CN117371940A

CN117371940A - Holographic intelligent control method and system for financial credit and debit management

Info

Publication number: CN117371940A
Application number: CN202311334297.0A
Authority: CN
Inventors: 杨阳; 张劲; 张莉
Original assignee: DIGITAL CHINA ADVANCED SYSTEMS SERVICES CO LTD
Current assignee: DIGITAL CHINA ADVANCED SYSTEMS SERVICES CO LTD
Priority date: 2023-10-16
Filing date: 2023-10-16
Publication date: 2024-01-09

Abstract

The invention belongs to the technical field of information supervision, and discloses a holographic intelligent control method and system for financial credit and debit management. The data collection module is used for automatically collecting the credit attribute data related to the project; the model training module performs model training based on the collected data and supports various machine learning algorithms; the matching execution module runs the model in the starting stage of the new project, automatically identifies the most matched credit attribute with the project, and allocates resources and strategies for the project. The invention integrates the credit attribute and the traditional project management, and creates a project management tool of the credit attribute through the credit management flow; and simultaneously, a unified credit and debit knowledge base is created, and all delivery pieces are managed and filed in a unified way. The invention can improve 50-75% of the efficiency of the credit-wound project management, enhance the communication coordination efficiency among departments and improve the safety of the project management.

Description

Holographic intelligent control method and system for financial credit and debit management

Technical Field

The invention belongs to the technical field of information supervision, and particularly relates to a holographic intelligent control method, a holographic intelligent control system, a holographic intelligent control medium and computer equipment for financial information creation management.

Background

At present, after the supervision puts forward the requirement on the credit substitution ratio, when the bank manages the credit project, the credit project needs to be executed by combining the supervision requirement and project management experience. In the existing industry management mode, most of the traditional project management software is used for performing credit-creation project management, and project starting, project planning, project execution and monitoring and project ending processes are used for performing project management by default. In the project management process, each flow progress is universal, and project management work of a universal project can be completed. And the special configuration requirements and management requirements in the credit items cannot be satisfied.

In the project management process, each flow progress is universal, and project management work of a universal project can be completed. Because of the specificity of the created project, many indexes, processes and parameters focused in project management do not have relevant parameter configuration capability in the traditional project management tools. Meanwhile, the content of the credit and debit project management needed by different financial institutions is different, and when the financial institutions use project management software to manage credit and debit projects, a large amount of offline management work is needed to be matched, so that the management efficiency is very low. Meanwhile, documents involved in project management are difficult to manage, and the project management method has challenges for project knowledge accumulation and credit and debit project confidentiality work. The existing project management system has no project management function provided for the credit attribute requirement, and the credit project has extremely strong credit attribute and needs to carry out the transformation, the flow modification and the configuration update of the requirement for the supervision requirement. The traditional project management software with functions cannot provide convenient and fast flow change operation, and cannot meet the requirements of the credit on high-frequency and multi-demand of financial institutions. In project management, related delivery members and file management are conventionally performed by performing document management and collection by means of centralized uploading or staged uploading of documents. In this process, there is a risk of omission, loss, leakage, etc.

For the analysis, the prior art has the following main defects and technical problems to be solved in industrial application:

1. the particularity of the credit item cannot be satisfied: traditional project management tools are mainly designed for general projects, and often cannot provide effective support and management for projects with specificity such as credit-creation projects, such as focused indexes, flows, parameters and the like.

2. Lack of personalized configuration: the content of the credit creation project management faced by different financial institutions is different, but the traditional project management tools often lack enough personalized configuration capability and cannot meet the specific requirements of different institutions.

3. Inefficient management: because the traditional project management tool cannot completely meet the management requirement of the credit-created project, financial institutions often need to cooperate with a large amount of offline management work, which greatly reduces the efficiency of project management.

4. Document management is difficult: in project management, related documents and data are often scattered in various places, so that effective management and confidentiality are not easy to perform, and meanwhile, the accumulation and inheritance of project knowledge are challenged.

5. Lack of pertinence: the credit items have extremely strong characteristics and need to be subjected to required transformation, flow modification and configuration update aiming at supervision requirements. However, conventional project management software often cannot provide convenient and fast process change operations, and cannot meet the requirement of frequent changes.

6. Document management risk: in traditional project management, uploading and management of documents often have risks of omission, loss, leakage and the like, and smooth project progress and information safety cannot be guaranteed.

In summary, the technical problems to be solved include: how to meet the specificity of the created project, provide personalized project management tools, improve project management efficiency, improve document management modes, meet the targeted requirements and reduce document management risks.

Disclosure of Invention

Aiming at the problems existing in the prior art, the invention provides a holographic intelligent control method, a system, a medium and computer equipment for financial information creation management.

The invention is realized in that a system for holographic intelligent control of financial information creation management comprises:

a data collection module for automatically collecting data of the credit attributes related to the plurality of completed and ongoing projects, the module receiving the data through API interface, document parsing or manual input;

a model training module for performing model training based on the collected data; the module can support a plurality of machine learning algorithms including but not limited to support vector machines, logistic regression and neural networks, and employ random gradient descent for model training;

And the matching execution module is used for running the trained model in the starting stage of the new project so as to automatically identify the credit attribute which is most matched with the project and allocate corresponding resources and strategies for the project.

Further, the system further comprises:

a supervision requirement analysis sub-module for analyzing rules and guidelines text related to financial originals using natural language processing algorithms, particularly word frequency-inverse document frequency models and long and short term memory networks;

and the automatic updating rule base sub-module is used for automatically updating or inserting corresponding supervision rules and requirements in the project management tool according to the analysis result.

Further, the system further comprises:

a real-time analysis sub-module for performing real-time analysis, in particular time series analysis using dynamic time warping;

and the decision support sub-module is used for automatically triggering corresponding management adjustment and intervention measures based on the real-time analysis result, and can continuously optimize the decision effect by utilizing the reinforcement learning algorithm.

Further, the system further comprises:

a knowledge graph construction sub-module for coding and connecting all the information creation attributes related to the project and the flow nodes and storing the information in a graph database;

And the recommendation system sub-module is used for automatically providing a recommendation and a solution for project managers when the project managers need to make decisions by applying a collaborative filtering algorithm based on the knowledge graph and the historical data.

Further, the system further comprises:

a hash value generation sub-module for generating a unique hash value for each delivery piece; regenerating a hash value each time the delivery piece is modified or updated; storing the hash value for each delivery member in an accessible database or storage facility; creating and maintaining a version history record containing all hash values for each delivery for subsequent tracking and auditing;

a security checking sub-module, configured to perform integrity checking on data stored in the system using a hash function and a merck tree algorithm; when the data is changed, automatically starting the integrity check; periodically maintaining and updating the merck tree to ensure that the data it reflects is up to date; providing a corresponding alarm mechanism when data inconsistency is detected;

an automatic label generation sub-module for content analysis of all deliveries using natural language processing and image recognition algorithms;

and a metadata generation sub-module for automatically generating tags and metadata associated with the delivery member based on the results of the content analysis, the metadata being available for further searching and management.

Another object of the present invention is to provide a method for holographic intelligent control of financial information creation management based on the above system, which is characterized in that the method comprises the following steps:

automatically collecting data of the credit attribute related to a plurality of completed and ongoing projects through an API interface, document analysis or manual input mode;

based on the collected data, performing model training by using a support vector machine, logistic regression and a neural network machine learning algorithm, and performing model optimization by applying random gradient descent;

analyzing the rule or guideline text related to financial credit by using a word frequency-inverse document frequency model and a long-short-term memory network;

according to the analysis result, automatically updating or inserting corresponding supervision rules and requirements in the project management tool;

running the trained model to automatically identify the most matched credit attribute with the new project and allocate corresponding resources and strategies for the project;

and implementing dynamic time warping to perform time sequence analysis, and automatically triggering corresponding management adjustment and intervention measures by using a reinforcement learning algorithm.

Further, the method further comprises:

step one, coding and connecting all the credit attribute related to the project and the flow node by using a knowledge graph, and storing the information in a graph database;

Step two, providing decision suggestions for project managers based on the knowledge graph and the historical data by using a collaborative filtering algorithm;

generating hash values for all delivery pieces and storing the hash values so as to ensure that all modifications and updates can be effectively tracked and verified;

step four, content analysis is carried out on all delivery pieces by using natural language processing and an image recognition algorithm;

and fifthly, automatically generating labels and metadata related to the delivery piece based on the content analysis result.

Further, the first step comprises the following steps:

defining the credit attribute related to the project and the flow node as the entity and the relation in the knowledge graph;

constructing a knowledge graph by using programming languages and libraries, wherein nodes in the knowledge graph represent entities and edges represent relationships among the entities;

storing the constructed knowledge graph in a graph database;

the knowledge graph is queried using the query language of the graph database to find the credit attributes associated with a particular item or process node.

Further, the second step comprises the following steps:

preparing a knowledge graph and historical data, wherein the knowledge graph comprises credit attributes related to the project and the flow nodes, and the historical data comprises decisions made by a project manager in past projects and results of the projects;

Constructing a user-project matrix, wherein rows represent project managers and columns represent projects, and each element represents the evaluation of the project manager on the project or the success rate of the project;

calculating the similarity between rows or columns in the user-project matrix;

a recommendation is generated based on the similarity, other items most similar to each item are found for each item, and then the items are recommended to the item manager.

Further, the method of step three comprises the steps of:

selecting a one-way hash function;

generating a hash value for each delivery piece, and generating a unique character string corresponding to the content of the delivery piece one by transferring the content of the delivery piece as input to a selected hash function;

storing the hash value of each delivery piece in a safe place, generating a new hash value and storing each time the delivery piece is updated;

when the modification of the delivery member needs to be verified, a hash value is generated again for the delivery member, and the newly generated hash value is compared with the stored hash value to verify whether the delivery member is modified.

Further, the implementation steps of the fourth step are as follows:

the first step: natural language processing analysis delivery piece text content

Natural language processing is used to analyze the textual content of the delivery member. The method comprises the following steps:

1. Data preprocessing: including cleaning data (e.g., eliminating punctuation and stop words), converting to lowercase, etc.

2. Word segmentation: a process of decomposing text into words or phrases.

3. Word vectorization: converting the text data into digital form allows the model to understand.

4. Model training: the model is trained using algorithms (e.g., deep learning or machine learning) to understand and analyze the text data.

And a second step of: image recognition algorithm for analyzing image content of delivery member

The image recognition is used to analyze the image content of the delivery member. The method comprises the following steps:

1. image preprocessing: including adjusting the image size, normalizing, etc.

2. Feature extraction: useful features are extracted from the image, such as color, shape, texture, etc.

3. Model training: the image data is understood and analyzed using some algorithm (e.g., deep learning or machine learning) to train the model.

For the above two steps, a pre-training model such as the BERT or res net model is pre-trained on a large amount of data, processing new, unseen data.

And a third step of: generating tags and metadata based on content analysis results

And generating labels and metadata related to the delivery piece according to the result of the steps. For example, if the NLP model finds that text mainly surrounds the "environmental" topic, a "environmental" tag is added. Also, if the image recognition model finds a lot of green in the image, a "green" label is added.

Metadata is data about data that describes some basic characteristics of the data. Such as the date of creation, author, size, etc. of the delivery.

The above is an implementation flow. The specific implementation can be adjusted according to the requirements and resources. In practical application, optimization and debugging are also needed, and the accuracy and efficiency of the model are improved.

In combination with the technical scheme and the technical problems to be solved, the technical scheme to be protected has the following advantages and positive effects:

firstly, the invention has obvious technical progress in the aspect of financial credit management, integrates a plurality of steps such as data collection, model training, matching execution, automatic updating rule and the like into one system, can effectively improve the management efficiency and standardability of financial credit projects, simultaneously reduces the investment of manpower resources, and realizes the intellectualization and automation of financial credit management.

The invention combines the credit attribute with the traditional project management capability to design the best practice of credit project management tool; and combining the supervision requirement with the credit creation project management, and creating a project management tool of the credit creation attribute through a credit creation management flow. Meanwhile, a unified credit and debit knowledge base is created, delivery pieces related to all flow nodes in credit and debit project management are managed in a unified mode, unified archiving is conducted in a process management mode, and safety and manageability of files are guaranteed.

Second, the invention reduces management efficiency in the financial institution credit project management process. Compared with the application of the traditional project management tool in the credit and trauma project management, the financial credit and trauma management holographic system solves the problem that the traditional project management mode cannot be attached to the pain points of the actual demand of the credit and trauma project through the process setting of credit and trauma attributes and a flexible process change mechanism, reduces the management efficiency of a financial institution in the actual credit and trauma project management process, does not need to pass a large number of verification and attempt to comb out the management mode, and provides a credit and trauma project management optimal path (which can be customized according to the actual condition of the institution) through the industry best practice provided by the system, so that 50 to 75 percent of credit and trauma project management efficiency is improved.

Meanwhile, by the system, the communication and coordination efficiency of each credit management department is improved, sensitive data and information in credit projects are uniformly managed and monitored by the system, and the safety of credit project management is improved.

Thirdly, the technical progress of the invention is as follows:

1. automation and machine learning integration: by integrating various data collection, analysis and model training means, the manual intervention is greatly reduced, and the efficiency and accuracy are improved.

2. The multiple modules work cooperatively: the collaborative work of a plurality of modules such as data collection, model training, matching execution and the like ensures comprehensive and deep information creation management.

3. Real-time resolution of regulations and supervision: through natural language processing algorithm, the system can analyze rules and guidelines related to financial credit and debit in real time, and the compliance risk is greatly reduced.

4. Dynamic adjustment and decision support: the system can perform real-time analysis, can automatically trigger management adjustment and intervention measures based on the reinforcement learning algorithm, and further increases management flexibility.

5. Knowledge graph and recommendation system: through a knowledge graph and a collaborative filtering algorithm, the system can provide more accurate decision suggestions for project managers.

6. Content analysis and metadata management: through natural language processing and image recognition techniques, the system can automatically generate tags and metadata associated with the delivery, which not only facilitates subsequent searching and management, but also increases the value of the data asset.

7. Task and milestone intelligent management: through data analysis and machine learning, the system can automatically perform task decomposition and allocation, as well as predict and adjust milestones of the project.

8. Intelligent supervision and reporting: the system can automatically generate various reports meeting the supervision requirements, and is integrated with the supervision and assessment system, so that the acceptance process of the project is greatly accelerated.

The invention clearly shows the uniqueness and advancement of the system and the method in the aspect of financial credit management, and has obvious technical progress and practical value.

Drawings

FIG. 1A is a schematic diagram of a financial information creation management holographic intelligent control system provided by an embodiment of the invention;

FIG. 1B is a schematic diagram of a financial information creation management holographic intelligent control system according to an embodiment of the invention;

FIG. 1C is a schematic diagram of a financial information creation management holographic intelligent control system provided by an embodiment of the invention;

FIG. 1D is a schematic diagram of a financial information creation management holographic intelligent control system provided by an embodiment of the invention;

FIG. 2 is a flowchart of a method for intelligently controlling financial information and creation management holographic according to an embodiment of the present invention;

FIG. 3 is a flowchart of an intelligent adaptive project management method according to an embodiment of the present invention;

FIG. 4 is a flow chart of a method for intelligent task and milestone management provided by an embodiment of the invention;

FIG. 5 is a flow chart of an intelligent supervision and reporting method according to an embodiment of the present invention;

fig. 6 is a schematic diagram of a holographic intelligent control method for financial information creation management according to an embodiment of the present invention.

Detailed Description

The present invention will be described in further detail with reference to the following examples in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.

The invention provides a system for holographic intelligent control of financial information creation management, which comprises:

a data collection module for automatically collecting data of the credit attributes associated with the plurality of completed and ongoing projects, the module receiving the data via an API interface, document parsing or manual input. The data collection module is primarily tasked with automatically collecting and collating data for the credit attributes associated with a plurality of completed and ongoing items. The data sources include API interfaces, document parsing, or manual input.

The following is the working principle of the data collection module:

1) The API interface receives data: the module automatically obtains data related to the project by calling a specific API interface. For example, when project information is stored in a remote database or cloud service, the API requests to retrieve such data. In practice, HTTP requests (e.g., GET or POST requests) are used to call these APIs and retrieve the returned data.

2) Document parsing receives data: when project-related data is stored in documents, such as PDF, word, or Excel files, the module uses document parsing techniques (such as OCR techniques or text parsing algorithms) to read and parse the documents to obtain the desired data.

3) Manual input of received data: in some cases, manual data entry is required. For example, some data cannot be obtained automatically through API or document parsing, or when manual review and verification is required, a User Interface (UI) is constructed to receive manual input from the user.

After data is collected, the module performs data cleaning and preprocessing to ensure the quality and consistency of the data, making it suitable for subsequent model training and analysis. The processing includes processing missing values, outliers, converting data types, selecting features, engineering features, etc.

And the model training module is used for carrying out model training based on the collected data. The module can support a variety of machine learning algorithms including, but not limited to, support Vector Machines (SVMs), logistic regression, and neural networks, and model training using optimization algorithms such as random gradient descent (SGD).

The implementation of the model training module based on Python is as follows. The scheme uses a 'scikit-learn' and a machine learning library, which supports a plurality of machine learning algorithms. To achieve training of the neural network model, 'Keras' is used, which is an advanced API for deep learning, and random gradient descent (SGD) is used as an optimizer.

```python

from sklearn import svm

from sklearn.linear_model import LogisticRegression

from keras.models import Sequential

from keras.layers import Dense

from keras.optimizers import SGD

from sklearn.model_selection import train_test_split

# load data

#X,y

# dividing data set

X_train,X_test,y_train,y_test＝train_test_split(X,y,test_size＝0.2,random_state＝42)

# support vector machine

svm_clf＝svm.SVC()

svm_clf.fit(X_train,y_train)

svm_predictions＝svm_clf.predict(X_test)

# logistic regression

logreg_clf＝LogisticRegression()

logreg_clf.fit(X_train,y_train)

logreg_predictions＝logreg_clf.predict(X_test)

# neural network

nn_model＝Sequential()

nn_model.add(Dense(12,input_dim＝8,activation＝'relu'))

nn_model.add(Dense(8,activation＝'relu'))

nn_model.add(Dense(1,activation＝'sigmoid'))

sgd＝SGD(lr＝0.01,decay＝1e-6,momentum＝0.9,nesterov＝True)

nn_model.compile(loss＝'binary_crossentropy',optimizer＝sgd,metrics＝['accuracy'])

nn_model.fit(X_train,y_train,epochs＝150,batch_size＝10)

nn_predictions＝nn_model.predict_classes(X_test)

```

The code segments are first loaded with data and then divided into training and testing sets. Next, a model is created and trained for each algorithm (support vector machine, logistic regression, and neural network) and then predictions are made on the test set.

For neural networks, a simple multi-layer perceptron (MLP) model is used, consisting of two hidden layers and one output layer. A random gradient descent (SGD) optimizer is used and sets a learning rate (lr), a weight decay (decay), a momentum (momentum), and a Nesterov momentum (Nesterov).

Implementation it is assumed that the data has been properly preprocessed and formatted, and that these machine learning algorithms are suitable for use. For neural networks, the structure and parameters of the model need to be adjusted to suit specific tasks and data.

The match execution module will use the trained model to predict new items and automatically identify the attributes that best match the items. Based on these attributes, the module will automatically allocate the corresponding resources and policies. The following is a simple implementation scheme based on Python:

In the above code, the "Matcher" class predicts new project data ("new_project_data") using a pre-trained model (here, "trained_model"), and the result is the attribute of the project ("attributes"). Then, the 'assignment_resources' method assigns corresponding resources and policies for the items according to the attributes.

As shown in fig. 1A, the system for intelligently controlling financial information and creation management according to the embodiment of the present invention includes:

the data collection module is used for automatically collecting data of the credit attribute related to a plurality of completed and ongoing projects, and can receive the data through an API interface, document analysis or manual input mode;

the model training module is connected with the data collection module and is used for carrying out model training based on the collected data; the module can support a plurality of machine learning algorithms including but not limited to support vector machines, logistic regression and neural networks, and employ random gradient descent for model training;

and the matching execution module is connected with the model training module and is used for running the trained model in the starting stage of the new project so as to automatically identify the credit attribute which is most matched with the project and allocate corresponding resources and strategies for the project.

The specific working principle of the three modules is as follows:

and a data collection module: the module is responsible for automatically collecting data of the credit attributes associated with numerous completed and ongoing projects. The data receiving mode comprises an API interface, document analysis and manual input. The API interface is used for exchanging data with other systems or services; document parsing processes documents of various formats, such as PDF, word or Excel files, from which required data is extracted; the manual input processes data that cannot be obtained in an automated manner.

Model training module: the module is connected with the data collection module and performs model training based on the collected data. The module supports numerous machine learning algorithms including, but not limited to, support Vector Machines (SVMs), logistic regression (Logistic Regression), and Neural Networks (Neural Networks). Each algorithm has the characteristics and the application field, and the algorithm is selected and used according to the requirements and the data characteristics. In the model training process, the optimization algorithm adopts random gradient descent (Stochastic Gradient Descent, SGD) to search the optimal parameters of the model.

And the matching execution module is used for: the module is connected with a model training module and is responsible for running a trained model in a new project starting stage, automatically identifying the most matched credit attribute with the project, and distributing corresponding resources and strategies for the project. The module inputs project related information by using a trained model, and the model outputs the credit attribute which is most matched with the project. Based on the attributes, the system automatically allocates the most appropriate resources and policies for the project.

As shown in fig. 1B, the system further includes:

the supervision requirement analysis module is used for analyzing rules and guidelines text related to financial originals by using a natural language processing algorithm, in particular a word frequency-inverse document frequency model and a long-short time memory network;

the automatic updating rule base module is connected with the supervision requirement analysis module and is used for automatically updating or inserting corresponding supervision rules and requirements in the project management tool according to analysis results;

the real-time analysis module is connected with the automatic updating rule base module and is used for implementing real-time analysis, in particular to time sequence analysis by using dynamic time warping;

the decision support module is connected with the real-time analysis module and used for automatically triggering corresponding management adjustment and intervention measures based on the real-time analysis result, and the sub-module can continuously optimize the decision effect by utilizing the reinforcement learning algorithm.

The working principle of the four modules is as follows:

the regulatory requirement parsing module applies Natural Language Processing (NLP) algorithms, particularly word frequency-inverse document frequency (TF-IDF) models and Long and Short Term Memory (LSTM) networks, to parse rules or guideline text related to financial credit. -word frequency-inverse document frequency (TF-IDF) model: the statistical method is used to evaluate the importance of a term to a document in a document set or corpus. The TF-IDF value increases as the number of occurrences of a term in a document increases, while decreasing as the frequency of its occurrence in the corpus increases. -Long and Short Time Memory (LSTM) network: the LSTM network is a special RNN designed to solve the long-term dependency problem of RNNs. LSTM networks can remember long-term patterns in training data, with an effect on parsing complex regulatory or guideline text.

The automatic updating rule base module is connected with the supervision requirement analysis module, and corresponding supervision rules and requirements are automatically updated or inserted into the project management tool according to analysis results. The module contains a rules engine that parses the administrative rules and converts them into executable project management rules.

The real-time analysis module is connected with the automatic updating rule base module for implementing real-time analysis. In particular, dynamic Time Warping (DTW) is used for time series analysis. DTW is a method of comparing two time series even though they are somewhat different in time axis or change in rate. DTW is useful for real-time monitoring and identification of rule violations.

The decision support module is connected with the real-time analysis module, and corresponding management adjustment and intervention measures are automatically triggered according to the real-time analysis result. The module utilizes reinforcement learning algorithms to continuously optimize decision effects. Reinforcement learning is a machine learning method in which an agent learns how to behave in an environment to maximize a jackpot. In this case, the circumstances are the compliance of the regulatory rules, rewards are the degree of compliance of the rules, or penalties/penalties avoided, etc.

As shown in fig. 1C, the system further includes:

the knowledge graph construction module is used for encoding and connecting all the credit attributes related to the project and the flow node and storing the information in a graph database;

and the recommendation system module is connected with the knowledge graph construction module and is used for automatically providing a recommendation and a solution for project managers when the project managers need to make decisions based on the knowledge graph and the historical data by applying a collaborative filtering algorithm.

The knowledge graph construction module is responsible for collecting, processing, encoding and connecting all information related to the project and the process nodes. Such information includes project requirements, team member skills, project schedules, progress, risk, and other factors that affect project success.

The working principle of the knowledge graph construction module comprises the following steps:

1) And (3) data collection: the modules collect data from various sources, such as project documents, meeting records, team feedback, and other sources containing useful information.

2) And (3) data processing: the module cleans and preprocesses the collected data, removes noise, solves missing values and abnormal values, and makes the data suitable for further analysis.

3) Information coding: the module encodes the preprocessed data into a format that is understood and processed by the computer. This mainly involves converting text data into numbers or symbols for storage and querying in a graph database.

4) And (3) information connection: the module connects the encoded information to form a comprehensive knowledge graph. This includes identifying items, people, tasks, and other entities and attributes related to the links.

The recommendation system module automatically provides suggestions and solutions for project managers in decision making by applying a collaborative filtering algorithm based on the knowledge graph and the historical data.

The working principle of the recommendation system module is as follows:

1) And (3) data acquisition: the module obtains necessary information from knowledge maps and historical data, such as status of projects, team member's ability, past decisions and results.

2) User-item matrix generation: the module generates a user-project matrix in which each element represents a user's (project manager) rating or feedback on a project.

3) Collaborative filtering: the module analyzes the user-project matrix using collaborative filtering algorithms to find similar projects or users, providing suggestions for project managers. Collaborative filtering algorithms are of two types: collaborative user-based filtering and collaborative project-based filtering.

4) Recommendation generation: the module generates recommendations based on the results of collaborative filtering, including actions to be taken, risks to be avoided, and resources or skills required.

The knowledge graph provides data and context for generating recommendations that the recommendation system uses to help project managers make better decisions.

As shown in fig. 1D, the system further comprises:

a hash value generation sub-module for generating a unique hash value for each delivery piece; regenerating a hash value each time the delivery piece is modified or updated; storing the hash value for each delivery member in an accessible database or storage facility; a version history record containing all hash values for each delivery is created and maintained for subsequent tracking and auditing.

The security checking sub-module performs integrity checking on data stored in the system by using a hash function and a merck tree algorithm; when the data is changed, automatically starting the integrity check; periodically maintaining and updating the merck tree to ensure that the data it reflects is up to date; when a data inconsistency is detected, a corresponding alert mechanism is provided.

The automatic label generating module is connected with the security checking module and is used for carrying out content analysis on all delivery pieces by using natural language processing and an image recognition algorithm;

and the metadata generation module is connected with the automatic label generation module and is used for automatically generating labels and metadata related to the delivery piece based on the result of the content analysis, and the metadata can be used for further searching and management.

As shown in fig. 2, the method for holographic intelligent control of financial information creation management based on the system provided by the embodiment of the invention comprises the following steps:

s1: automatically collecting data of the credit attribute related to a plurality of completed and ongoing projects through an API interface, document analysis or manual input mode;

s2: based on the collected data, performing model training by using a support vector machine, logistic regression and a neural network machine learning algorithm, and performing model optimization by applying random gradient descent;

s3: analyzing the rule or guideline text related to financial credit by using a word frequency-inverse document frequency model and a long-short-term memory network;

s4: according to the analysis result, automatically updating or inserting corresponding supervision rules and requirements in the project management tool;

s5: running the trained model to automatically identify the most matched credit attribute with the new project and allocate corresponding resources and strategies for the project;

s6: and implementing dynamic time warping to perform time sequence analysis, and automatically triggering corresponding management adjustment and intervention measures by using a reinforcement learning algorithm.

S7: coding and connecting all the information creation attributes related to the project and the flow nodes by using the knowledge graph, and storing the information in a graph database; providing decision suggestions for project managers based on the knowledge graph and the historical data by using a collaborative filtering algorithm;

S8: generating hash values for all deliveries and storing to ensure that all modifications and updates can be effectively tracked and verified;

s9: the hash function and digital signature ensure data integrity, calculate a hash value for each delivery, and sign the hash value using a private key, thereby verifying data integrity and origin. Natural language processing and image recognition algorithms analyze all delivery piece content to understand the meaning of the data and enable more efficient searches and queries. Based on the content analysis results, tags and metadata associated with the deliveries are generated, which are used to organize and categorize the data for convenient retrieval and use. The distributed database stores all deliveries, relevant tags, metadata and digital signatures, and simultaneously records the modification history of each delivery to achieve data traceability.

As shown in fig. 3, an embodiment of the present invention provides an intelligent adaptive project management method, including:

s11: automatically generating or optimizing a workflow template based on historical project data and a standard flow of the existing financial industry by using a machine learning algorithm;

s12: natural language processing and data analysis techniques are integrated to automatically detect and identify important supervisory changes associated with a project and send early warning to project managers and related team members in real time.

As shown in fig. 4, an embodiment of the present invention provides an intelligent task and milestone management method, including:

s21: automatically decomposing tasks according to project requirements and targets by utilizing a machine learning algorithm, and automatically distributing the tasks according to the capabilities and resources of team members;

s22: automatically predicting project progress using data analysis techniques and adjusting milestones according to the prediction; the system will automatically send notification of milestone adjustments to project managers and team members.

As shown in fig. 5, an embodiment of the present invention provides an intelligent supervision and reporting method, including:

s31: automatically generating various reports meeting the supervision requirements, such as progress reports and financial reports, through data analysis and text generation technologies;

s32: collecting and analyzing each dimension data of the item in real time, and automatically generating daily reports, weekly reports and monthly reports according to the data;

s33: and the system is integrated with a supervision and assessment system, and a machine learning algorithm is used for automatically carrying out preliminary document and information examination so as to accelerate the acceptance process of the project.

In the embodiment of the present invention, as shown in fig. 6, the financial credit management holographic system of the embodiment of the present invention creates an integrated financial industry credit management tool by accumulation of best practices in industry, in combination with management functions for start, execution, monitoring, and ending in the project management tool. The system designs the credit-creating project management function specially aiming at the financial industry according to the supervision requirement and the demand of the financial institution on credit-creating reconstruction, and focuses more on the flow setting of the credit-creating project management requirement. Users can classify different projects through the credit and debit project attributes, and provide corresponding project milestone time nodes, delivery member requirements and templates of the nodes, problem management, report management and other functions. Meanwhile, aiming at the actual conditions of different financial institutions and the different management granularities, the system supports the customization of the management flow, and a user can manage and modify the management flow of the credit-created project according to the change of the supervision requirement in the project execution process, so that the result of rapidly adapting to the conditions of different credit-created projects is achieved.

In the embodiment of the invention, the financial credit management holographic system monitors and manages the life cycle and the supervision and assessment requirements of the whole financial credit project in an omnibearing way through the attention and design of five types of objects. The five objects are organization and role, credit engineering, assessment index, project group and report.

In the embodiment of the invention, in the project and the role, the financial credit management holographic system divides the organization into three major categories of credit lead group, credit execution group and credit review expert. The credit-wound lead group consists of a line lead and related responsible persons, and plays a role in deciding on the whole credit-wound project; the credit execution group consists of a project group manager, a project manager and project members and is responsible for the relevant floor work of credit projects; the review expert consists of enterprises and professional third party teams, and promotes and optimizes the standard reaching condition of the credit establishment of the financial institutions.

In the embodiment of the invention, in the credit engineering, a financial credit management holographic system is integrally unfolded in combination with credit work of a financial institution for 5-8 years, and the credit architecture transformation of the financial institution is realized by focusing on the settlement from the aspects of organization construction, fund input condition, annual task completion, acceptance data collection and credit capability. The powerful knowledge base management mechanism aims at orderly management of templates, schemes, plans and the like of the credit task, provides process document management for problems, reports and solutions in the project delivery process, opens resources for project wholesalers and effectively promotes orderly expansion of credit projects. In the future, a strong problem-solving knowledge base is provided through accumulation of industry best practices, and the industry enterprises are helped to carry out related credit-creating and customs-striving work.

In the embodiment of the invention, the financial information and creation management holographic system carries out index decomposition aiming at each subclass around the four major classes and twenty-two subclasses of the supervision requirements, and maps the delivery object, the assessment score and the supervision requirements one by one, thereby providing real-time monitoring and management of the supervision requirement completion condition for the financial institutions.

In the embodiment of the invention, the financial credit management holographic system is supported by annual promotion of the whole project around the organization of the entity or virtual project group, and provides support for orderly, high-quality and efficient completion of the project. Firstly, decomposing projects according to an office system, a financial machine, a general service system, a key service system, an attack task and other six tasks, and carrying out project propulsion by a project group; secondly, dividing the project into three types of milestone types of a business system, basic software and hardware and an attack and closure task, and customizing the project in a differentiated way according to construction standards and actual conditions of the project; again, predefining the corresponding deliveries for each milestone and providing the relevant deliveries templates issued by the supervision as relevant references for the financial institution; finally, setting up a problem risk management mechanism, wherein the risk and the problem can be built in the project group and the project, and the problem risk solving capability in the project management process is enhanced.

In the embodiment of the invention, the financial credit management holographic system simultaneously provides report management functions for supervision, project group and project group. Supervision-oriented: the planning stage system can manage the reporting scheme and related documents, the implementation stage can carry out data statistics and collection in a month report mode and the like, and the project acceptance stage can carry out data model inspection acceptance in a corresponding delivery report mode in the supervision and inspection system. The management conditions of each dimension of the project can be collected and managed by a daily report, weekly report and monthly report management system of best practice.

As shown in fig. 3, the financial information creation management holographic intelligent control system provided by the embodiment of the invention includes:

the organization type dividing module is used for dividing the organization into three major categories, namely a credit-wound leading group, a credit-wound executing group and a credit-wound review expert; the credit-wound lead group consists of a line lead and related responsible persons, and plays a role in deciding on the whole credit-wound project; the credit execution group consists of a project group manager, a project manager and project members and is responsible for the relevant floor work of credit projects; the review expert consists of enterprises and professional third party teams and promotes and optimizes the standard reaching condition of the credit establishment of the financial institutions;

The credit architecture transformation module is used for combining the construction of a financial institution from an organization, the investment condition of funds, the annual task completion degree, the acceptance data collection and credit ability precipitation to realize the credit architecture transformation of the financial institution;

the assessment index requirement module is used for surrounding the four major classes and twenty-two minor classes of the supervision requirements, the financial information creation management holographic system carries out index decomposition aiming at each minor class, and maps delivery objects, assessment score items and supervision requirements one by one, thereby providing real-time monitoring and management of supervision requirement completion conditions for financial institutions;

and the whole project supporting module is used for organizing around the entity or virtual project group, and supporting the ordered, high-quality and high-efficiency completion of the projects through annual pushing support of the whole projects.

Aiming at the concept and the strategy of the holographic intelligent control method for financial credit and debit management, the following technical angle of application is specifically realized.

The embodiment of the invention provides a holographic intelligent control method for financial information creation management, which comprises the following steps:

1. intelligent matching of credit attributes to project management capabilities

And (3) data collection: data about the nature of the originality (e.g., trust, innovation, etc.) is collected from each completed and ongoing project.

Model training: the completed project is analyzed and cases of success and failure are marked. The model training is performed using algorithms such as SVM, random forest or neural network.

Implementation and application: in the new project start-up phase, the model is run to identify the most matching credit attributes to the project.

2. Real-time supervision and adaptive adjustment

Analysis of supervision requirements: and analyzing the text of the rule or the guideline by using an NLP algorithm, and extracting key information.

Automatically updating a rule base: a rules engine is designed to automatically update the project management tool whenever a new regulatory requirement occurs.

3. Intelligent management of a credit knowledge base

Knowledge graph construction: and encoding and connecting the information creation attributes of all the items and the flow nodes. These relationships are stored using a graph database.

Recommendation system: based on the historical data and the knowledge graph, a recommendation algorithm is designed. When project manager needs to make decision, advice is automatically provided.

4. Intelligent process management and unified archiving

Automatic label generation: the delivery is content analyzed using NLP and image recognition algorithms. Based on the analysis results, tags and metadata are automatically generated.

The following is a specific implementation method for each sub-module or step:

1. Intelligent matching of credit attributes to project management capabilities

And (3) data collection: data is automatically crawled from various project management tools or databases through the API interface.

A data standardization and cleaning flow is designed.

Model training: model training was performed using the Scikit-learn library of Python or the TensorFlow framework. The dataset needs to be annotated to identify which items are successful and which are failed.

Implementation and application: the trained model is integrated into the project management tool through RESTful APIs or micro-service architecture.

2. Real-time supervision and adaptive adjustment

Analysis of supervision requirements: legal text is segmented, part-of-speech tagged, etc., using an NLP library (e.g., space or NLTK).

Automatically updating a rule base: using triggers or Webhook, when a new regulatory requirement is issued, a rule update process is automatically started.

And (3) self-adaptive adjustment: real-time analysis is performed using a real-time data stream processing framework (e.g., apache Kafka or Spark Streaming).

3. Intelligent management of a credit knowledge base

Knowledge graph construction: the map database (e.g., neo4 j) is used to store the credit attributes and project relationships.

Recommendation system: recommendation is performed using collaborative filtering or content-based recommendation algorithms.

4. Intelligent process management and unified archiving

Automatic label generation: the text delivery is analyzed using the NLP library. For images or other non-text deliverables, an image recognition library (e.g., openCV) may be used.

As a specific application of the present invention, a virtual project management software company is described below as a detailed example.

1) Intelligent matching of credit attributes to project management capabilities

Model training: the model was trained using the SVM algorithm of the Scikit-learn library. The input data includes various item indexes, member evaluations, etc., and the label is "success" or "failure".

Implementation and application: the model is integrated into the company's internal project management platform, and when a new project is started, the model recommends which trust attributes (trust, innovation, etc.) are best suited for the project.

2) Real-time supervision and adaptive adjustment

Automatically updating a rule base: the rule base is stored in an SQL database, and the newly analyzed rule is automatically added by the Python script.

Knowledge graph construction: and using a Neo4j graph database to establish the relation between nodes and edges by using all the items and the credit attributes.

Recommendation system: when a project manager queries a decision problem, the system recommends a solution using a content-based recommendation algorithm based on knowledge maps and historical data.

3) Intelligent process management and unified archiving

Automatic label generation: the text is parsed using the API, and tags and metadata are then automatically generated based on the results of these analyses.

By these specific embodiments, the company can efficiently perform project management while also ensuring efficient dissemination of regulatory compliance and knowledge.

The processing process of the signals and the data in the embodiment of the invention comprises the following steps:

data collection and pretreatment: various types of data such as text, numbers, images, sounds, etc. are collected. And (5) preprocessing technologies such as standardization, normalization, denoising and the like are applied.

Feature engineering and model training: and performing feature engineering by using PCA, feature selection algorithm and the like. And performing model training by applying algorithms such as gradient lifting, deep learning and the like.

Real-time analysis and decision support: real-time analysis is performed using time-series analysis, anomaly detection algorithms, and the like. If the score output by the model is below a certain threshold, management adjustments are automatically triggered.

Post-processing and feedback loop: feedback is collected using a/B tests, model evaluation metrics (e.g., F1 score, accuracy, etc.). The model is trimmed or retrained.

In the process of combining the mathematical algorithm and the model of artificial intelligence to intellectualize the technical scheme, the following steps are adopted:

1. Data collection and preprocessing

In the data collection phase, various data for project management, such as project progress, cost, quality, risk, etc., may be collected from multiple sources by means of automated data collection tools and processed using specific data cleansing and normalization methods in preparation for subsequent machine learning model training.

2. Feature engineering and model training

At this stage, features in the project management data may be automatically extracted using artificial intelligence techniques such as deep learning, natural language processing, etc., such as extracting keywords and emotion information in the project document using natural language processing techniques, and then training a machine learning model using these features to predict the outcome of the project management.

3. Real-time analysis and decision support

At this stage, project management may be monitored and analyzed in real-time using artificial intelligence techniques, such as predicting trends in project progress, cost, etc. using machine learning models, and using such information to provide decision support for project management, such as solutions and adjustment strategies provided to project management personnel using recommendation systems.

4. Post-processing and feedback loop

At this stage, project summary reports and best practices can be automatically generated using artificial intelligence techniques, and these information are used to feed back into model training, iterating model optimization, and continuously improving project management effects.

Through the intelligent improvement scheme, the project management process is automated, manual intervention is reduced, and project management efficiency and quality are improved. Meanwhile, through continuous data accumulation and model optimization, the method and the system can better understand and master the project management rule and provide more valuable support for future project management.

As an optimization scheme of the embodiment of the invention, the method specifically comprises the following steps:

intelligent matching of credit attributes to project management capabilities

According to the invention, the attribute of project manager and the requirement of the project are matched by using a Support Vector Machine (SVM), and the model is optimized by random gradient descent (SGD), so that the matching accuracy is improved. Attribute data and project demand data of project manager are collected. The data is pre-processed, e.g., normalized. Training is performed by using an SVM model, and SGD is optimized. And carrying out matching prediction on the new project manager or project.

1) And (3) data collection: attribute data and project demand data of project manager are collected. This should include a range of features such as project manager experience, skill, past project success rates, etc., as well as detailed information about project needs such as the size of the project, expected completion time, skill required, etc.

2) Data preprocessing: the step of data preprocessing includes normalizing the data (so that all features are within the same range), processing missing values, or performing feature engineering, such as creating new features or deleting irrelevant features.

3) Model training: training is performed by using an SVM model. In the training process, the attributes of the project manager and the project requirements are used as input, and the matched project manager and project are used as targets. This step requires parameter adjustments to find the optimal model parameters.

4) Model optimization: the model training process is optimized by using SGD. SGD is an effective optimization algorithm, finding the optimal solution of model parameters. This step requires adjustment of the learning rate or other super-parameters to optimize the model learning effect.

5) Prediction and evaluation: after model training and optimization are completed, matching prediction is carried out on new project management personnel or projects by using the model. At the same time, the prediction performance of the model is evaluated, for example, by calculating the accuracy of prediction or the precision recall, etc.

(II) real-time supervision and adaptive adjustment

The invention analyzes real-time data by using LSTM network, analyzes time sequence data by using Dynamic Time Warping (DTW), and makes real-time decision and adjustment by reinforcement learning.

And performing natural language processing by using a TF-IDF model, and extracting features. And applying the LSTM model to analyze the time sequence data. The similarity analysis of the time series was performed using the DTW algorithm. And combining reinforcement learning to make real-time decision and adjustment.

(III) Intelligent management of information creation knowledge base

Establishing a knowledge graph by utilizing graph theory, recommending information by collaborative filtering, and finding out the most relevant information in the knowledge graph by utilizing PageRank algorithm.

And establishing a knowledge graph by using graph theory. And applying a collaborative filtering algorithm to conduct knowledge recommendation. The most relevant knowledge is found using the PageRank or shortest path algorithm.

(IV) Intelligent process management and unified archiving

1) Data integrity assurance by hash function and Merkle Tree

The hash function and Merkle Tree ensure data integrity. The hash value of each file is stored in a database after being calculated, and the data tampering condition is checked by utilizing Merkle Tree. When the file is modified, the hash value is recalculated, compared with the hash value stored in the database, and if the hash value is different, the file is indicated to be modified.

2) Random forest for tag generation and document classification

The random forest is used as a machine learning algorithm to generate labels and classified documents. Training the random forest model predicts new document tags using tagged documents.

3) K-means or hierarchical clustering for document classification archiving

And taking the K-means and hierarchical clustering as an unsupervised learning algorithm to classify and archive the documents. The random forest generated tags are used as input, and the clustering algorithm classifies documents with similar tags into one category.

Fifth, data collection and preprocessing

The invention performs data description through mean value and variance in statistics and performs data preprocessing through Z-score normalization and PCA.

The data were descriptive analyzed using statistical methods. Data preprocessing and dimension reduction were performed using Z-score normalization and PCA.

Feature engineering and model training

The invention performs feature screening by using analysis of variance (ANOVA), selects the most useful features by information gain, and performs feature selection by using forward selection and backward cancellation.

And performing analysis of variance and information gain calculation. Forward selection and backward elimination of the screening feature are applied. The model is trained using the screened features.

Seventh real-time analysis and decision support

The invention performs time series analysis by using ARIMA model or Fourier transform and performs state estimation by using Kalman filtering.

The ARIMA model or Fourier transform is applied to perform time series analysis. The system state is estimated using kalman filtering.

Eighth post-processing and feedback loop

The invention carries out model evaluation through a confusion matrix and an ROC curve, and uses gradient lifting or AdaBoost to carry out model optimization.

Model performance evaluation was performed using confusion matrix and ROC curve. Model optimization is performed by gradient lifting or AdaBoost.

By integrating the various models and algorithms, an intelligent system is established by the method of the invention, so as to realize holographic intelligent control of financial credit and debit management, thereby improving efficiency and accuracy.

The embodiment of the invention also provides an intelligent self-adaptive project management method, which comprises the following steps:

the workflow templates are automatically generated or optimized based on historical project data and standard flow of the existing financial industry using machine learning algorithms. Natural Language Processing (NLP) and data analysis techniques are integrated to automatically detect and identify important regulatory changes associated with a project and send early warning to project managers and related team members in real time.

The embodiment of the invention also provides an intelligent knowledge base and document management method, which comprises the following steps:

task templates, solutions and plans that are most relevant to the current project are automatically searched and recommended in the knowledge base using NLP and machine learning algorithms. Advanced text generation algorithms are employed to automatically create documents related to the project delivery process, such as scheduling, execution reports, delivery reports, etc.

The embodiment of the invention also provides an intelligent task and milestone management method, which specifically comprises the following steps:

and automatically decomposing the task according to project requirements and targets by utilizing a machine learning algorithm, and automatically distributing the task according to the capabilities and resources of team members. Project progress is automatically predicted using data analysis techniques and milestones are adjusted based on the prediction. The system will automatically send notification of milestone adjustments to project managers and team members.

The embodiment of the invention also provides an intelligent supervision and reporting method, which specifically comprises the following steps:

various reports meeting the supervision requirements, such as progress reports, financial reports and the like, are automatically generated through data analysis and text generation technologies. And collecting and analyzing each dimension data of the project, such as time schedule, quality, cost and the like, and automatically generating daily reports, weekly reports and monthly reports according to the data. And the system is integrated with a supervision and assessment system, and a machine learning algorithm is used for automatically carrying out preliminary document and information examination so as to accelerate the acceptance process of the project.

It should be noted that the embodiments of the present invention can be realized in hardware, software, or a combination of software and hardware. The hardware portion may be implemented using dedicated logic; the software portions may be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or special purpose design hardware. Those of ordinary skill in the art will appreciate that the apparatus and methods described above may be implemented using computer executable instructions and/or embodied in processor control code, such as provided on a carrier medium such as a magnetic disk, CD or DVD-ROM, a programmable memory such as read only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The device of the present invention and its modules may be implemented by hardware circuitry, such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., as well as software executed by various types of processors, or by a combination of the above hardware circuitry and software, such as firmware.

The foregoing is merely illustrative of specific embodiments of the present invention, and the scope of the invention is not limited thereto, but any modifications, equivalents, improvements and alternatives falling within the spirit and principles of the present invention will be apparent to those skilled in the art within the scope of the present invention.

Claims

1. A system for holographic intelligent control of financial credit management, comprising:

a model training module for performing model training based on the collected data; the module can support a plurality of machine learning algorithms, including a support vector machine, a logistic regression and a neural network, and performs model training by applying random gradient descent;

2. The system for holographic intelligent control of financial information creation management of claim 1, further comprising:

3. The system for holographic intelligent control of financial information creation management of claim 1, further comprising:

4. The system for holographic intelligent control of financial information creation management as in claim 1, further comprising:

5. The system for holographic intelligent control of financial information creation management as in claim 1, further comprising:

6. A method for holographic intelligent control of financial credit management based on the system of any of claims 1 to 6, the method comprising the steps of:

7. The method for holographic intelligent control of financial information creation management of claim 7, the method further comprising:

step four, content analysis is carried out on all delivery pieces by using natural language processing and an image recognition algorithm; based on the results of the content analysis, tags and metadata associated with the delivery are automatically generated.

8. The method for holographic intelligent control of financial information creation management of claim 7, wherein step one comprises the steps of:

Storing the constructed knowledge graph in a graph database;

9. The method for intelligent control of financial information creation management hologram according to claim 7, wherein step two comprises the steps of:

calculating the similarity between rows or columns in the user-project matrix;

10. The method for intelligent control of financial information creation management hologram according to claim 7, wherein said method of step three comprises the steps of:

selecting a one-way hash function;