Intelligent analysis and natural language report generation system and method for traffic simulation results

Abstract

The intelligent analysis and natural language report generation system and method of traffic simulation results belong to the field of intelligent traffic. In order to solve the problem of automatic and intelligent deep analysis of traffic simulation results. The user interaction layer, the data access layer, the intelligent processing layer, the report generation layer and the output layer are connected in sequence, and the output layer is connected to the user interaction layer again; the user interaction layer includes a Web graphical user interface, a mobile terminal application module and a voice / text input interface; the data access layer includes a simulation data receiving module, a project configuration management module and an evaluation type and template selection module; the intelligent processing layer includes an intelligent analysis engine and a natural language generation engine, the intelligent analysis engine includes a key indicator identification unit, an influence degree evaluation unit and a correlation problem positioning unit, and the natural language generation engine includes a domain knowledge base, a text generation unit, a conclusion generation unit and a suggestion generation unit. The present application realizes the whole process automation from simulation data to professional analysis report.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of traffic simulation technology and artificial intelligence natural language processing, specifically involving a system and method for intelligent analysis of traffic simulation results and generation of natural language reports. Background Technology

[0002] Traffic impact assessment is a crucial component of urban traffic planning and management. Traditional methods rely on specialized simulation software (such as TransCAD and Vissim) for traffic simulation analysis. After the simulation is completed, professionals need to spend a significant amount of time manually processing the data, creating charts, and writing analysis reports. This process suffers from the following problems: high reliance on manual labor: engineers with specialized knowledge are required to interpret the results and write the report; low efficiency: it typically takes several hours or even days from simulation completion to report output; poor consistency: reports written by different personnel have inconsistent formats and levels of detail; high interpretation threshold: the raw simulation data is difficult for non-professionals to understand. Summary of the Invention

[0003] The technical problem to be solved by this invention is to realize intelligent in-depth analysis of traffic simulation results and automated and standardized report generation. It proposes an intelligent analysis and natural language report generation system and method for traffic simulation results.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A traffic simulation result intelligent analysis and natural language report generation system includes a user interaction layer, a data access layer, an intelligent processing layer, a report generation layer, and an output layer, wherein the user interaction layer, data access layer, intelligent processing layer, report generation layer, and output layer are connected in sequence, and the output layer is then connected to the user interaction layer.

[0006] The user interaction layer includes a web graphical user interface, a mobile application module, and a voice / text input interface;

[0007] The data access layer includes a simulation data receiving module, a project configuration management module, and an evaluation type and template selection module;

[0008] The intelligent processing layer includes an intelligent analysis engine and a natural language generation engine. The intelligent analysis engine includes a key indicator identification unit, an impact assessment unit, and a related problem location unit. The natural language generation engine includes a domain knowledge base, a text generation unit, a conclusion generation unit, and a suggestion generation unit.

[0009] The report generation layer includes a multi-format exporter, a data population and rendering engine, and a template selector;

[0010] The output layer includes a web interactive dashboard, a PDF report generation module, and a Word report generation module.

[0011] Furthermore, the key indicator identification unit in the intelligent analysis engine adopts a multi-dimensional feature extraction algorithm to automatically identify key performance indicators from simulation data, including traffic capacity, service level, and delay time; the related problem location unit adopts a bottleneck identification model that integrates multiple indicators to identify traffic bottlenecks caused or aggravated by intervention measures; and the impact assessment unit establishes a multi-level assessment indicator system, combining quantitative calculation and qualitative analysis to conduct graded assessment of the impact.

[0012] Furthermore, the domain knowledge base in the natural language generation engine is connected to the text generation unit, conclusion generation unit, and suggestion generation unit respectively. The domain knowledge base is constructed through a combination of expert rule initialization and data-driven optimization, including professional analysis logic and expression standards in the field of traffic engineering. The text generation unit automatically generates professional text that conforms to traffic engineering standards based on template and rule-driven methods, combined with a deep learning language model. The conclusion generation unit uses information extraction and text summarization techniques to extract key information from multi-dimensional analysis results and generate conclusive statements. The suggestion generation unit uses an inference engine based on the domain knowledge base to generate management suggestions.

[0013] A method for intelligent analysis and natural language report generation system of traffic simulation results includes the following steps:

[0014] S1. The data receiving layer receives traffic simulation results data and then configures project information;

[0015] S2. Input the simulation data obtained in step S1 into the intelligent analysis engine in the intelligent processing layer to calculate the indicators, and then perform key problem location and impact assessment to obtain the structured data output by the intelligent analysis engine.

[0016] S3. Input the structured data output by the intelligent analysis engine obtained in step S2 into the natural language generation engine in the intelligent processing layer, and fill and synthesize it into natural language text that conforms to traffic engineering specifications based on the rules, templates and mapping relationships in the domain knowledge base.

[0017] S4. Input the natural language text obtained in step S3 into the report generation layer. Based on the preset template, integrate the text content, visualization charts and raw data to generate a draft of the traffic simulation result report.

[0018] S5. The output layer outputs a complete report in PDF, Word, and Web formats simultaneously, based on the initial draft of the traffic simulation results report.

[0019] S6. Input the complete report obtained in step S5 into the user interaction layer for result display, and the user can conduct interactive data exploration through the Web graphical user interface.

[0020] Furthermore, the specific implementation method of step S1 includes the following steps:

[0021] S1.1. The simulation results data are divided into output data of the basic scenario and output data of the intervention scenario. The intervention scenarios include traffic planning scenario, traffic control scenario, and demand management scenario.

[0022] S1.2. Configure project information for traffic simulation results data. The configuration projects include project name, major category of intervention scenario and corresponding specific measure category, specific measures, evaluation year, evaluation period, evaluation scope, and simulation dimension.

[0023] Furthermore, the specific implementation method of step S2 includes the following steps:

[0024] S2.1. Input the simulation data obtained in step S1 into the intelligent analysis engine in the intelligent processing layer for index calculation. The index calculation includes regional level index, road segment level index, and intersection level index.

[0025] Regional indicators are used to calculate the total travel demand of the entire network or a designated area, the total traffic flow of the area, the total vehicle kilometers of the area, the average speed of the area, the total congestion mileage of the area, and the congestion cost of the area.

[0026] The road segment level indicators are used to calculate the traffic flow, road segment saturation, average speed, road segment density, and road segment delay time for each road segment.

[0027] Intersection-level indicators calculate the node traffic, service level, overflow degree, and total node delay for each intersection;

[0028] S2.2. Construct a bottleneck identification model that integrates multiple indicators. Based on the regional, road segment, and intersection indicators obtained in step S1, calculate the comprehensive severity score and locate key issues.

[0029] First, calculate the saturation deterioration score S_vc. The calculation formula is as follows:

[0030]

[0031] Among them, VOC baseline VOCs intervention It represents the saturation level of a road segment or intersection under the basic and intervention schemes, where w_voc is the weight of the saturation index and max is the maximum value function.

[0032] The speed reduction score S_speed is calculated using the following formula:

[0033]

[0034] Among them, Speed baseline Speed intervention It refers to the speed of the basic plan and intervention plan for road segments or intersections, where w_speed is the weight of the speed index;

[0035] Define the downgrade score mapping table with grades from A to F, and calculate the service level downgrade score S_los using the following formula:

[0036]

[0037] Where LOS_Degrade_Score is the degrade score, and w_los is the weight of the service level indicator;

[0038] The overall severity score S_total is obtained by summing the scores of all indicators for an analysis unit. The calculation formula is as follows:

[0039] S_total = S_vc + S_speed + S_los

[0040] Set a comprehensive threshold T_bottleneck and a single veto threshold, then determine the key issues based on whether S_total > T_bottleneck or VOC. intervention > 0.95 or LOS intervention If the rating is F, then the analysis unit is marked as a critical bottleneck, where LOS intervention Service level rating for the intervention program;

[0041] S2.3. The overall impact of the intervention measures is assessed using the entropy weight-TOPSIS-based quantitative assessment model, and the impact levels are classified.

[0042] Furthermore, the specific implementation method of step S3 includes the following steps:

[0043] S3.1. Construct a domain knowledge base, including an intervention-assessment mapping library, a measure-suggestion rule library, and a traffic semantic vector library;

[0044] The data structure of the intervention-assessment mapping library is a lookup table with intervention category and specific measures as keys. The core fields include a list of core assessment dimensions, an assessment indicator weight vector, indicator-conclusion semantic mapping rules, and conclusion template ID.

[0045] The data structure of the action-recommendation rule base is a set of rules with problem type, intervention type, and severity level as keys. Core fields include a list of recommended actions, triggering conditions, and a priority score P, where P is calculated using the following formula:

[0046]

[0047] Where EffectScore is the effect score and CostScore is the cost score, which are predefined by expert experience, and α and β are the adjustment coefficients for the effect score and cost score, respectively;

[0048] The data structure of the traffic semantic vector library is a finely tuned professional word embedding model or lookup table. The core content includes degree adverb mapping relationships, state adjective mapping relationships, and causal association lexicon.

[0049] The domain knowledge base is constructed and iterated through a combination of hard-coded expert rules and dynamic optimization using machine learning.

[0050] S3.2. For the structured data output by the intelligent analysis engine obtained in step S2, obtain the corresponding core assessment dimension list and weight vector W from the intervention-assessment mapping library, and transform the numerical changes into qualitative descriptions according to the indicator-conclusion semantic mapping rules.

[0051] Then, the problem type is searched in the measure-suggestion rule base, and all the retrieved suggested measures are sorted according to their priority score P, and the top-N measures are output;

[0052] Then, the natural language generation engine uses a traffic semantic vector library to assemble qualitative descriptions and suggested measures into natural language text that conforms to traffic engineering specifications.

[0053] Furthermore, the specific implementation method of the domain knowledge base in step S3 includes the following steps:

[0054] For feedback-based learning-based weight optimization, the method of hard-coding expert rules is used. For the evaluation index weight vector, the gradient descent method is used for adjustment, and the calculation formula is as follows:

[0055]

[0056] in, The original weights of the evaluation index m to be adjusted. η is the adjusted weight of the metric m to be evaluated, η is the learning rate, and FeedbackStrength is the feedback strength, which is quantified by user behavior.

[0057] The domain knowledge base, which is hard-coded by expert rules, is dynamically optimized and expanded using machine learning. By using association rule mining or clustering methods to analyze historical simulation projects and their final adopted manual solutions, each historical project is first represented as a feature vector. Then, the Apriori algorithm is used to discover frequent itemsets. Finally, new rules with confidence scores greater than a preset threshold are added to the measure-suggestion rule base.

[0058] Furthermore, the interactive data verification and exploration method in step S6 involves the user exploring the data based on the web interactive dashboard after receiving the web report.

[0059] The beneficial effects of this invention are:

[0060] This invention discloses an intelligent analysis and natural language report generation system for traffic simulation results. By integrating an intelligent analysis engine and a natural language generation engine, it achieves automatic parsing, multi-dimensional professional analysis, key issue identification, and impact assessment of multi-source traffic simulation data. Ultimately, it intelligently generates standardized reports in three formats: PDF, Word, and a web-based interactive dashboard, containing data charts, professional conclusions, and targeted management recommendations. This invention reduces the traditional report generation process, which relies on manual expert work and takes hours or even days, to minutes, significantly improving efficiency and lowering the barrier to entry for traffic simulation technology.

[0061] The present invention discloses an intelligent analysis and natural language report generation system for traffic simulation results, which realizes automated and intelligent in-depth analysis of traffic simulation results. It solves the problems of traditional methods that rely on traffic engineers to manually interpret data, which is inefficient and easily affected by subjective experience. It allows machines to replace engineers in completing the transformation from raw data to structured knowledge.

[0062] The intelligent analysis and natural language report generation system for traffic simulation results described in this invention achieves seamless integration of the entire process, solving the problem of disconnect between simulation result analysis and final report generation, which previously required manual sorting, conversion, and writing. It realizes an end-to-end automated system from receiving simulation data to outputting a complete report.

[0063] The intelligent analysis and natural language report generation system for traffic simulation results described in this invention makes the simulation results easy for non-professionals to understand, solving the problems of the original simulation data and traditional reports being highly professional and difficult to understand. Through natural language and visualization technology, the complex analysis results are transformed into intuitive and easy-to-understand narrative text and management suggestions.

[0064] The intelligent analysis and natural language report generation system for traffic simulation results described in this invention meets diverse output needs, improves the practicality and coverage of reports, and solves the problem of traditional reports having a single format and being difficult to adapt to different application scenarios (such as formal archiving, scheme modification, and online reporting). It provides a flexible output mechanism that can generate reports in multiple formats simultaneously. Attached Figure Description

[0065] Figure 1 This is a schematic diagram of the structure of an intelligent analysis and natural language report generation system for traffic simulation results according to the present invention;

[0066] Figure 2 This is a structural block diagram of an intelligent analysis and natural language report generation system for traffic simulation results as described in this invention;

[0067] Figure 3 This is a flowchart of the method for intelligent analysis and natural language report generation system of traffic simulation results according to the present invention. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.

[0069] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention.

[0070] To further understand the invention's content, features, and effects, the following specific embodiments are provided, along with accompanying drawings. Figure 1 - Appendix Figure 3 Detailed explanation is as follows:

[0071] Example 1:

[0072] This embodiment provides an intelligent analysis and natural language report generation system for traffic simulation results. The system aims to achieve in-depth mining and problem diagnosis of simulation data through an intelligent analysis engine, and automatically convert the diagnostic results into structured professional reports through a natural language generation engine.

[0073] A traffic simulation result intelligent analysis and natural language report generation system includes a user interaction layer, a data access layer, an intelligent processing layer, a report generation layer, and an output layer, wherein the user interaction layer, data access layer, intelligent processing layer, report generation layer, and output layer are connected in sequence, and the output layer is then connected to the user interaction layer.

[0074] The user interaction layer includes a web graphical user interface, a mobile application module, and a voice / text input interface;

[0075] The data access layer includes a simulation data receiving module, a project configuration management module, and an evaluation type and template selection module;

[0076] The intelligent processing layer includes an intelligent analysis engine and a natural language generation engine. The intelligent analysis engine includes a key indicator identification unit, an impact assessment unit, and a related problem location unit. The natural language generation engine includes a domain knowledge base, a text generation unit, a conclusion generation unit, and a suggestion generation unit.

[0077] The report generation layer includes a multi-format exporter, a data population and rendering engine, and a template selector;

[0078] The output layer includes a web interactive dashboard, a PDF report generation module, and a Word report generation module.

[0079] Furthermore, the key indicator identification unit in the intelligent analysis engine adopts a multi-dimensional feature extraction algorithm to automatically identify key performance indicators from simulation data, including traffic capacity, service level, and delay time; the related problem location unit adopts a bottleneck identification model that integrates multiple indicators to identify traffic bottlenecks caused or aggravated by intervention measures; and the impact assessment unit establishes a multi-level assessment indicator system, combining quantitative calculation and qualitative analysis to conduct graded assessment of the impact.

[0080] Furthermore, the domain knowledge base in the natural language generation engine is connected to the text generation unit, conclusion generation unit, and suggestion generation unit respectively. The domain knowledge base is constructed through a combination of expert rule initialization and data-driven optimization, including professional analysis logic and expression standards in the field of traffic engineering. The text generation unit automatically generates professional text that conforms to traffic engineering standards based on template and rule-driven methods, combined with a deep learning language model. The conclusion generation unit uses information extraction and text summarization techniques to extract key information from multi-dimensional analysis results and generate conclusive statements. The suggestion generation unit uses an inference engine based on the domain knowledge base to generate management suggestions.

[0081] Furthermore, the Web graphical user interface (GUI) provides a complete web-based operating interface, supporting full project lifecycle management. This module adopts a responsive design and includes four main functional areas: a project configuration panel, a data upload interface, real-time progress monitoring, and result visualization. Users can complete complex configurations through intuitive drag-and-drop operations. The system verifies input validity in real time and provides intelligent prompts, ensuring smooth workflow and data accuracy.

[0082] The mobile application provides convenient mobile office support for field staff. Optimized for mobile devices, this module includes three main functions: simplified operation processes, real-time push notifications, and mobile report viewing. It supports offline data collection and synchronization, automatically uploading data when the network is restored, ensuring continuity of field work and timeliness of data.

[0083] The voice / text input interface enables a natural language-based human-computer interaction experience. This module integrates advanced speech recognition and natural language understanding technologies, supporting multi-turn dialogue and contextual understanding. Users can describe their traffic needs via voice or text, and the system automatically parses the user's intent and generates corresponding configuration parameters, significantly reducing the barrier to entry for system use.

[0084] The simulation data receiving module is designed to process multi-source simulation data input in a unified manner. This module supports data format parsing from mainstream traffic simulation software, including TransCAD, VISSIM, and AIMSUN. It automatically detects outliers through data quality verification algorithms and provides data cleaning and transformation functions to ensure the integrity and comparability of data from the base scenario and the intervention scenario.

[0085] The project configuration management module manages basic project information and simulation parameter configurations. This module adopts a layered configuration architecture, supporting automatic adaptation of different parameter templates based on project type (planning / control). It includes a built-in intervention measure classification library covering typical scenarios such as road segment control, intelligent traffic control, and integrated traffic organization, and supports flexible switching between simulation dimensions (macro / micro).

[0086] The assessment type and template selection module provides professional assessment frameworks and output templates. This module integrates multiple standard assessment systems, including traffic impact assessment, environmental assessment, and economic benefit assessment. It intelligently recommends appropriate assessment types and report templates based on project characteristics, and supports online editing and custom expansion of templates.

[0087] The key performance indicator (KPI) identification unit is designed to automatically extract and identify core performance indicators. This unit employs a multi-dimensional feature extraction algorithm to automatically identify KPIs from massive amounts of simulation data, including throughput, service level, and delay time.

[0088] The key issue location unit is designed to automatically and accurately identify new or exacerbated traffic bottlenecks caused by intervention measures from a massive number of road segments and nodes using a multi-indicator fusion bottleneck identification model.

[0089] The function of the impact assessment unit is to quantitatively evaluate the degree of impact of traffic interventions. This unit establishes a multi-level assessment index system, combining quantitative calculations and qualitative analysis to classify and assess the degree of impact. Fuzzy comprehensive evaluation is used to handle uncertainties, outputting standardized impact levels and risk warnings.

[0090] The text generation unit converts numerical results into professional text descriptions. Based on template and rule-driven methods, combined with a deep learning language model, this unit automatically generates professional text that conforms to traffic engineering specifications. It supports multilingual output and text generation with varying levels of detail to meet the reading needs of different users.

[0091] The conclusion generation unit is designed to extract core findings from the analysis and generate summary conclusions. This unit employs information extraction and text summarization techniques to extract key information from multi-dimensional analysis results, generating accurate and concise conclusive statements. A built-in confidence assessment mechanism ensures the reliability and scientific rigor of the conclusions.

[0092] The suggestion generation unit's function is to generate targeted management recommendations based on the analysis results. This unit combines a domain knowledge base and a case library, using an inference engine to generate specific and feasible management recommendations. The recommendations cover multiple aspects such as traffic organization optimization, adjustment of control measures, and infrastructure improvement, and include priority and implementation difficulty assessments.

[0093] The template selector intelligently matches and selects report templates. Based on project type, assessment needs, and user preferences, this module automatically selects the most suitable report template from the template library. It supports intelligent template recommendations and manual adjustments to ensure a high degree of match between the report structure and content requirements.

[0094] The data population and rendering engine dynamically populates analysis content into report templates. This engine employs a component-based design, supporting intelligent layout and rendering of different content types, including text, charts, and tables. An adaptive layout algorithm ensures optimal display across various screen sizes, providing real-time preview and adjustment capabilities.

[0095] The multi-format exporter enables simultaneous output of reports in multiple formats. This module integrates various document processing engines, supporting one-click generation of PDF, Word, HTML, and other formats. Unified style management ensures content consistency across different formats, and it provides advanced features such as batch processing and scheduled generation.

[0096] The PDF report generation module is designed to produce professional-looking PDF reports. It employs a high-quality typesetting engine to ensure both professionalism and aesthetic appeal. It supports accurate rendering of complex charts, mathematical formulas, and technical symbols, and provides security features such as password protection and digital signatures to meet the archiving requirements of formal documents.

[0097] The Word report generation module generates editable Word document reports. This module maintains the integrity and editability of the document format, supporting subsequent manual modifications and collaborative editing. Standardized styles ensure a clear document structure, facilitating correct display across different versions of Word.

[0098] The web-based interactive dashboard provides an online, interactive data exploration platform. Built with modern web technologies, this responsive dashboard supports multi-dimensional data filtering, drill-down, and interconnected analysis. It integrates a rich set of visualization components, allowing users to customize analytical views through drag-and-drop operations for in-depth data exploration and insightful discovery.

[0099] The technical implementation details of this embodiment include: natural language generation using a combination of rule-based templates and machine learning; report templates using the Jinja2 template engine, supporting dynamic content generation; multi-format output implemented through their respective rendering engines (WeasyPrint for PDF, python-docx for Word, Jinja2 for HTML); and visualization charts generated based on the Plotly library, supporting interactive functions.

[0100] Example 2:

[0101] A method for intelligent analysis and natural language report generation system of traffic simulation results as described in Embodiment 1 includes the following steps:

[0102] S1. The data receiving layer receives traffic simulation results data and then configures project information;

[0103] Furthermore, the specific implementation method of step S1 includes the following steps:

[0104] S1.1. The simulation results data are divided into output data of the basic scenario and output data of the intervention scenario. The intervention scenarios include traffic planning scenario, traffic control scenario, and demand management scenario.

[0105] S1.2. Configure project information for traffic simulation results data. The configuration projects include project name, major category of intervention scenario and corresponding specific measure category, specific measures, evaluation year, evaluation period, evaluation scope, and simulation dimension.

[0106] Project configuration information is shown in Table 1:

[0107] Table 1

[0108]

[0109] The specific measures for intervention projects are categorized as follows: traffic planning (Table 2), traffic control (Table 3), and demand management (Table 4).

[0110] Table 2

[0111]

[0112] Table 3

[0113]

[0114] Table 4

[0115]

[0116] S2. Input the simulation data obtained in step S1 into the intelligent analysis engine in the intelligent processing layer to calculate the indicators, and then perform key problem location and impact assessment to obtain the structured data output by the intelligent analysis engine.

[0117] Furthermore, the specific implementation method of step S2 includes the following steps:

[0118] S2.1. Input the simulation data obtained in step S1 into the intelligent analysis engine in the intelligent processing layer for index calculation. The index calculation includes regional level index, road segment level index, and intersection level index.

[0119] Regional indicators are used to calculate the total travel demand of the entire network or a designated area, the total traffic flow of the area, the total vehicle kilometers of the area, the average speed of the area, the total congestion mileage of the area, and the congestion cost of the area.

[0120] Furthermore, the formula for calculating Total Travel Demand (TVD) is as follows:

[0121]

[0122] Where, d ij Let I be the travel demand from starting point i to ending point j, where I is the set of all starting points in the region, and J is the set of all starting points in the region.

[0123] The formula for calculating Total Flow Rate (TVF) is:

[0124]

[0125] Among them, flow ij Let i be the travel flow from origin i to destination j;

[0126] The formula for calculating the total vehicle mileage (TVK) for a given area is:

[0127]

[0128] in, Let the traffic flow be for road segment a. Let A be the length of road segment 'a', and A be the set of road segments within the region.

[0129] Regional average speed The calculation formula is:

[0130]

[0131] in, The travel time for route segment a;

[0132] The formula for calculating the total congestion mileage (CL) is:

[0133]

[0134] Where I is the indicator function, when the road segment speed V a When the value is below the congestion threshold, the value is 1.

[0135] The formula for calculating congestion cost (CC) is:

[0136]

[0137] in, Free-flow time, VOT stands for Time Value;

[0138] The road segment level indicators are used to calculate the traffic flow, road segment saturation, average speed, road segment density, and road segment delay time for each road segment.

[0139] Furthermore, the traffic flow on the road segment is Flow. a The formula for calculating road segment saturation is: VOC a =Flow a / Capacity a ;

[0140] The formula for calculating the average speed of a road segment is: Speed a =L a / T a ;

[0141] The formula for calculating road segment density is: Density a= Flow a / Speed a ;

[0142] The formula for calculating road segment delay time is: ;

[0143] Intersection-level indicators calculate the node traffic, service level, overflow degree, and total node delay for each intersection;

[0144] Furthermore, the formula for calculating the node traffic at each intersection is as follows:

[0145]

[0146] Among them, I n For all inlet channels of node n;

[0147] The formula for calculating total node delay is:

[0148] ;

[0149] The formula for calculating the overflow level is:

[0150]

[0151] Among them, overflow_index n Here, `in` represents the overflow level at the intersection, and `overflow_index` represents all the approach lanes of node `n`. i Lqueue represents the overflow level value of the inlet lane i at intersection n. i Lstorage is the queue length for the import lane. i Lbuffer represents the physical storage length of the imported channel i. i This is the safety buffer value for the import lane i;

[0152] The formula for calculating service level is:

[0153]

[0154] According to the "Road Capacity Manual" standard, delays are mapped to AF levels; as shown in Table 5:

[0155] Table 5

[0156]

[0157] S2.2. Construct a bottleneck identification model that integrates multiple indicators. Based on the regional, road segment, and intersection indicators obtained in step S1, calculate the comprehensive severity score and locate key issues.

[0158] First, calculate the saturation deterioration score S_vc. The calculation formula is as follows:

[0159]

[0160] Among them, VOC baseline VOCs interventionIt represents the saturation level of a road segment or intersection under the basic and intervention schemes, where w_voc is the weight of the saturation index and max is the maximum value function.

[0161] The speed reduction score S_speed is calculated using the following formula:

[0162]

[0163] Among them, Speed baseline Speed intervention It refers to the speed of the basic plan and intervention plan for road segments or intersections, where w_speed is the weight of the speed index;

[0164] Define the downgrade score mapping table with grades from A to F, and calculate the service level downgrade score S_los using the following formula:

[0165]

[0166] Where LOS_Degrade_Score is the degrade score, and w_los is the weight of the service level indicator;

[0167] The overall severity score S_total is obtained by summing the scores of all indicators for an analysis unit. The calculation formula is as follows:

[0168] S_total = S_vc + S_speed + S_los

[0169] Set a comprehensive threshold T_bottleneck and a single veto threshold, then determine the key issues based on whether S_total > T_bottleneck or VOC. intervention >0.95 or LOS intervention If the rating is F, then the analysis unit is marked as a critical bottleneck, where LOS intervention Service level rating for the intervention program;

[0170] S2.3. The overall impact of the intervention measures is assessed using the entropy weight-TOPSIS-based quantitative assessment model, and the impact levels are classified.

[0171] Furthermore, the specific implementation method of the traffic intervention impact quantification assessment model based on entropy weight-TOPSIS includes the following steps:

[0172] S2.3.1. Construct an evaluation matrix by selecting m evaluation indicators (such as total vehicle hours in the region, average speed, average saturation of main roads, number of severely congested road sections, etc.) to form a 2×m matrix under the basic scenario and the intervention scenario.

[0173] S2.3.2. Calculate entropy weights. The entropy weight method is an objective weighting method that determines the weight based on the degree of variation in the indicator data. The greater the degree of variation (i.e., the greater the change in the indicator before and after the scene), the higher its weight.

[0174] Calculate the entropy value e of the j-th index. j :

[0175]

[0176] Where, p ij It represents the proportion of the indicator value after standardization, where k is a constant.

[0177] Calculate the weight w of the j-th indicator j :

[0178]

[0179] S2.3.3. TOPSIS ranking: Calculate the relative proximity of the intervention scenario to the scenario with the best overall performance across all metrics and the scenario with the worst overall performance across all metrics, and calculate the relative proximity C. i :

[0180]

[0181] in, and These are the distances from the intervention scenario to the positive and negative ideal solutions, respectively; C i The closer the value is to 1, the better the effect of the intervention; the closer it is to 0, the greater the negative impact.

[0182] S2.3.4. Classify the impact level based on the calculated C. i Values ​​are set to categorize and classify the degree of influence.

[0183] C i <0.3: Impact level = severe (negative); 0.3 <= C i <0.6: Impact level = Moderate (Negative); 0.6 <= C i <0.8: Impact level = Slight (Negative / Neutral); C i >=0.8: Impact level = positive (positive).

[0184] S3. Input the structured data output by the intelligent analysis engine obtained in step S2 into the natural language generation engine in the intelligent processing layer, and fill and synthesize it into natural language text that conforms to traffic engineering specifications based on the rules, templates and mapping relationships in the domain knowledge base.

[0185] Furthermore, the specific implementation method of step S3 includes the following steps:

[0186] S3.1. Construct a domain knowledge base, including an intervention-assessment mapping library, a measure-suggestion rule library, and a traffic semantic vector library;

[0187] The data structure of the intervention-assessment mapping library is a lookup table with intervention category and specific measures as keys. The core fields include a list of core assessment dimensions, an assessment indicator weight vector, indicator-conclusion semantic mapping rules, and conclusion template ID.

[0188] Furthermore, the core evaluation dimension list defines the set of indicators that require focused attention (e.g., {saturation, average delay, queue length}). The evaluation indicator weight vector W uses the analytic hierarchy process (AHP) or entropy weighting to calculate the importance weights of each indicator. For example, for "signal optimization," delay has a greater weight than traffic flow; for "new road construction," diversion rate has a greater weight than queue length. The indicator-conclusion semantic mapping rule associates indicator changes with qualitative descriptions. The conclusion template ID is associated with a pre-defined conclusion generation template.

[0189] The data structure of the action-recommendation rule base is a set of rules with problem type, intervention type, and severity level as keys. Core fields include a list of recommended actions, triggering conditions, and a priority score P, where P is calculated using the following formula:

[0190]

[0191] Where EffectScore is the effect score and CostScore is the cost score, which are predefined by expert experience, and α and β are the adjustment coefficients for the effect score and cost score, respectively;

[0192] Furthermore, the list of recommended measures consists of a series of specific action items; the triggering condition is the logical condition for implementing the recommendation; the priority score is the priority calculated based on the difficulty of implementation and the expected effect; and the preconditions are the prerequisites required to implement the recommendation (such as "recommended only on non-main roads").

[0193] The data structure of the traffic semantic vector library is a finely tuned professional word embedding model or lookup table. The core content includes degree adverb mapping relationships, state adjective mapping relationships, and causal association lexicon.

[0194] The domain knowledge base is constructed and iterated through a combination of hard-coded expert rules and dynamic optimization using machine learning.

[0195] Furthermore, the degree adverb mapping includes the rate of change Δ∈[0,10%) → slightly, mildly; the rate of change Δ∈[10%,30%) → significant, obvious; the rate of change Δ∈[30%,∞) → drastic, greatly;

[0196] The state adjective mapping includes saturation V / C∈[0,0.6)→unobstructed flow; saturation V / C∈[0.6,0.8)→stable flow, basically stable; saturation V / C∈[0.8,1.0)→unstable flow, close to saturation; saturation V / C≥1.0→oversaturated, congested.

[0197] The causal association thesaurus includes "lead to", "cause", "the main reason lies in", and "further lead to".

[0198] S3.2. For the structured data output by the intelligent analysis engine obtained in step S2, obtain the corresponding core assessment dimension list and weight vector W from the intervention-assessment mapping library, and transform the numerical changes into qualitative descriptions according to the indicator-conclusion semantic mapping rules.

[0199] Then, the problem type is searched in the measure-suggestion rule base, and all the retrieved suggested measures are sorted according to their priority score P, and the top-N measures are output;

[0200] Then, the natural language generation engine uses a traffic semantic vector library to assemble qualitative descriptions and suggested measures into natural language text that conforms to traffic engineering specifications.

[0201] Furthermore, the specific implementation method of the domain knowledge base in step S3 includes the following steps:

[0202] For feedback-based learning-based weight optimization, the method of hard-coding expert rules is used. For the evaluation index weight vector, the gradient descent method is used for adjustment, and the calculation formula is as follows:

[0203]

[0204] in, The original weights of the evaluation index m to be adjusted. η is the adjusted weight of the metric m to be evaluated, η is the learning rate, and FeedbackStrength is the feedback strength, which is quantified by user behavior.

[0205] The domain knowledge base, which is hard-coded by expert rules, is dynamically optimized and expanded using machine learning. By using association rule mining or clustering methods to analyze historical simulation projects and their final adopted manual solutions, each historical project is first represented as a feature vector. Then, the Apriori algorithm is used to discover frequent itemsets. Finally, new rules with confidence scores greater than a preset threshold are added to the measure-suggestion rule base.

[0206] Furthermore, the system analyzes a large number of historical simulation projects and their ultimately adopted manual solutions to discover new and effective "problem-response" pairs. Association rule mining (such as the Apriori algorithm) or cluster analysis is used. First, each historical project is represented as a feature vector (problem features, adopted measures). Then, the Apriori algorithm is used to discover frequent itemsets, such as: {upstream congestion, overflow, associated intersections} → {optimize the phase difference of associated intersection signals} (confidence: 85%). Finally, new rules with high confidence (confidence > preset threshold) are added to the measure-suggestion rule base.

[0207] Furthermore, the knowledge base access process involves collaborative invocation of the hybrid inference engine;

[0208] During system operation, the aforementioned knowledge base is invoked collaboratively by a hybrid inference engine.

[0209] Input: "structured knowledge" generated by the intelligent analysis engine (e.g., {intersection X, average delay, baseline: 45s, intervention value: 110s}).

[0210] Diagnostic Phase: Based on the intervention type, the engine retrieves the corresponding core assessment dimension list and weight vector W from the intervention-assessment mapping library. According to the indicator-conclusion semantic mapping rules, the numerical changes are transformed into qualitative descriptions (e.g., "Delay time has increased significantly, far exceeding the acceptable range").

[0211] Prescription Phase: The engine uses the problem type = "intersection delay surge", intervention type = "road closure", and severity level = "severe" as keys to search the action-suggestion rule base. All retrieved suggested actions are sorted according to their priority score P, and the top-N suggestions are output.

[0212] Text generation stage: The natural language generation engine uses a traffic semantic vector library to assemble "qualitative descriptions" and "suggested measures" into fluent natural language paragraphs that conform to the context of traffic engineering.

[0213] The generation of descriptive text, summary conclusions, and management recommendations is based on rules, templates, and mapping relationships in the "domain knowledge base." It intelligently fills in and synthesizes the structured data output by the "intelligent analysis engine" into natural language text that conforms to traffic engineering specifications. The specific generation process is as follows:

[0214] The workflow of the text generation unit:

[0215] Input: A data point output by the intelligent analysis engine, such as {Object: Road segment A, Indicator: Saturation, Base value: 0.75, Intervention value: 0.95, Change rate: +26.7%}.

[0216] Processing: Query the "Traffic Semantic Vector Library" and "Intervention-Evaluation Mapping Library" in the domain knowledge base. Based on the rate of change: +26.7%, the degree adverb "significant" is matched. Based on the intervention value: 0.95, the state adjective "approaching saturation" is matched. According to the rules, the sentence template is formed: "[Object]'s traffic pressure [degree adverb] increases, saturation rises from [base value] to [intervention value] (at [state adjective]), the increase reaches [rate of change], and it has become one of the main congestion bottlenecks."

[0217] Output: The completed natural language sentence: "Traffic pressure on road segment A has increased significantly, with the saturation level rising from 0.75 to 0.95 (close to saturation), an increase of 26.7%, making it one of the main congestion bottlenecks."

[0218] The workflow of the conclusion generation unit:

[0219] Input: The collection of all "critical issue location" results (e.g., 10 road segments are severely congested, and the total regional delay increases by 15%).

[0220] Processing: Query the pre-defined comprehensive conclusion logic rules in the "Intervention-Assessment Mapping Library" of the domain knowledge base. Execute the rule: IF (Percentage of severely congested road sections > 10%) AND (Increase in total regional delays > 5%) THEN Conclusion Tone = "Significant Negative Impact". Call the conclusion template corresponding to this conclusion tone and fill in the key data points (such as delay increase value, key bottleneck name).

[0221] Output: Comprehensive conclusion paragraph: "Overall, the road closure had a significant negative impact on the operation of the regional road network. While the main traffic demand was diverted, it exceeded the capacity of surrounding roads, resulting in a systemic increase in delays of approximately 15%."

[0222] Recommended workflow for generating cells:

[0223] Input: Conclusion tone ("significant negative impact"), intervention type ("road closure"), specific problem list ([road segment A congestion, intersection B overflow]).

[0224] Processing: Using (Problem Type = "Congestion", Intervention Type = "Road Closure", Severity Level = "Severe") as the key, query the "Measures-Suggestions Rule Base" in the domain knowledge base. A series of suggested measures were retrieved, such as ["Optimize signal timing at surrounding intersections", "Add traffic guidance signs", "Suggest adopting a time-based closure scheme"]. These suggestions were then sorted according to their "Priority Score P".

[0225] Output: A sorted list of targeted management recommendations.

[0226] S4. Input the natural language text obtained in step S3 into the report generation layer. Based on the preset template, integrate the text content, visualization charts and raw data to generate a draft of the traffic simulation result report.

[0227] Further specific work includes:

[0228] Template and content matching:

[0229] The report generator retrieves the most suitable report template from the template library based on the assessment type selected by the user in step 1 (such as traffic impact assessment) and the core conclusion tone identified by the system in step 3 (such as "significant negative impact"). This template has pre-set corresponding chapter structures (such as "Overview", "Current Situation Analysis", "Intervention Effect", "Main Conclusions", "Management Recommendations", etc.).

[0230] Dynamic Content Population and Rendering: Text Population: Descriptive text, summaries, and management recommendations output by the natural language generation engine are automatically populated into the corresponding sections of the template. Chart Insertion: Key charts generated by the intelligent analysis engine (such as regional traffic flow heatmaps, key road segment saturation trend charts, and intersection delay comparison bar charts) are inserted into the corresponding positions of the text descriptions in the report. Data Table Integration: The most important summary data (such as core KPI change comparison tables) are embedded into the report in the form of structured tables for readers to view specific values. Automatic Style Adjustment: The rendering engine automatically adjusts the page layout according to the content length to ensure aesthetically pleasing and standardized text and image layout, avoiding chart misalignment or excessive white space on the page.

[0231] S5. The output layer outputs a complete report in PDF, Word, and Web formats simultaneously, based on the initial draft of the traffic simulation results report.

[0232] Furthermore, the specific output is as follows:

[0233] PDF format report (for formal archiving and circulation):

[0234] Generation method: The report is converted into a high-quality, fixed-format PDF document using PDF rendering engines such as WeasyPrint.

[0235] Features: Beautifully formatted, tamper-proof, and meets the requirements for formal document submission and archiving. Supports security features such as headers and footers, watermarks, and digital signatures.

[0236] Word format report (for subsequent editing and adjustments):

[0237] Generation method: Generate editable .docx documents using libraries such as python-docx.

[0238] Features: It retains the complete document structure and style, and users (especially professionals) can customize, add explanations, or extract parts of the content based on the report, making it extremely flexible.

[0239] Web-based interactive dashboards (for online reporting and in-depth exploration):

[0240] Generation method: Generate HTML pages based on Jinja2 templates, integrate interactive chart components generated by libraries such as Plotly, and deploy them as web applications.

[0241] Features: This is not a static page, but a feature-rich dashboard. Users can: Filter and drill down: Use dropdown menus, sliders, and other controls to filter data by region, time period, indicator type, and other dimensions. Chart interaction: Hover the mouse over the data to view details, click the legend to show / hide data series, and zoom in / out on chart ranges. Linked analysis: Clicking on a region on the map will simultaneously update related line charts and bar charts, displaying detailed information about that region.

[0242] S6. Input the complete report obtained in step S5 into the user interaction layer for result display, and the user can conduct interactive data exploration through the Web graphical user interface.

[0243] Furthermore, the interactive data verification and exploration method in step S6 involves the user exploring the data based on the web interactive dashboard after receiving the web report.

[0244] Furthermore, for example, interactive data verification and exploration: if there is doubt about the conclusion in the report that "traffic surges on a certain alternative path", the path can be located directly on the dashboard to view the detailed change curves of its traffic and speed, verify the reliability of the conclusion, and discover details not mentioned in the report.

[0245] Report content fine-tuning: The system provides a simple feedback mechanism. For example, if users believe that the generated "management suggestions" are not prioritized in a realistic way, they can reorder or annotate the suggestions on the web interface. This feedback can be used to optimize the system's suggestion generation algorithm.

[0246] Generation process backtracking: The system records key analysis steps and decision points (such as why a certain road segment was identified as a "critical change point"). Advanced users can query this backtracking information through the interface, enhancing their trust in the automatically generated reports.

[0247] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0248] Although this application has been described above with reference to specific embodiments, various modifications can be made and components can be replaced with equivalents without departing from the scope of this application. In particular, as long as there is no structural conflict, the features in the specific embodiments disclosed in this application can be combined with each other in any way. The lack of an exhaustive description of these combinations in this specification is merely for the sake of brevity and resource conservation. Therefore, this application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for intelligent analysis and natural language report generation of traffic simulation results, characterized in that, Includes the following steps: S1. The data receiving layer receives traffic simulation results data and then configures project information; The simulation results data are divided into output data of the basic scenario and output data of the intervention scenario. The intervention scenario includes traffic planning scenario, traffic control scenario, and demand management scenario. The configuration items include the project name, major category of intervention scenarios and corresponding specific measure categories, specific measures, evaluation year, evaluation period, evaluation scope, and simulation dimensions; S2. Input the simulation result data obtained in step S1 into the intelligent analysis engine in the intelligent processing layer to calculate the indicators, and then perform key problem location and impact assessment to obtain the structured data output by the intelligent analysis engine. S2.

1. Input the simulation result data obtained in step S1 into the intelligent analysis engine in the intelligent processing layer for index calculation. The index calculation includes regional level index, road segment level index, and intersection level index. S2.

2. Construct a bottleneck identification model that integrates multiple indicators. Based on the regional, road segment, and intersection indicators obtained in step S2.1, calculate the comprehensive severity score and locate key issues. First, calculate the saturation deterioration score S_vc. The calculation formula is as follows: ； Among them, VOC baseline VOCs intervention It represents the saturation level of a road segment or intersection under the basic and intervention plans. w_voc is the weight of the saturation index, and max is the maximum value function. The speed reduction score S_speed is calculated using the following formula: ； Among them, Speed baseline Speed intervention It refers to the speed of the basic plan and intervention plan for road segments or intersections, where w_speed is the weight of the speed index; Define the downgrade score mapping table with grades from A to F, and calculate the service level downgrade score S_los using the following formula: ； Where LOS_Degrade_Score is the degrade score, and w_los is the weight of the service level indicator; The overall severity score S_total is obtained by summing the scores of all indicators for an analysis unit. The calculation formula is as follows: S_total = S_vc + S_speed + S_los; Set a comprehensive threshold T_bottleneck and a single veto threshold, then identify and judge key issues based on whether S_total > T_bottleneck or VOC. intervention > 0.95 or LOS intervention If the rating is F, then the analysis unit is marked as a critical bottleneck, where LOS intervention Service level rating for the intervention program; S2.

3. The overall impact of the intervention measures is assessed using a traffic intervention impact quantification assessment model based on entropy weight-TOPSIS, and the impact levels are classified. S3. Input the structured data output by the intelligent analysis engine obtained in step S2 into the natural language generation engine in the intelligent processing layer, and fill and synthesize it into natural language text that conforms to traffic engineering specifications based on the rules, templates and mapping relationships in the domain knowledge base. S4. Input the natural language text obtained in step S3 into the report generation layer. Based on the preset template, integrate the text content, visualization charts and raw data to generate a draft of the traffic simulation result report. S5. The output layer outputs a complete report in PDF, Word, and Web formats simultaneously, based on the initial draft of the traffic simulation results report. S6. Input the complete report obtained in step S5 into the user interaction layer for result display, and the user can conduct interactive data exploration through the Web graphical user interface.

2. The intelligent analysis and natural language report generation method for traffic simulation results according to claim 1, characterized in that, In step S2, the regional indicators are calculated as follows: total travel demand of the entire network or a designated area, total regional traffic flow, total regional vehicle kilometers, regional average speed, total regional congestion mileage, and regional congestion cost. The road segment level indicators are used to calculate the traffic flow, road segment saturation, average speed, road segment density, and road segment delay time for each road segment. Intersection-level indicators calculate the node traffic, service level, overflow degree, and total node delay for each intersection.

3. The intelligent analysis and natural language report generation method for traffic simulation results according to claim 2, characterized in that, The specific implementation method of step S3 includes the following steps: S3.

1. Construct a domain knowledge base, including an intervention-assessment mapping library, a measure-suggestion rule library, and a traffic semantic vector library; The data structure of the intervention-assessment mapping library is a lookup table with intervention category and specific measures as keys. The core fields include a list of core assessment dimensions, an assessment indicator weight vector, indicator-conclusion semantic mapping rules, and conclusion template ID. The data structure of the action-recommendation rule base is a set of rules with problem type, intervention type, and severity level as keys. Core fields include a list of recommended actions, triggering conditions, and a priority score P, where P is calculated using the following formula: ； Where EffectScore is the effect score and CostScore is the cost score, which are predefined by expert experience, and α and β are the adjustment coefficients for the effect score and cost score, respectively; The data structure of the traffic semantic vector library is a finely tuned professional word embedding model or lookup table. The core content includes degree adverb mapping relationships, state adjective mapping relationships, and causal association lexicon. The domain knowledge base is constructed and iterated through a combination of hard-coded expert rules and dynamic optimization using machine learning. S3.

2. For the structured data output by the intelligent analysis engine obtained in step S2, obtain the corresponding core assessment dimension list and weight vector W from the intervention-assessment mapping library, and transform the numerical changes into qualitative descriptions according to the indicator-conclusion semantic mapping rules. Then, the problem type is searched in the measure-suggestion rule base, and all the retrieved suggested measures are sorted according to their priority score P, and the top-N measures are output; Then, the natural language generation engine uses a traffic semantic vector library to assemble qualitative descriptions and suggested measures into natural language text that conforms to traffic engineering specifications.

4. The intelligent analysis and natural language report generation method for traffic simulation results according to claim 3, characterized in that, The specific implementation method of the domain knowledge base in step S3 includes the following steps: For feedback-based learning-based weight optimization, the method of hard-coding expert rules is used. For the evaluation index weight vector, the gradient descent method is used for adjustment, and the calculation formula is as follows: ； in, The original weights of the evaluation index m to be adjusted. η is the adjusted weight of the metric m to be evaluated, η is the learning rate, and FeedbackStrength is the feedback strength, which is quantified by user behavior. The domain knowledge base, which is hard-coded by expert rules, is dynamically optimized and expanded using machine learning. By using association rule mining or clustering methods to analyze historical simulation projects and their final adopted manual solutions, each historical project is first represented as a feature vector. Then, the Apriori algorithm is used to discover frequent itemsets. Finally, new rules with confidence scores greater than a preset threshold are added to the measure-suggestion rule base.

5. The intelligent analysis and natural language report generation method for traffic simulation results according to claim 4, characterized in that, The interactive data exploration in step S6 involves users exploring data based on the web interactive dashboard after receiving the web report.

6. A system for intelligent analysis and natural language report generation of traffic simulation results, implementing the method for intelligent analysis and natural language report generation of traffic simulation results as described in claim 1, characterized in that, The system includes a user interaction layer, a data access layer, an intelligent processing layer, a report generation layer, and an output layer, which are connected in sequence, and the output layer is then connected to the user interaction layer. The user interaction layer includes a web graphical user interface, a mobile application module, and a voice / text input interface; The data access layer includes a simulation result data receiving module, a project configuration management module, and an evaluation type and template selection module; The intelligent processing layer includes an intelligent analysis engine and a natural language generation engine. The intelligent analysis engine includes a key indicator identification unit, an impact assessment unit, and a related problem location unit. The natural language generation engine includes a domain knowledge base, a text generation unit, a conclusion generation unit, and a suggestion generation unit. The report generation layer includes a multi-format exporter, a data population and rendering engine, and a template selector; The output layer includes a web interactive dashboard, a PDF report generation module, and a Word report generation module.

7. The intelligent analysis and natural language report generation system for traffic simulation results according to claim 6, characterized in that, The key performance indicator identification unit in the intelligent analysis engine uses a multi-dimensional feature extraction algorithm to automatically identify key performance indicators from the simulation result data, including traffic capacity, service level, and delay time. The problem location unit uses a bottleneck identification model that integrates multiple indicators to identify traffic bottlenecks caused or exacerbated by intervention measures; the impact assessment unit establishes a multi-level assessment indicator system, combining quantitative calculation and qualitative analysis to conduct graded assessments of the impact.

8. The intelligent analysis and natural language report generation system for traffic simulation results according to claim 7, characterized in that, The domain knowledge base in the natural language generation engine connects the text generation unit, conclusion generation unit, and suggestion generation unit. The domain knowledge base is constructed through a combination of expert rule initialization and data-driven optimization, including professional analysis logic and expression standards in the field of traffic engineering. The text generation unit, based on template and rule-driven methods and combined with a deep learning language model, automatically generates professional text conforming to traffic engineering standards. The conclusion generation unit uses information extraction and text summarization techniques to extract key information from multi-dimensional analysis results and generate conclusive statements. The suggestion generation unit uses an inference engine based on the domain knowledge base to generate management suggestions.

Images