Assistant decision-making system and method for water transportation engineering project management based on intelligent agent

Abstract

The invention discloses an agent-based water transportation engineering project management aid decision-making system and method. The system comprises a data acquisition unit; the data storage unit is used for storing the collected data; the data preprocessing unit is used for preprocessing the data acquired by the data acquisition unit; the knowledge graph construction unit is used for constructing a knowledge base; the updating and maintaining unit is used for regularly updating and maintaining the knowledge base so as to reflect the latest industry specifications, engineering cases and scientific achievements; the intelligent model unit is used for establishing a model; and the model training unit is used for training the model. According to the invention, improvement is carried out from four aspects of automatic data processing, efficient and accurate integration of a comprehensive knowledge graph, establishment of an intelligent model, training for application and decision support and optimization, an efficient, accurate and high-adaptability auxiliary decision-making system and method are provided, and the efficiency and quality of project management are greatly improved.

Description

Technical Field

[0001] The present invention belongs to the field of electronic digital data processing, and more specifically, it particularly relates to a water transport engineering project management auxiliary decision-making system based on an intelligent agent. Meanwhile, the present invention also relates to a water transport engineering project management auxiliary decision-making method based on an intelligent agent. Background Art

[0002] Water transport project management is a complex task involving project planning, risk assessment, resource allocation and other aspects. Traditional water transport project management methods rely on manual processing of large amounts of data and information, which is not only inefficient but also prone to errors. Project managers need to process data from different sources, including text materials, drawings, reports, etc. The integration and analysis of these data places high demands on managers.

[0003] Existing data collection mainly relies on manual entry and processing, which is not only time-consuming and labor-intensive, but also increases the risk of data errors. Irregularities in data storage and preprocessing make it difficult to ensure data reliability and integrity, affecting the accuracy of subsequent data analysis and project management decisions. In addition, although the field of water transport engineering has rich expertise and experience, the existing knowledge management system fails to effectively integrate and apply this knowledge. The transfer and sharing of knowledge mostly rely on personal experience and oral communication, and lack of systematic and formal knowledge management tools, resulting in inefficient knowledge utilization. Therefore, we propose an agent-based water transport engineering project management decision-making support system and method. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings existing in the prior art and propose an intelligent agent-based water transport engineering project management decision-making support system and method. The present invention makes improvements in four aspects: automated data processing, efficient and accurate integration of comprehensive knowledge graphs, establishment, training and application of intelligent models, and decision support and optimization. It provides an efficient, accurate and adaptable decision-making support system and method, which greatly improves the efficiency and quality of project management.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The agent-based water transport engineering project management auxiliary decision-making system includes:

[0007] A data layer, the data layer comprising:

[0008] Data collection unit, using OCR to automatically identify and extract water transport engineering data from various documents;

[0009] A data storage unit for storing the collected data, wherein the data storage unit needs to ensure the reliability and integrity of the data;

[0010] The data preprocessing unit is used to preprocess the data collected by the data collection unit, including data cleaning and format conversion, to ensure data quality and the efficiency of subsequent processing;

[0011] The knowledge management layer includes:

[0012] The knowledge graph construction unit is used to integrate multiple data types such as text, images, and videos in the field of water transport engineering to build a knowledge base;

[0013] Update and maintenance unit, used to regularly update and maintain the knowledge base to reflect the latest industry standards, engineering cases and scientific research results. Maintenance includes the addition, update and abandonment of knowledge.

[0014] An intelligent processing layer, the intelligent processing layer comprising:

[0015] An intelligent model unit, used to establish a model that simulates the decision-making process for project planning, risk assessment, and resource allocation;

[0016] The model training unit is used to train the model to ensure accurate output results.

[0017] The method for auxiliary decision-making of water transport engineering project management based on intelligent agent adopts the above system and comprises the following steps:

[0018] S1. Extract data from water transport engineering documents, drawings and reports using optical character recognition;

[0019] S2. Preprocess the extracted data, including data cleaning and format conversion, to improve data quality and processing efficiency;

[0020] S3. Build and maintain a knowledge graph containing knowledge in the field of water transport engineering, which integrates structured and unstructured data and is regularly updated to reflect the latest industry specifications, engineering cases and scientific research results;

[0021] S4. Build and train machine learning-based models to simulate and predict planning, risk assessment, and resource allocation for water transport engineering projects;

[0022] S5. Coordinate the outputs of all models through the decision support module and output decision support;

[0023] S6. Establish a feedback mechanism to collect feedback and project results data to continuously optimize the model and knowledge graph.

[0024] Preferably, step S1 specifically comprises:

[0025] S11. Convert all physical documents into digital images through scanners;

[0026] S12, determine the layout of text, tables and graphics, and adjust the clarity, noise removal and brightness of the image;

[0027] S13, converting the graphic text in the image into an editable text format;

[0028] S14, correcting errors that occur during recognition;

[0029] S15. Integrate the identified data with data collected from existing databases, real-time data streams, and other data sources to form extracted data.

[0030] Preferably, in step S14, the error in correcting the identification is specifically:

[0031] S141, segment the given sentence, and the segmentation result will serve as the basis for the next step;

[0032] S142, finding a confusion set of corresponding words in the sentence, wherein the confusion set is a pre-prepared set consisting of common confusing characters, including similar pronunciations and similar glyphs;

[0033] S143, for each replaceable character in the sentence, enumerate each character in its confusion set to replace the original character, and obtain a candidate sentence set;

[0034] S144, calculating the score of each candidate sentence, and the sentence with the highest score will be selected as the final output;

[0035] The score calculation formula for calculating each candidate sentence in step S144 is as follows:

[0036] ;

[0037] Among them, the conditional probability of each item can be calculated by maximum likelihood estimation, and the expression is:

[0038] ;

[0039] In the formula, N(c l−1 ,c l ) is a binary group (c l−1 ,c l ) appears in the data, N(c l−1 ) is the character c l−1 The number of occurrences; the above formula is calculated through the following process:

[0040] Count each tuple (c l−1 ,c l ) appears at a frequency N(c l−1 ,cl ) and a single character c l−1 The frequency of occurrence N(c l−1 );

[0041] According to the above formula, the conditional probability P(c l ∣c l−1 ), the probability represents that given the previous character c l−1 In the case of l Probability of occurrence;

[0042] Multiply all the conditional probabilities together to get the probability of the entire string, and determine the final output based on the probability of the entire string.

[0043] Preferably, the step S3 of constructing and maintaining a knowledge graph containing knowledge in the field of water transport engineering is specifically as follows:

[0044] S31. Integrate structured data, semi-structured data and unstructured data in the field of water transport engineering, including industry specifications, engineering cases and scientific research results. The integrated data is stored in the form of RDF or graph database;

[0045] S32. Construct a knowledge graph ontology, wherein the knowledge graph ontology needs to clarify concepts and entities within the domain and determine the logical relationships between them;

[0046] S33. Design a conceptual hierarchy, build a hierarchical relationship between concepts, and ensure the accuracy and hierarchy of each concept;

[0047] S34, identifying named entities specific to the field of water transport engineering from the original data, such as project names, technical terms, and locations, and extracting relationships between entities based on the identified entity and relationship patterns;

[0048] Among them, the basic unit of the knowledge graph is a triple consisting of entity-relationship-entity, which is used to present complex information in a simple and clear manner.

[0049] Preferably, the establishment and training of the model based on machine learning in step S4 specifically adopts a regression model or a classification model. If the regression model is used to predict a specific value, the classification model is used if the classification problem includes risk level and feasibility.

[0050] The training and verification of the model specifically includes:

[0051] Data segmentation, dividing the dataset into training and test sets, or using cross-validation methods to evaluate the performance of the model on unseen data;

[0052] Model training, using training set data to train the model, and adjusting model parameters through optimization algorithms to minimize prediction errors;

[0053] Model validation, evaluating model performance through validation sets or cross-validation to ensure that the model has good generalization ability;

[0054] The step S4 also includes model evaluation and optimization, specifically:

[0055] Performance indicators, use accuracy, recall, and F1 score indicators to evaluate the performance of classification models, and for regression models, use the mean square error indicator;

[0056] Hyperparameter tuning optimizes model performance by adjusting the number of trees in a random forest or the depth of a decision tree in a machine learning algorithm.

[0057] Preferably, step S5 is specifically:

[0058] S51. According to the specific requirements of the project and the risk tolerance, set the decision threshold and establish a rule engine to automatically make decisions according to the set rules;

[0059] S52. Collect data from each model, including time prediction data, cost prediction data, and risk assessment data;

[0060] S53, cleaning the collected data, including removing outliers and processing missing values, and then normalizing the data to convert it into a unified dimension;

[0061] S54, assigning different weights to each model, and integrating the outputs of each model using a weighted average formula to obtain a comprehensive prediction result;

[0062] S55. Define the objective function of decision optimization, choose to minimize total cost or maximize benefits, and establish constraints based on the actual situation of the project;

[0063] S56. Based on the constraints and the specifically defined decision optimization objective function, output a detailed report of the decision, including charts, data analysis results and specific recommendations.

[0064] Preferably, in step S54, the weight distribution expression is:

[0065] ;

[0066] Where C is the comprehensive cost forecast, C1, C2, C 3、 C n are the output results of the model, w1, w 2、 w3 and w n is the weight of the one-to-one corresponding model output result;

[0067] The weight is calculated by calculating the performance ratio. For n models, the accuracy of each model i is p i , the performance ratio r of each model i , the accuracy of each model j is p j， The expression is:

[0068] ;

[0069] Since the weight w of each model i The performance ratio can be used directly, and the sum of all weights is equal to 1, which is expressed as:

[0070] ;

[0071] Since the sum of all weights equals 1, the weight expression for each model is:

[0072] ;

[0073] In step S56, the constraint conditions and the specifically defined decision optimization objective function are expressed as follows:

[0074] ;

[0075] Where 0≤x i ≤B i , i=1,2,...,n;

[0076] In the above formula, x i is the investment amount of the i-th project, B is the total investment, and B i is the maximum investment of the ith project, which includes resource volume, production capacity, and risk tolerance.

[0077] Preferably, step S6 specifically comprises:

[0078] S61. Clarify the purpose of collecting feedback, including improving model performance, increasing user satisfaction, or optimizing knowledge graph content. Based on the purpose, determine key performance indicators to measure the effect of feedback on system improvement.

[0079] S62. Choose the appropriate feedback collection method based on the target audience;

[0080] S63. Clean and classify the collected data to facilitate subsequent analysis and processing, and start a feedback mechanism to actively encourage users to provide feedback;

[0081] S64. Identify commonalities or important trends from the feedback through qualitative analysis and rank them according to their importance and urgency to determine priorities for processing;

[0082] S65. Based on the collected feedback, formulate a specific action plan to improve the knowledge graph and actually apply the improvement measures to the system.

[0083] Technical effects and advantages of the present invention: Compared with the prior art, the intelligent agent-based water transport engineering project management decision-making support system and method provided by the present invention has the following advantages:

[0084] By using OCR technology to automatically extract data from documents, combined with pre-processing steps of data cleaning and format conversion, the system significantly improves the efficiency and accuracy of data processing. This process reduces manual entry errors and saves a lot of human resources, allowing the project management team to focus more on decision-making and analysis;

[0085] By building a comprehensive knowledge graph that includes multiple data types such as text, images, and videos, we have achieved systematic and formal management of knowledge. This integration has promoted the accumulation, transfer, and sharing of knowledge, improved the efficiency of knowledge utilization, and provided rich information support for project decision-making.

[0086] By building and training machine learning-based models, we can simulate and predict key aspects of water transport engineering projects, including project planning, risk assessment, and resource allocation. The model building and verification process ensures the accuracy of the output results and provides a powerful decision support tool for project management.

[0087] The output of all models is coordinated through the decision support module, which optimizes the project management process. Through the automated decision-making process, the project planning cycle is shortened, the timeliness and effectiveness of risk management and the rationality of resource allocation are improved, and the efficiency and quality of project management are improved overall.

[0088] Project data and user feedback are continuously collected through the feedback mechanism to optimize the model and knowledge graph. The iterative update mechanism enables the system to adapt to changing project requirements and complex engineering environments, and continuously improve system performance and user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0089] Figure 1 This is an architecture diagram of the water transport engineering project management auxiliary decision-making system based on the intelligent agent of the present invention;

[0090] Figure 2 It is a flow chart of the agent-based decision-making support method for water transport engineering project management of the present invention;

[0091] Figure 3 The following is a specific flow chart of extracting data in an embodiment of the present invention. DETAILED DESCRIPTION

[0092] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0093] The present invention provides an agent-based water transport engineering project management decision-making support system and method, which has the following advantages:

[0094] Automated data processing uses OCR technology to automatically identify and extract water transport engineering data from various documents, significantly improving the efficiency and accuracy of data collection; through data cleaning and format conversion, data quality and processing efficiency are guaranteed; comprehensive knowledge graphs are built to include a variety of data types such as text, images, and videos in the field of water transport engineering. Through regular updates and maintenance, the latest industry specifications, engineering cases, and scientific research results are effectively integrated and reflected;

[0095] Establishment and training of intelligent models. A machine learning-based model was established to simulate and predict key aspects of water transport engineering project planning, risk assessment and resource allocation. Model training and verification were used to ensure that the model could accurately output results and provide reliable decision support for project management. Decision support and optimization were carried out through the decision support module to coordinate the output of all models and provide comprehensive decision support for water transport engineering project management. At the same time, a feedback mechanism was established to continuously collect project data and user feedback for optimizing models and knowledge graphs, thereby improving the overall performance of the system and user satisfaction.

[0096] like Figure 1 As shown, the system includes:

[0097] Data Layer

[0098] The data acquisition unit uses optical character recognition technology to automatically identify and extract water transport engineering data from various documents such as texts, drawings and reports; multi-source data integration, in addition to traditional text materials, also integrates information from sensors, monitoring systems and other real-time data sources.

[0099] The data storage unit stores the collected data. The data storage unit needs to ensure the reliability and integrity of the data and implement regular data backup and disaster recovery plans to prevent data loss or damage.

[0100] The data preprocessing unit is used to preprocess the data collected by the data acquisition unit, including data cleaning and format conversion, to ensure data quality and efficiency of subsequent processing, and convert the data into a unified format for subsequent processing and analysis.

[0101] Knowledge management layer

[0102] The knowledge graph construction unit is used to integrate multiple data types such as text, images, and videos in the field of water transport engineering to build a knowledge base; identify and establish associations and links between different data to form a networked knowledge structure.

[0103] The update and maintenance unit is used to regularly update and maintain the knowledge base to reflect the latest industry standards, engineering cases and scientific research results. Maintenance includes the addition, update and abandonment of knowledge, and uses user feedback and actual operation results as part of knowledge updating to continuously optimize the accuracy and practicality of the knowledge base.

[0104] Intelligent processing layer

[0105] The intelligent model unit builds models to simulate the planning, risk assessment and resource allocation decision-making process of water transport engineering projects, and adopts machine learning models to adapt to different decision-making needs.

[0106] The model training unit is used to train the model to ensure that it can accurately output results and verify the performance and generalization ability of the model through cross-validation and actual test data sets.

[0107] The system integration and optimization includes:

[0108] User interface, providing a friendly user interface that enables users to easily interact with the system, input data and obtain decision support.

[0109] Feedback loop, establish a feedback mechanism, collect user feedback and project results data, and use it to continuously optimize the model and knowledge graph.

[0110] Security and privacy protection ensures the security of the system, protects data from unauthorized access, and complies with relevant data protection regulations.

[0111] The above improvements in four aspects, namely, automated data processing, efficient and accurate integration of comprehensive knowledge graphs, establishment, training and application of intelligent models, and decision support and optimization, provide an efficient, accurate and adaptable decision support system and method, greatly improving the efficiency and quality of project management.

[0112] This embodiment also provides Figure 2 The agent-based water transport engineering project management decision-making support method shown in the figure adopts the above system and includes the following steps:

[0113] S1. Extract data from water transport engineering documents, drawings and reports using optical character recognition; Figure 3 , step S1 is specifically as follows:

[0114] S11. Convert all physical documents into digital images through scanners;

[0115] S12, determine the layout of text, tables and graphics, and adjust the clarity, noise removal and brightness of the image;

[0116] S13, converting the graphic text in the image into an editable text format;

[0117] S14, correcting errors that occur during identification; wherein, in step S14, correcting errors that occur during identification, specifically:

[0118] S141, segment the given sentence, and the segmentation result will serve as the basis for the next step;

[0119] S142, finding a confusion set of corresponding words in the sentence, where the confusion set is a pre-prepared set of common confusing characters, including similar pronunciations and similar glyphs;

[0120] S143, for each replaceable character in the sentence, enumerate each character in its confusion set to replace the original character, and obtain a candidate sentence set;

[0121] S144, calculating the score of each candidate sentence, and the sentence with the highest score will be selected as the final output;

[0122] The score calculation formula for calculating each candidate sentence in step S144 is as follows:

[0123] ;

[0124] Among them, the conditional probability of each item can be calculated by maximum likelihood estimation, and the expression is:

[0125] ;

[0126] In the formula, N(c l−1 ,c l ) is a binary group (c l−1 ,c l ) appears in the data, N(c l−1 ) is the character c l−1 The number of occurrences; the above formula is calculated through the following process:

[0127] Count each tuple (c l−1 ,c l ) appears at a frequency N(c l−1 ,c l ) and a single character c l−1 The frequency of occurrence N(c l−1 );

[0128] According to the above formula, the conditional probability P(c l ∣c l−1 ), the probability represents that given the previous character c l−1 In the case of l Probability of occurrence;

[0129] Multiply all conditional probabilities to get the probability of the entire string, and determine the final output based on the probability of the entire string;

[0130] S15, integrating the identified data with data collected from existing databases, real-time data streams, and other data sources to form extracted data;

[0131] For example, in order to correct errors, the probability score of each candidate sentence is calculated. Assume that relevant conditional probability data calculated from a large number of water transport engineering documents are available. These data can be used to calculate the probability score of each sentence. If the conditional probabilities of "the channel depth must reach 5 meters" and "the bow depth must reach 5 meters" are required, the specific results are as follows:

[0132] P(channel|)=0.01; P(depth|channel)=0.8; P(must|depth)=0.7; P(reach|must)=0.6; P(5 meters|reach)=0.9;

[0133] According to the above, the probability score is calculated as follows:

[0134] P(channel|)×P(depth|channel)×P(must|depth)×P(reach|must)×P(5m|reach)=0;

[0135] According to the above calculation, the probability scores of the two sentences are compared. Since "the channel depth must reach 5 meters" has a higher score, it is selected as the final output, thus correcting the OCR's misrecognition.

[0136] S2, preprocessing the extracted data, including data cleaning and format conversion, to improve data quality and processing efficiency; step S2 is specifically:

[0137] S21: Data cleaning

[0138] S211: Remove irrelevant data:

[0139] Purpose: To ensure that the dataset contains only information related to water transport engineering project management;

[0140] Method: Screen and remove all irrelevant data entries using predefined keywords and project requirements;

[0141] S212: Correcting incorrect data:

[0142] Challenge: Character recognition errors that may be introduced during the OCR conversion process;

[0143] Solution: Use an automatic correction algorithm to correct incorrect text recognition;

[0144] S213: Handling missing values:

[0145] Problem: Document damage or scanning problems may result in missing information;

[0146] Treatment: Fill missing values ​​as appropriate, using the mean, median, or mode, or remove the entire column if there are too few data points;

[0147] S214: Eliminate duplicate data:

[0148] Necessity: Multiple data sources may generate duplicate data entries;

[0149] Implementation: Ensure the uniqueness of each data by detecting and deleting duplicate rows;

[0150] S22: Format Conversion

[0151] S221: Unified data format:

[0152] Purpose: To unify data from all data sources into a standard format for easy subsequent processing;

[0153] Method: Convert date, number and text formats to ensure they are consistent across all documents;

[0154] S222: Data type conversion:

[0155] Requirements: Some analytical tools and models require specific data types;

[0156] Operation: Convert a string to a numeric value, or vice versa, to meet the needs of data analysis;

[0157] S223: Standardization and Normalization:

[0158] Purpose: To adjust the data scale to a standard range, such as [0,1] or [-1,1];

[0159] Methods: Apply standardization (e.g., Z-score standardization) or normalization techniques to improve the performance of the algorithm;

[0160] S224: Encoding conversion:

[0161] Scenario: When processing categorical data, some models require numerical input;

[0162] Technique: Convert categorical variables into numerical form using one-hot encoding or label encoding.

[0163] S3, construct and maintain a knowledge graph containing knowledge in the field of water transport engineering, which integrates structured and unstructured data and is regularly updated to reflect the latest industry standards, engineering cases and scientific research results; Step S3 constructs and maintains a knowledge graph containing knowledge in the field of water transport engineering, specifically:

[0164] S31. Integrate structured data, semi-structured data and unstructured data in the field of water transport engineering, including industry specifications, engineering cases and scientific research results. The integrated data is stored in the form of RDF or graph database;

[0165] S32. Construct the knowledge graph ontology. The knowledge graph ontology needs to clarify the concepts and entities in the field and determine the logical relationships between them.

[0166] S33. Design a conceptual hierarchy, build a hierarchical relationship between concepts, and ensure the accuracy and hierarchy of each concept;

[0167] S34, identifying named entities specific to the field of water transport engineering from the original data, such as project names, technical terms, and locations, and extracting relationships between entities based on the identified entity and relationship patterns;

[0168] Among them, the basic unit of the knowledge graph is a triple consisting of entity-relationship-entity, which is used to present complex information in a simple and clear way.

[0169] S4, establishing and training a model based on machine learning for simulating and predicting the planning, risk assessment and resource allocation of water transport engineering projects; the establishment and training of the model based on machine learning in step S4 specifically adopts a regression model or a classification model. If the regression model is used for predicting specific values, the classification model is used for classification problems including risk level and feasibility;

[0170] The training and verification of the model specifically includes:

[0171] Data segmentation, dividing the dataset into training and test sets, or using cross-validation methods to evaluate the performance of the model on unseen data;

[0172] Model training, using training set data to train the model, and adjusting model parameters through optimization algorithms to minimize prediction errors;

[0173] Model validation, evaluating model performance through validation sets or cross-validation to ensure that the model has good generalization ability;

[0174] Step S4 also includes model evaluation and optimization, specifically:

[0175] Performance indicators, use accuracy, recall, and F1 score indicators to evaluate the performance of classification models, and for regression models, use mean square error indicators;

[0176] Hyperparameter tuning optimizes model performance by adjusting the number of trees in a random forest or the depth of a decision tree in a machine learning algorithm.

[0177] S5. The outputs of all models are coordinated through the decision support module, and decision support is output; step S5 is specifically as follows:

[0178] S51. According to the specific requirements and risk tolerance of the project, set the decision threshold and establish a rule engine to automatically make decisions according to the set rules;

[0179] S52. Collect data from each model, including time prediction data, cost prediction data, and risk assessment data;

[0180] S53, cleaning the collected data, including removing outliers and processing missing values, and then normalizing the data to convert it into a unified dimension;

[0181] S54, assign different weights to each model, and use the weighted average formula to integrate the outputs of each model to obtain a comprehensive prediction result; In step S54, the weight assignment expression is:

[0182] ;

[0183] Where C is the comprehensive cost forecast, C1, C2, C 3、 C n are the output results of the model, w1, w 2、 w3 and w n is the weight of the one-to-one corresponding model output result;

[0184] The weights are calculated by calculating the performance ratio. For n models, the accuracy of each model i is p i , the performance ratio r of each model i , the accuracy of each model j is p j， The expression is:

[0185] ;

[0186] Since the weight w of each model i The performance ratio can be used directly, and the sum of all weights is equal to 1, which is expressed as:

[0187] ;

[0188] Since the sum of all weights equals 1, the weight expression for each model is:

[0189] ;

[0190] In step S56, the constraint conditions and the specifically defined decision optimization objective function are expressed as follows:

[0191] ;

[0192] Where 0≤x i ≤B i , i=1,2,...,n;

[0193] In the above formula, x i is the investment amount of the i-th project, B is the total investment, and B i is the maximum investment of the i-th project, which includes resources, production capacity, and risk tolerance;

[0194] S55. Define the objective function of decision optimization, choose to minimize total cost or maximize benefits, and establish constraints based on the actual situation of the project;

[0195] S56. Based on the constraints and the specifically defined decision optimization objective function, output a detailed report of the decision, including charts, data analysis results and specific recommendations.

[0196] For example, let's put the above into practice. Let's assume that we are managing a water transport project that includes multiple sub-projects, including dredging, wharf construction, and waterway development. Each sub-project has different cost forecasts, time forecasts, and risk assessment data. The goal is to maximize the overall benefits of the project within a limited budget while controlling risks, as follows:

[0197] Setting decision thresholds and rules engine:

[0198] Decision threshold: Based on the specific requirements of the project, we set a threshold of cost overrun not exceeding 10% and risk level below medium.

[0199] Rules Engine: If the cost forecast of any sub-project exceeds the budget by 10% or the risk assessment result is high, the system will automatically mark it and recommend re-evaluation or adjustment plan.

[0200] Collect model data:

[0201] Collect data from the time prediction model, cost prediction model and risk assessment model of each sub-project.

[0202] Data cleaning and normalization:

[0203] The collected data is subjected to outlier removal, for example, predictions where the cost is much lower than the market price, and missing values ​​are processed and filled with the mean value.

[0204] Normalization is performed to convert all cost data into a unified dimension for easy comparison and integration.

[0205] Weight distribution and comprehensive prediction:

[0206] Suppose we have three models: cost prediction model C1, time prediction model C2 and risk assessment model C3.

[0207] According to historical data, the model's accuracy is p1=0.9, p2=0.85, and p3=0.95 respectively.

[0208] Calculate the performance ratio: r1=w1 / p1,r2=w2 / p2,r3=w3 / p3.

[0209] Since the sum of all weights is 1, we have w1+w2+w3=1.

[0210] By solving this system of equations, we find the weight of each model, and then use the weighted average formula to integrate the outputs of each model to obtain the comprehensive prediction result, as follows:

[0211] Calculate the weights of each model, w1, w2, and w3, based on the model's performance ratios r1, r2, and r3 and the model's accuracy p1, p2, and p3. i Defined as weight w i Divide by the accuracy p i The goal is to find weights such that the sum of all weights equals 1.

[0212] Known data: p1=0.9; p2=0.85; p3=0.95;

[0213] The relationship between weight and performance ratio, r1=w 1 / p1; r2==w 2 / p2; r3==w 3 / p 3；

[0214] The constraint of total weight, w1+w2+w3=1;

[0215] Calculating weights: Since the sum of all weights equals 1, use the following steps to calculate each weight:

[0216] Calculate the performance ratio of each model, r1=w1 / 0.9; r2=w2 / 0.85; r3=w3 / 0.95

[0217] Assume that the performance ratios are equal, that is, r1=r2=r3=r:r=w1 / 0.9=w2 / 0.85=w3 / 0.95;

[0218] Since w1+w2+w3=1, express w1, w2, and w3 as functions of r: w1=0.9r; w2=0.85r; w3=0.95r;

[0219] Substitute the total weight constraint: r+0.85r+0.95r=1;

[0220] Simplified: 2.7r=1;

[0221] The solution is: r = 1 / 2.7 ≈ 0.3704;

[0222] Calculate each weight:

[0223] w1=0.9×0.3704≈0.3334;

[0224] w2=0.85×0.3704≈0.3148;

[0225] w3=0.95×0.3704≈0.3518;

[0226] According to the above calculation, the weight of each model is as follows:

[0227] w1≈0.3334;

[0228] w2≈0.3148;

[0229] w3≈0.3518;

[0230] These weights are used in a weighted average formula to combine the outputs of each model to obtain a composite forecast.

[0231] Define the decision optimization objective function:

[0232] In this case, we choose to minimize the total cost as the optimization objective.

[0233] The constraints include: the investment amount of each sub-project does not exceed its maximum investment, and the total investment does not exceed the budget.

[0234] Output decision report:

[0235] Based on the defined constraints and optimization objective function, the system outputs a detailed decision report.

[0236] The report includes charts such as the cost distribution diagram of each sub-project, data analysis results such as which projects are over budget or have too high risks, and specific suggestions such as adjusting the budget of a sub-project or changing the execution strategy.

[0237] S6. Establish a feedback mechanism to collect feedback and project result data to continuously optimize the model and knowledge graph; Step S6 is specifically as follows:

[0238] S61. Clarify the purpose of collecting feedback, including improving model performance, increasing user satisfaction, or optimizing knowledge graph content. Based on the purpose, determine key performance indicators to measure the effect of feedback on system improvement.

[0239] S62. Choose the appropriate feedback collection method based on the target audience;

[0240] S63. Clean and classify the collected data to facilitate subsequent analysis and processing, and start a feedback mechanism to actively encourage users to provide feedback;

[0241] S64. Identify commonalities or important trends from the feedback through qualitative analysis and rank them according to their importance and urgency to determine priorities for processing;

[0242] S65. Based on the collected feedback, formulate a specific action plan to improve the knowledge graph and actually apply the improvement measures to the system.

[0243] The present invention has the following advantages:

[0244] By using OCR technology to automatically extract data from documents, combined with pre-processing steps of data cleaning and format conversion, the system significantly improves the efficiency and accuracy of data processing. This process reduces manual entry errors and saves a lot of human resources, allowing the project management team to focus more on decision-making and analysis;

[0245] By building a comprehensive knowledge graph that includes multiple data types such as text, images, and videos, we have achieved systematic and formal management of knowledge. This integration has promoted the accumulation, transfer, and sharing of knowledge, improved the efficiency of knowledge utilization, and provided rich information support for project decision-making.

[0246] By building and training machine learning-based models, we can simulate and predict key aspects of water transport engineering projects, including project planning, risk assessment, and resource allocation. The model building and verification process ensures the accuracy of the output results and provides a powerful decision support tool for project management.

[0247] The output of all models is coordinated through the decision support module, which optimizes the project management process. Through the automated decision-making process, the project planning cycle is shortened, the timeliness and effectiveness of risk management and the rationality of resource allocation are improved, and the efficiency and quality of project management are improved overall.

[0248] Project data and user feedback are continuously collected through the feedback mechanism to optimize the model and knowledge graph. The iterative update mechanism enables the system to adapt to changing project requirements and complex engineering environments, and continuously improve system performance and user satisfaction.

[0249] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Agent-based water transport project management decision support system, It is characterized in that The system includes: A data layer, the data layer comprising: Data collection unit, using OCR to automatically identify and extract water transport engineering data from various documents; A data storage unit for storing the collected data, wherein the data storage unit needs to ensure the reliability and integrity of the data; The data preprocessing unit is used to preprocess the data collected by the data collection unit, including data cleaning and format conversion, to ensure data quality and the efficiency of subsequent processing; The knowledge management layer includes: The knowledge graph construction unit is used to integrate multiple data types such as text, images, and videos in the field of water transport engineering to build a knowledge base; Update and maintenance unit, used to regularly update and maintain the knowledge base to reflect the latest industry standards, engineering cases and scientific research results. Maintenance includes the addition, update and abandonment of knowledge. An intelligent processing layer, the intelligent processing layer comprising: An intelligent model unit, used to establish a model that simulates the decision-making process for project planning, risk assessment, and resource allocation; The model training unit is used to train the model to ensure accurate output results.

2. A method for assisting decision-making in water transport project management based on an intelligent agent, the method adopting the system of claim 1, It is characterized in that The steps include: S1. Extract data from water transport engineering documents, drawings and reports using optical character recognition; S2. Preprocess the extracted data, including data cleaning and format conversion, to improve data quality and processing efficiency; S3. Build and maintain a knowledge graph containing knowledge in the field of water transport engineering, which integrates structured and unstructured data and is regularly updated to reflect the latest industry specifications, engineering cases and scientific research results; S4. Build and train machine learning-based models to simulate and predict planning, risk assessment, and resource allocation for water transport engineering projects; S5. Coordinate the outputs of all models through the decision support module and output decision support; S6. Establish a feedback mechanism to collect feedback and project results data to continuously optimize the model and knowledge graph.

3. According to claim 2, the agent-based water transport engineering project management decision-making support method, It is characterized in that Step S1 is specifically as follows: S11. Convert all physical documents into digital images through scanners; S12, determine the layout of text, tables and graphics, and adjust the clarity, noise removal and brightness of the image; S13, converting the graphic text in the image into an editable text format; S14, correcting errors that occur during recognition; S15. Integrate the identified data with data collected from existing databases, real-time data streams, and other data sources to form extracted data.

4. According to claim 3, the agent-based water transport engineering project management decision-making support method, It is characterized in that In step S14, the error in correction identification is specifically: S141, segment the given sentence, and the segmentation result will serve as the basis for the next step; S142, finding a confusion set of corresponding words in the sentence, wherein the confusion set is a pre-prepared set consisting of common confusing characters, including similar pronunciations and similar glyphs; S143, for each replaceable character in the sentence, enumerate each character in its confusion set to replace the original character, and obtain a candidate sentence set; S144, calculating the score of each candidate sentence, and the sentence with the highest score will be selected as the final output; The score calculation formula for calculating each candidate sentence in step S144 is as follows: ； Among them, the conditional probability of each item can be calculated by maximum likelihood estimation, and the expression is: ； In the formula, N(c l−1 ,c l ) is a binary group (c l−1 ,c l ) appears in the data, N(c l−1 ) is the character c l−1 The number of occurrences; the above formula is calculated through the following process: Count each tuple (c l−1 ,c l ) appears at a frequency N(c l−1 ,c l ) and a single character c l−1 The frequency of occurrence N(c l−1 ); According to the above formula, the conditional probability P(c l ∣c l−1 ), the probability represents that given the previous character c l−1 In the case of l Probability of occurrence; Multiply all the conditional probabilities together to get the probability of the entire string, and determine the final output based on the probability of the entire string.

5. According to claim 2, the agent-based water transport engineering project management decision-making support method, It is characterized in that Step S3 of constructing and maintaining a knowledge graph containing knowledge in the field of water transport engineering is as follows: S31. Integrate structured data, semi-structured data and unstructured data in the field of water transport engineering, including industry specifications, engineering cases and scientific research results. The integrated data is stored in the form of RDF or graph database; S32. Construct a knowledge graph ontology, wherein the knowledge graph ontology needs to clarify concepts and entities within the domain and determine the logical relationships between them; S33. Design a conceptual hierarchy, build a hierarchical relationship between concepts, and ensure the accuracy and hierarchy of each concept; S34, identifying named entities specific to the water transport engineering field, such as project names, technical terms, and locations, from the raw data, and extracting relationships between entities based on the identified entity and relationship patterns; Among them, the basic unit of the knowledge graph is a triple consisting of entity-relationship-entity, which is used to present complex information in a simple and clear manner.

6. The agent-based decision-making support method for water transport engineering project management according to claim 2, It is characterized in that The establishment and training of a model based on machine learning described in step S4 specifically adopts a regression model or a classification model. If a regression model is used to predict a specific value, a classification model is used if a classification problem includes risk level and feasibility. The training and verification of the model specifically includes: Data segmentation, dividing the dataset into training and test sets, or using cross-validation methods to evaluate the performance of the model on unseen data; Model training, using training set data to train the model, and adjusting model parameters through optimization algorithms to minimize prediction errors; Model validation, evaluating model performance through validation sets or cross-validation to ensure that the model has good generalization ability; The step S4 also includes model evaluation and optimization, specifically: Performance indicators, use accuracy, recall, and F1 score indicators to evaluate the performance of classification models, and for regression models, use the mean square error indicator; Hyperparameter tuning optimizes model performance by adjusting the number of trees in a random forest or the depth of a decision tree in a machine learning algorithm.

7. The agent-based decision-making support method for water transport engineering project management according to claim 2, It is characterized in that Step S5 is specifically as follows: S51. According to the specific requirements of the project and the risk tolerance, set the decision threshold and establish a rule engine to automatically make decisions according to the set rules; S52. Collect data from each model, including time prediction data, cost prediction data, and risk assessment data; S53, cleaning the collected data, including removing outliers and processing missing values, and then normalizing the data to convert it into a unified dimension; S54, assigning different weights to each model, and integrating the outputs of each model using a weighted average formula to obtain a comprehensive prediction result; S55. Define the objective function of decision optimization, choose to minimize total cost or maximize benefits, and establish constraints based on the actual situation of the project; S56. Based on the constraints and the specifically defined decision optimization objective function, output a detailed report of the decision, including charts, data analysis results and specific recommendations.

8. The agent-based decision-making support method for water transport engineering project management according to claim 7, It is characterized in that In step S54, the weight distribution expression is: ； Where C is the comprehensive cost forecast, C 1 , C 2 , C 3、 C n are the output results of the model, w 1 、w 2、 w 3 and w n is the weight of the one-to-one corresponding model output result; The weight is calculated by calculating the performance ratio. For n models, the accuracy of each model i is p i , the performance ratio r of each model i , the accuracy of each model j is p j， The expression is: ； Since the weight w of each model i The performance ratio can be used directly, and the sum of all weights is equal to 1, which is expressed as: ； Since the sum of all weights equals 1, the weight expression for each model is: ； In step S56, the constraint conditions and the specifically defined decision optimization objective function are expressed as follows: ； Where 0≤x i ≤B i , i=1,2,...,n; In the above formula, x i is the investment amount of the i-th project, B is the total investment, and B i is the maximum investment of the ith project, which includes resource volume, production capacity, and risk tolerance.

9. The agent-based decision-making support method for water transport engineering project management according to claim 2, It is characterized in that Step S6 is specifically as follows: S61. Clarify the purpose of collecting feedback, including improving model performance, increasing user satisfaction, or optimizing knowledge graph content. Based on the purpose, determine key performance indicators to measure the effect of feedback on system improvement. S62. Choose the appropriate feedback collection method based on the target audience; S63. Clean and classify the collected data to facilitate subsequent analysis and processing, and start a feedback mechanism to actively encourage users to provide feedback; S64. Identify commonalities or important trends from the feedback through qualitative analysis and rank them according to their importance and urgency to determine priorities for processing; S65. Based on the collected feedback, formulate a specific action plan to improve the knowledge graph and actually apply the improvement measures to the system.

Images