to t) Fi) / k

[0026] σ=√((∑(i=1 to n) (Fi -μ)²) / n)

[0027] Where t represents the current time point or the currently observed data point; k represents the time window size or period of the moving average; μ represents the overall mean or expected value of the data set, μ=(∑(i=1 to n) Fi) / n;

[0028] If |Fi - MA(t)|>3σ, it is judged as an abnormal financing event.

[0029] Preferably, the machine learning prediction module includes:

[0030] Construct a financing success rate prediction model, the objective function is:

[0031] obj(θ) = ∑(i=1 to n) l(yi, i) + ∑(k=1 to K) Ω(fk)

[0032] Where l is the loss function, Ω is the regularization term, fk is the decision tree; θ represents the parameter set of the model; yi represents the true value of the i-th financing amount; i represents the model's predicted value for the i-th financing amount; K represents the total number of decision trees in the model.

[0033] Preferably, the machine learning prediction module includes:

[0034] Construct a financing valuation prediction model, the expression is:

[0035] ft = σ(Wf · [ht-1, xt] + bf)

[0036] it = σ(Wi · [ht-1, xt] + bi)

[0037] t = tanh(WC · [ht-1, xt] + bC)

[0038] Ct = ft * Ct-1 + it * t

[0039] ot = σ(Wo · [ht-1, xt] + bo)

[0040] ht = ot * tanh(Ct)

[0041] Among them, ft represents the output vector of the forget gate; Wf represents the weight matrix of the forget gate; ht-1 represents the hidden state vector at the previous moment; xt represents the input vector at the current moment; bf represents the bias vector of the forget gate; it represents the output vector of the input gate; Wi represents the weight matrix of the input gate; bi represents the bias vector of the input gate; t represents the candidate cell state vector; WC represents the weight matrix of the candidate cell state; bC represents the bias vector of the candidate cell state; Ct represents the cell state vector at the current moment; ot represents the output vector of the output gate; Wo represents the weight matrix of the output gate; bo represents the bias vector of the output gate; ht represents the hidden state vector at the current moment; σ represents the sigmoid activation function, and tanh represents the hyperbolic tangent activation function.

[0042] Preferably, the machine learning prediction module includes:

[0043] The collaborative filtering algorithm based on matrix decomposition is used to realize investor matching recommendation, and the expression is:

[0044] R ≈ U * V^T

[0045] Among them, R is the user-item rating matrix, U is the user latent factor matrix, V is the item latent factor matrix, and T represents the matrix transpose operation.

[0046] Preferably, the 3D visualization module includes: using WebGL technology to implement a 3D force-directed graph, and the node position update formula is:

[0047] x = x + (Fx / |F|) * min(|F|,temperature)

[0048] y = y + (Fy / |F|) * min(|F|, temperature)

[0049] Among them, temperature is the cooling parameter that controls the node movement speed; x and y represent the current coordinates of the node on the two-dimensional plane; Fx and Fy represent the components of the resultant force acting on the node in the x and y directions; F represents the magnitude of the resultant force acting on the node, |F|=√(Fx² + Fy²).

[0050] Preferably, the blockchain data recording module includes: using the distributed ledger Hyperledger Fabric to build a permissioned blockchain network, adopting the PBFT consensus algorithm, and the view number v calculation formula is: v = (v + 1) mod 2^64.

[0051] Preferably, the security management module includes: using AES-256 algorithm to encrypt data, including: round key addition, byte replacement, row shift and column mixing; implementing role-based access control and supporting fine-grained permission management.

[0052] The intelligent enterprise financing map device based on distributed architecture provided by the present invention is equipped with the intelligent enterprise financing map system based on distributed architecture.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] (1) Improve financing efficiency: Through intelligent matching algorithms, help companies quickly find suitable investors and shorten the financing cycle;

[0055] (2) Reduce investment risks: Use machine learning models to conduct comprehensive assessments of companies and provide investors with scientific decision-making basis;

[0056] (3) Optimize the supervision process: monitor the dynamics of the financing market in real time, detect abnormal situations in a timely manner, and improve supervision efficiency;

[0057] (4) Promote resource integration: Build a complete industrial chain map to promote cooperation between upstream and downstream enterprises and optimize resource allocation;

[0058] (5) Support policy formulation: Provide industry development trend analysis to government departments and assist in the formulation of industrial policies;

[0059] (6) Enhance data value: Integrate multi-source data, explore data value, and provide data support for corporate strategic planning;

[0060] (7) Enhance market transparency: Increase market information transparency and promote healthy competition through visual display;

[0061] (8) Promoting technological innovation: Encouraging enterprises to increase investment in research and development, improve core competitiveness, and attract high-quality investment. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0063] Figure 1 It is a data processing flow chart;

[0064] Figure 2 This is a diagram of the machine learning model architecture;

[0065] Figure 3 This is a schematic diagram of the 3D financing map;

[0066] Figure 4 Flowchart for blockchain data recording. DETAILED DESCRIPTION

[0067] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0068] Embodiment 1:

[0069] The intelligent enterprise financing graph system of the present invention is designed based on a distributed architecture and mainly includes the following modules:

[0070] 1. Distributed data acquisition module:

[0071] Public data API interface: connect to third-party data sources such as crunchbase and pitchbook;

[0072] Web crawler: Use the Scrapy framework to achieve scheduled crawling of the target website;

[0073] Manual input: Provides web form and Excel import functions;

[0074] 2. Real-time data processing module:

[0075] Building a message queue based on Apache Kafka;

[0076] Real-time stream processing using Apache Flink;

[0077] Enables real-time aggregation, anomaly detection, and trend analysis;

[0078] 3. Distributed storage module:

[0079] Use Hadoop Distributed File System (HDFS) to store raw data;

[0080] Use HBase to store structured and semi-structured data;

[0081] Use Elasticsearch to implement full-text retrieval;

[0082] 4. Machine learning prediction module:

[0083] Financing success rate prediction: based on XGBoost algorithm;

[0084] Financing valuation prediction: using LSTM neural network;

[0085] Investor matching recommendation: based on collaborative filtering algorithm;

[0086] 5. 3D Visualization Module:

[0087] Use WebGL technology to achieve 3D rendering;

[0088] Use the Three.js library to encapsulate the underlying graphics interface;

[0089] Realize interactive corporate financing relationship map;

[0090] 6. Blockchain data recording module:

[0091] Build a permissioned blockchain network based on Hyperledger Fabric;

[0092] Achieve tamper-proof storage of key data;

[0093] Support smart contracts to automatically execute business rules;

[0094] 7. Security management module:

[0095] Data encryption: using AES-256 algorithm;

[0096] Access control: Implement role-based access control (RBAC);

[0097] Security Audit: Record system operation logs and support security incident tracing.

[0098] Embodiment 2:

[0099] like Figure 1 As shown in the figure, the data collection and preprocessing process mainly includes the following steps:

[0100] 1. Multi-source data collection:

[0101] a) Regularly obtain relevant data on corporate financing, investment institutions, etc. through public data API;

[0102] b) Use the Scrapy framework to crawl the latest financing news and company information from the specified website;

[0103] c) Provides a manual input interface, supporting Excel batch import and Web form single entry;

[0104] 2. Data cleaning:

[0105] a) De-duplicate data: Use hashing algorithms to quickly identify and delete duplicate records;

[0106] b) Correct format errors: standardize the formats of date, amount and other fields;

[0107] c) Fill in missing values: use mean, mode or advanced interpolation algorithm according to data type;

[0108] 3. Data conversion:

[0109] a) Standardized field names: unify field names from different sources;

[0110] b) Unify date format: convert all dates to "YYYY-MM-DD" format;

[0111] c) Currency conversion: uniformly converted to US dollars, using the real-time exchange rate API;

[0112] 4. Data Standardization:

[0113] Normalize the numerical data, and the calculation formula is:

[0114] X_norm = (X - X_min) / (X_max - X_min)

[0115] Among them, X_min and X_max are the minimum and maximum values ​​of the field respectively;

[0116] 5. Data Storage:

[0117] a) The original data is stored in HDFS to retain data traceability;

[0118] b) Structured data is stored in HBase, and the storage mode is designed according to "row key + column family";

[0119] c) Create an Elasticsearch index to support full-text retrieval and complex queries.

[0120] Embodiment 3:

[0121] like Figure 1 As shown in the figure, the real-time data processing process is implemented based on Kafka and Flink, and mainly includes the following steps:

[0122] 1. Kafka Topics Design:

[0123] Financing rounds: record the details of each round of financing;

[0124] Investors: store dynamic information of investment institutions and individual investors;

[0125] Financing amount: Real-time update of changes in financing amount of each enterprise;

[0126] 2. Flink Job Implementation:

[0127] a) Real-time aggregation:

[0128] Calculation of total financing amount: Total_Funding = ∑(i=1 to n) Fi;

[0129] Average financing amount: Avg_Funding = (∑(i=1 to n) Fi) / n;

[0130] Among them, Fi is the financing amount of the i-th round, and n is the number of financing rounds;

[0131] b) Anomaly Detection:

[0132] Moving average: MA(t) = (∑(i=t-k+1 to t) Fi) / k;

[0133] Standard deviation: σ = √((∑(i=1 to n) (Fi - μ)²) / n);

[0134] Abnormal determination rule: If |Fi - MA(t)|>3σ, it is determined to be an abnormal financing event;

[0135] Among them, t represents the current time point or the currently observed data point. In the moving average formula, t represents the end time point for calculating the average value; k represents the time window size or period of the moving average. For example, if a 10-day moving average is calculated, k=10. In the formula, from t-k+1 to t means taking the average of the most recent k data points; μ(mu) represents the overall mean or expected value of the data set. In the standard deviation formula, μ represents the arithmetic mean of all data points, which is calculated by dividing the sum of all data points by the total number of data points: μ= (∑(i=1 to n) Fi) / n.

[0136] In the moving average formula, t and k together define the time range for calculating the average value. In the standard deviation formula, μ is used to calculate the deviation of each data point from the average value. In the anomaly determination rule, the moving average value MA(t) at time t is compared with the current financing amount Fi, and the standard deviation σ is used as the basis for the determination threshold.

[0137] This method identifies possible abnormal financing events by comparing the difference between the current financing amount and the recent average level and taking into account the overall volatility (expressed as standard deviation). If the difference between a financing amount and the recent average exceeds 3 standard deviations, it is considered an outlier.

[0138] c) Trend analysis:

[0139] Financing heat index: Heat_Index(t) = (∑(i=t-m+1 to t) wi * Fi) / m;

[0140] Among them, wi is the time weight, m ​​is the sliding window size;

[0141] 3. Real-time data update:

[0142] The processing results are pushed to the front end in real time through Kafka to update the visualization graphs and analysis reports.

[0143] Embodiment 4:

[0144] like Figure 2 As shown, this system contains three core machine learning models:

[0145] 1. Financing success rate prediction model (XGBoost):

[0146] Feature engineering: including historical financing, team background, market size, etc.

[0147] Objective function: obj(θ) = ∑(i=1 to n) l(yi, i) + ∑(k=1 to K) Ω(fk);

[0148] The weight calculation of the leaf node of the tree is: wj = - (∑(i∈Ij) gi) / (∑(i∈Ij) hi + λ);

[0149] Among them, θ is the parameter set of the model; yi is the true value of the i-th financing amount, i is the model's predicted value for the i-th financing amount; K is the total number of trees in the model; Ω(fk) is the complexity penalty term of the k-th tree, which is used to control the model complexity and avoid overfitting; fk is the prediction function of the k-th tree; Ij is the set of samples falling on the j-th leaf node; gi is the first-order gradient of the i-th financing amount, that is, the first-order derivative of the loss function with respect to the predicted value; hi is the second-order gradient of the i-th financing amount, that is, the second-order derivative of the loss function with respect to the predicted value; λ is the regularization parameter, which is used to control the strength of regularization.

[0150] In the objective function, the first term is the loss function, which measures the difference between the model prediction value and the true value; the second term is the regularization term, which is used to control the complexity of the model.

[0151] The weight of a leaf node is the sum of the negative gradients of all samples of the node divided by the sum of the second-order gradients plus the regularization term. This minimizes the objective function and balances the model's performance on the training data with its generalization ability.

[0152] This model predicts the financing success rate through feature engineering (historical financing status, team background, market size, etc.) and optimizing the objective function, while using regularization to prevent overfitting.

[0153] 2. Financing valuation prediction model (LSTM):

[0154] Input data: time series data, including historical financing, financial indicators, market performance, etc.

[0155] LSTM unit core formula:

[0156] ft = σ(Wf · [ht-1, xt] + bf)

[0157] it = σ(Wi · [ht-1, xt] + bi)

[0158] t = tanh(WC · [ht-1, xt] + bC)

[0159] Ct = ft * Ct-1 + it * t

[0160] ot = σ(Wo · [ht-1, xt] + bo)

[0161] ht = ot * tanh(Ct)

[0162] Among them, ft is the output of the forget gate, which determines which information to discard from the cell state; Wf is the weight matrix of the forget gate; ht-1 is the hidden state at the previous moment; xt is the input data at the current moment, which includes time series data such as historical financing, financial indicators, and market performance in this financing valuation prediction model; bf is the bias term of the forget gate; it is the output of the input gate, which controls how much new information is added to the cell state; Wi is the weight matrix of the input gate; bi is the bias term of the input gate; t is the candidate cell state, which contains new information that may be added to the cell state; tanh is the hyperbolic tangent activation function, which compresses the value into the range of [-1, 1]; WC is the weight matrix of the candidate cell state; bC is the bias term of the candidate cell state; Ct is the cell state at the current moment, the main carrier of long-term memory; ot is the output of the output gate, which controls how much information is passed from the cell state to the hidden state; Wo is the weight matrix of the output gate; bo is the bias term of the output gate; ht is the hidden state at the current moment, which is also the output of the LSTM unit;

[0163] In this financing valuation prediction model, LSTM can effectively capture long-term dependencies through these gating mechanisms and cell states, thereby better utilizing time series data such as historical financing, financial indicators, and market performance to predict future financing valuations.

[0164] 3. Investor matching recommendation system:

[0165] Matrix decomposition: R ≈ U * V^T

[0166] Cosine similarity: similarity = cos(θ) = (A · B) / (||A|| * ||B||) = ∑(i=1to n) (Ai * Bi) / (√(∑(i=1 to n) Ai²) * √(∑(i=1 to n) Bi²))

[0167] Among them, R is the user-project rating matrix, which can be understood as the investor-enterprise interest or matching degree matrix in the investor matching scenario; U is the user latent factor matrix, each row represents the latent feature vector of an investor; V is the project latent factor matrix, each column represents the latent feature vector of an enterprise; T is the matrix transpose, V^T is the transposed matrix of V; the two vectors A and B can represent the feature vectors of two investors or the feature vectors of two enterprises; Ai and Bi are the i-th components of vectors A and B respectively; n is the dimension of the vector, that is, the number of features; ||A|| and ||B|| represent the Euclidean norm (length) of vectors A and B respectively.

[0168] In this investor matching recommendation system, matrix decomposition is used to learn the potential characteristics of investors and enterprises from historical data, while cosine similarity is used to calculate the similarity between investor-enterprise pairs for matching and recommendation. This method can effectively process large-scale sparse data and discover potential matching relationships.

[0169] Embodiment 5:

[0170] like Figure 3 As shown, the realization of the 3D financing relationship map mainly includes the following steps:

[0171] 1. Data preparation:

[0172] Read the data of enterprise nodes and financing relationship edges from HBase to build a graph data structure;

[0173] 2. Layout algorithm:

[0174] The Fruchterman-Reingold force-directed algorithm is used to achieve node layout;

[0175] Calculation of repulsive force: Fr = k² / d;

[0176] Gravitational calculation: Fa = d² / k;

[0177] Where k is the ideal edge length and d is the distance between nodes;

[0178] 3. Node location update:

[0179] x = x + (Fx / |F|) * min(|F|, temperature);

[0180] y = y + (Fy / |F|) * min(|F|, temperature);

[0181] z = z + (Fz / |F|) * min(|F|, temperature);

[0182] Among them, temperature is a cooling parameter that controls the node movement speed;

[0183] x, y, z represent the current coordinates of the node in three-dimensional space. After each iteration, these values ​​are updated to reflect the new position of the node. Fx, Fy, Fz represent the components of the resultant force acting on the node in the x, y, and z directions. This force may come from multiple sources, such as repulsion between nodes, gravitational force of edges, etc. |F| represents the magnitude (modulus) of the resultant force acting on the node, which can be calculated by the formula |F| = √(Fx² + Fy² + Fz²). temperature is a cooling parameter used to control the movement speed of the node. As the simulation progresses, the temperature usually decreases gradually, which makes the movement amplitude of the node smaller and helps the system reach a stable state.

[0184] 4. WebGL rendering:

[0185] Use the Three.js library to encapsulate the WebGL interface to achieve 3D scene rendering;

[0186] Create scenes, cameras, and renderers;

[0187] Add light sources: ambient light and parallel light;

[0188] Create a node using THREE.SphereGeometry;

[0189] Use THREE.Line to create edges;

[0190] Realize viewing angle control and zoom function;

[0191] 5. Interactive functions:

[0192] Hovering over a node displays detailed information;

[0193] Click on a node to expand the relationship;

[0194] Supports graph rotation, zooming and translation.

[0195] Embodiment 6:

[0196] like Figure 4 As shown in the figure, the blockchain data recording process based on Hyperledger Fabric includes:

[0197] 1. Network topology design:

[0198] Set up 4 organizations: regulatory agencies, enterprises, investors, and third-party data providers;

[0199] Each organization deploys 2 peer nodes;

[0200] Set up an Orderer node to be responsible for transaction sorting;

[0201] 2. Channel configuration:

[0202] Create a channel named "Funding Data Channel" and all organizations join the channel;

[0203] 3. Chaincode (smart contract) development:

[0204] Define asset structure: financing events, corporate information, investor information;

[0205] Implement CRUD operations: create, read, update and delete assets;

[0206] Define business logic: financing information verification, fund flow tracking, etc.;

[0207] 4. Consensus Mechanism:

[0208] Adopt PBFT (Practical Byzantine Fault Tolerance) algorithm;

[0209] View number calculation: v = (v + 1) mod 2^64;

[0210] Fault tolerance: It can tolerate f nodes to act maliciously, and the total number of nodes must satisfy n ≥ 3f + 1;

[0211] v represents the "view" number in the PBFT network. In PBFT, a view is a time period during which a specific node is designated as the primary or leader. The view number increases over time and is used to track the state of the network and changes in the primary node.

[0212] This formula means that each time the view changes, the view number is incremented by 1 and modulo 2^64. This ensures that the view number cycles through the range 0 to 2^64-1.

[0213] View change: When the network detects that there may be a problem with the current master node (such as sending invalid messages or failing to generate valid blocks in a timely manner), it will trigger a view change and switch to a new master node.

[0214] 5. Data on-chain process:

[0215] a) The client submits a financing transaction proposal;

[0216] b) The endorsing node simulates execution and signs;

[0217] c) The client collects endorsements and submits them to the Orderer node;

[0218] d) The Orderer node sorts the transactions and packages them into blocks;

[0219] e) The peer nodes of each organization verify and submit blocks;

[0220] 6. Data query:

[0221] Supports query by company name, financing round, time range, etc.

[0222] Realize the full process traceability of financing history.

[0223] Embodiment 7:

[0224] The steps to implement security management are as follows:

[0225] 1. Data encryption:

[0226] The AES-256 algorithm is used to encrypt sensitive data. The main steps include:

[0227] Key expansion: expand the 256-bit key into 14 rounds of subkeys;

[0228] Round key addition: state = state ⊕ roundKey;

[0229] Byte replacement: state[i,j] = S-box[state[i,j]];

[0230] Row shift: state[i] = state[i]<< <shift(i);

[0231] Column mixing: state[j] = M * state[j];

[0232] Where M is a fixed 4x4 matrix;

[0233] 2. Access Control:

[0234] Implement role-based access control (RBAC), mainly including:

[0235] Role definition: administrator, enterprise user, investor, ordinary user, etc.

[0236] Permission allocation: set fine-grained permissions for different data and functional modules;

[0237] Dynamic authorization: supports granting and revoking temporary permissions;

[0238] 3. Security Audit:

[0239] Record logs of user login, data access, operation behavior, etc.

[0240] Implement abnormal behavior detection, such as frequent failed logins, batch access to sensitive data, etc.;

[0241] Supports log visualization analysis and real-time alarms.

[0242] Through the above implementation methods, the present invention realizes an intelligent enterprise financing map system based on a distributed architecture, which has efficient data processing capabilities, accurate financing forecasting functions, intuitive 3D visualization, secure and reliable data storage, and a complete security management mechanism. The system can provide comprehensive, timely, and reliable financing information and decision-making support for enterprises, investors, and regulators. The intelligent enterprise financing map system of the present invention can significantly improve the efficiency and transparency of the financing market, reduce financing costs and risks, promote optimal resource allocation and industrial upgrading, and has broad application prospects and important economic and social benefits.

[0243] Those skilled in the art know that, in addition to implementing the system, device and its various modules provided by the present invention in a purely computer-readable program code, it is entirely possible to implement the same program in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers and embedded microcontrollers by logically programming the method steps. Therefore, the system, device and its various modules provided by the present invention can be considered as a hardware component, and the modules included therein for implementing various programs can also be considered as structures within the hardware component; the modules for implementing various functions can also be considered as both software programs for implementing the method and structures within the hardware component.

[0244] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. An intelligent enterprise financing graph system based on distributed architecture, It is characterized in that include: Distributed data collection module: collects enterprise financing related data from multi-source data interfaces; Real-time data processing module: Based on the distributed messaging system and stream processing framework, it processes and analyzes the collected data in real time; Distributed storage module: based on distributed system architecture and distributed non-relational database, it stores and manages large-scale structured and unstructured data; Machine learning prediction module: Integrates XGBoost regression model and long short-term memory network LSTM algorithm to predict financing success rate and valuation; 3D visualization module: Based on 3D drawing technology, it can intuitively display the corporate financing relationship map; Blockchain data recording module: Based on distributed ledger technology, it ensures that key data cannot be tampered with; Security management module: implements data encryption and role-based access control; Build a financing success rate prediction model, predict the financing success rate through feature engineering and optimization of the objective function, and use regularization to prevent overfitting; Construct the financing valuation prediction model LSTM. The core formula of the LSTM unit is: ft = σ(Wf · [ht-1, xt] + bf) it = σ(Wi · [ht-1, xt] + bi) t = tanh(WC · [ht-1,xt] + bC) Ct = ft * Ct-1 + it * t ot = σ(Wo · [ht-1, xt] + bo) ht = ot * tanh(Ct) Among them, ft represents the output vector of the forget gate; Wf represents the weight matrix of the forget gate; ht-1 represents the hidden state vector at the previous moment; xt represents the input vector at the current moment; bf represents the bias vector of the forget gate; it represents the output vector of the input gate; Wi represents the weight matrix of the input gate; bi represents the bias vector of the input gate; t represents the candidate cell state vector; WC represents the weight matrix of the candidate cell state; bC represents the bias vector of the candidate cell state; Ct represents the cell state vector at the current moment; ot represents the output vector of the output gate; Wo represents the weight matrix of the output gate; bo represents the bias vector of the output gate; ht represents the hidden state vector at the current moment; σ represents the sigmoid activation function, and tanh represents the hyperbolic tangent activation function; In the financing valuation prediction model, LSTM captures long-term dependencies through forget gates, input gates, output gating mechanisms, and cell states, thereby better utilizing historical financing, financial indicators, and market performance to predict future financing valuations. Adopting collaborative filtering algorithm based on matrix decomposition to achieve investor matching recommendation; Matrix decomposition: R ≈ U * V^T Cosine similarity: similarity = cos(θ) = (A · B) / (||A|| * ||B||)= ∑(i=1 to n)(Ai * Bi) / (√(∑(i=1 to n) Ai²) * √(∑(i=1 to n) Bi²)) Where R is the user-item rating matrix; U is the user latent factor matrix, each row represents the latent feature vector of an investor; V is the project latent factor matrix, each column represents the latent feature vector of an enterprise; T is the matrix transpose, V^T is the transpose matrix of V; A and B are two vectors, representing the feature vectors of two investors or the feature vectors of two enterprises; Ai and Bi are the i-th components of vectors A and B respectively; n is the dimension of the vector, that is, the number of features; ||A|| and ||B|| represent the Euclidean norms of vectors A and B respectively; Matrix factorization is used to learn the latent characteristics of investors and firms from historical data, and cosine similarity is used to calculate the similarity between investor-firm pairs for matching and recommendation.

2. According to the distributed architecture-based intelligent enterprise financing graph system of claim 1, It is characterized in that The distributed data acquisition module includes: collecting multi-source data through API interface, web crawler and manual input, building data processing pipeline using big data processing and distribution system Apache NiFi, and performing data cleaning, conversion and standardization.

3. According to the distributed architecture-based intelligent enterprise financing graph system of claim 1, It is characterized in that The real-time data processing module includes: creating topics for a distributed stream processing platform, storing information on financing rounds, investors, and financing amounts, and performing real-time aggregation, anomaly detection, and trend analysis; Calculation of total financing amount: Total_Funding = ∑(i=1 to n) Fi Among them, Fi is the i-th financing amount, and n is the number of financings; Anomaly detection using moving average and standard deviation: MA(t) = (∑(i=t-k+1 to t) Fi) / k σ=√((∑(i=1 to n) (Fi -μ)²) / n) Where t represents the current time point or the currently observed data point; k represents the time window size or period of the moving average; μ represents the overall mean or expected value of the data set, μ=(∑(i=1 to n) Fi) / n; If |Fi - MA(t)| > 3σ, it is considered an abnormal financing event.

4. According to claim 3, the intelligent enterprise financing graph system based on distributed architecture, It is characterized in that The machine learning prediction module includes: Construct a financing success rate prediction model, the objective function is: obj(θ) = ∑(i=1 to n) l(yi, i) + ∑(k=1 to K) Ω(fk) Where l is the loss function, Ω is the regularization term, fk is the decision tree; θ represents the parameter set of the model; yi represents the true value of the i-th financing amount; i represents the model's predicted value for the i-th financing amount; K represents the total number of decision trees in the model.

5. According to claim 1, the intelligent enterprise financing graph system based on distributed architecture, It is characterized in that The 3D visualization module includes: using WebGL technology to implement a 3D force-directed graph, and the node position update formula is: x = x + (Fx / |F|) * min(|F|,temperature) y = y + (Fy / |F|) * min(|F|, temperature) Among them, temperature is the cooling parameter that controls the node movement speed; x and y represent the current coordinates of the node on the two-dimensional plane; Fx and Fy represent the components of the resultant force acting on the node in the x and y directions; F represents the magnitude of the resultant force acting on the node, |F|=√(Fx² + Fy²).

6. The intelligent enterprise financing graph system based on distributed architecture according to claim 1, It is characterized in that The blockchain data recording module includes: using the distributed ledger Hyperledger Fabric to build a permissioned blockchain network, adopting the PBFT consensus algorithm, and the view number v calculation formula is: v = (v + 1) mod 2^64.

7. The intelligent enterprise financing graph system based on distributed architecture according to claim 1, It is characterized in that The security management module includes: using AES-256 algorithm to encrypt data, including round key addition, byte replacement, row shift and column mixing; implementing role-based access control and supporting fine-grained authority management.

8. An intelligent enterprise financing graph device based on distributed architecture, It is characterized in that Equipped with the intelligent enterprise financing map system based on distributed architecture as described in any one of claims 1 to 7.

Images