Engineering cost estimation method and system, electronic equipment and storage medium

Abstract

The invention provides a project cost estimation method and system, electronic equipment and a storage medium, and relates to the field of project cost. The method comprises the following steps: acquiring historical project cost sample data, and constructing a project cost knowledge graph; then, acquiring information of a project to be estimated, inputting the information into the knowledge graph to obtain a structured representation, and mapping the structured representation into the graph to obtain an association mode; thirdly, integrating the engineering information based on an association mode, extracting engineering influence factor characteristics, and generating a project cost influence factor representation vector through an attention mechanism; meanwhile, inputting the engineering information into a preset standard model, and outputting an engineering cost standard; and finally, constructing a project cost model based on the project cost influence factor representation vector, the project information and the cost standard, and inputting the project information into the model to obtain a final project cost estimation result. According to the method, through semantic understanding and correlation analysis of engineering information, the accuracy of engineering cost estimation is improved from the perspective of data driving.

Description

Technical Field

[0001] The present invention relates to the field of engineering cost, and in particular to an engineering cost estimation method, system, electronic equipment and storage medium. Background Art

[0002] Project cost estimation is an important part of engineering management. Its purpose is to provide a basis for engineering decision-making, investment control, bidding, etc. by predicting and evaluating the cost of engineering projects before engineering design or construction. Traditional engineering cost estimation mainly relies on the experience and knowledge of professionals. By referring to historical project data, cost indicators, quota standards, etc., the cost of new engineering projects is estimated and adjusted by analogy. However, traditional engineering cost estimation methods have certain limitations. First, the cost of engineering is affected by many factors, such as project scale, construction technology, materials and equipment, labor costs, construction environment, etc. There are often complex nonlinear relationships between these factors, which are difficult to accurately grasp through simple analogy and adjustment. Secondly, engineering projects have individual characteristics, and there are large differences between different projects. Historical data and empirical models are difficult to directly migrate and reuse. Furthermore, the traditional method does not make systematic and in-depth use of engineering cost knowledge, and it is difficult to fully explore and utilize implicit knowledge such as industry experience, standards and specifications. Therefore, the accuracy and efficiency of traditional engineering cost estimation cannot meet the increasingly complex needs of engineering management. Summary of the invention

[0003] The present application provides a construction cost estimation method, system, electronic device and storage medium to improve the accuracy of construction cost estimation.

[0004] In a first aspect, the present application provides a method for estimating construction cost, the method comprising: Acquire historical engineering cost sample data, and construct an engineering cost knowledge graph based on the historical engineering cost sample data; Acquire engineering information of the project to be estimated, input the engineering information into the engineering cost knowledge graph, and output a structured representation of the project to be estimated; Mapping the structured expression to the semantic representation in the engineering cost knowledge graph to obtain the association pattern of the engineering project to be estimated; Based on the association pattern, the engineering information is managed and integrated to obtain engineering integration information, and features are extracted from the engineering integration information to obtain feature descriptions of engineering influencing factors; The characteristic expressions of the engineering influencing factors are weighted and fused through the attention mechanism to generate a representation vector of the project cost influencing factors; Inputting the engineering information into a preset engineering cost standard model and outputting the engineering cost standard; A project cost model is constructed based on the project cost influencing factor characterization vector, the project information and the project cost standard, and the project information is input into the project cost model to obtain a project cost result.

[0005] In the above technical solution, by constructing a knowledge graph of engineering cost, the domain knowledge is structured and semantically represented, and the influencing factors are weighted and integrated using the attention mechanism, and the estimation process is constrained and guided by industry standards, an integrated and intelligent engineering cost estimation model is constructed. Compared with traditional methods, this method can consider various influencing factors more comprehensively and accurately, make full use of historical experience and industry knowledge, and dynamically grasp the characteristics of engineering projects, thereby significantly improving the accuracy, efficiency and practicality of engineering cost estimation.

[0006] In a second aspect of the present application, a construction cost estimation system is provided, the system comprising: A knowledge graph construction module, used to obtain historical engineering cost sample data and construct an engineering cost knowledge graph based on the historical engineering cost sample data; A preprocessing module, used to obtain engineering information of the project to be estimated, input the engineering information into the engineering cost knowledge graph, and output a structured representation of the project to be estimated; An association pattern determination module is used to map the structured expression into the semantic representation in the engineering cost knowledge graph to obtain an association pattern of the engineering project to be estimated; An integration and extraction module, used to manage and integrate the engineering information based on the association pattern to obtain engineering integration information, and to extract features from the engineering integration information to obtain a feature description of engineering influencing factors; An attention mechanism module, used for weighted fusion of the characteristic expressions of the engineering influencing factors through the attention mechanism to generate a representation vector of the project cost influencing factors; A standard selection module, used to input the engineering information into a preset engineering cost standard model and output the engineering cost standard; The cost module is used to construct a project cost model based on the project cost influencing factor representation vector, the project information and the project cost standard, and input the project information into the project cost model to obtain the project cost result.

[0007] In a third aspect of the present application, a computer storage medium is provided, wherein the computer storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the above method steps.

[0008] In the fourth aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the above method.

[0009] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. This invention constructs an integrated and intelligent engineering cost estimation model by constructing a knowledge graph of engineering cost, structuring and semantically representing domain knowledge, using the attention mechanism to weightedly fuse influencing factors, and combining industry standards to constrain and guide the estimation process. Compared with traditional methods, this method can more comprehensively and accurately consider various influencing factors, make full use of historical experience and industry knowledge, and dynamically grasp the characteristics of engineering projects, thereby significantly improving the accuracy, efficiency and practicality of engineering cost estimation. 2. This application improves the quality of engineering cost data through data preprocessing, reveals engineering cost knowledge elements through relational entity extraction, establishes a consistent industry knowledge pedigree through entity fusion, and realizes structured management of engineering cost knowledge through triple representation and knowledge graph construction.

[0010] 3. This application manages and integrates engineering information based on the association model, which is a key to deepening the application of engineering informationization. By adapting engineering information to the association model and optimizing, completing and formatting the adapted information, massive engineering management data is integrated into semantically rich and structurally standardized engineering integration information, so that the engineering management wisdom scattered in various disciplines and stages can be expressed in a systematic and explicit manner, greatly improving the sharing level and value density of engineering data. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 A schematic diagram of a process flow of a method for estimating construction cost provided in an embodiment of the present application; Figure 2 An architectural diagram of a construction cost estimation system provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in this application. DETAILED DESCRIPTION

[0012] In order to enable technicians in this field to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments.

[0013] In the description of the embodiments of the present application, words such as "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "for example" or "for example" is intended to present related concepts in a specific way.

[0014] In the description of the embodiments of the present application, the meaning of the term "multiple" refers to two or more. For example, multiple systems refer to two or more systems, and multiple screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. The terms "include", "comprise", "have" and their variations all mean "including but not limited to", unless otherwise specifically emphasized.

[0015] In order to facilitate the understanding of the method and system provided by the embodiments of the present application, before introducing the embodiments of the present application, the background of the embodiments of the present application is first introduced. Project cost estimation is a crucial link in the field of engineering management. Its main purpose is to provide a reliable basis for engineering decision-making, investment control, bidding, etc. by predicting and evaluating the cost of the project before the start of engineering design or construction. For a long time, traditional project cost estimation has mainly relied on the experience and knowledge of professionals. They make analog estimates and appropriate adjustments to the cost of new projects by referring to historical project data, cost indicators, quota standards, etc.

[0016] However, traditional methods of engineering cost estimation have some inherent limitations. First, engineering costs are often affected by a variety of factors, such as project scale, construction technology, materials and equipment, labor costs, construction environment, etc. There are usually intricate nonlinear relationships between these factors. It is difficult to accurately grasp their impact on the cost through simple analogies and adjustments. Secondly, each engineering project has its own uniqueness, and there are often large differences between different projects. Historical data and empirical models are difficult to directly migrate and reuse, and personalized analysis and adjustments are required for specific projects. In addition, the traditional method is not systematic and in-depth enough in the use of engineering cost knowledge, and it is difficult to fully explore and utilize the rich experience, standards and specifications accumulated in the industry, resulting in a waste of valuable knowledge resources. Therefore, in the face of increasingly complex engineering management needs, traditional engineering cost estimation can no longer fully meet the requirements in terms of accuracy and efficiency.

[0017] After the background introduction of the above content, those skilled in the art can understand the problems existing in the prior art. The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0018] Based on the above background technology, please refer to Figure 1 , Figure 1 A flow chart of a method for estimating construction cost provided in an embodiment of the present application. The system can be implemented by a computer program or can be run as an independent tool application. Specifically, in an embodiment of the present application, the method can be applied on a server, but can also be applied on electronic devices such as a server. A method for estimating construction cost includes the following steps: S101, obtaining historical engineering cost sample data, and constructing an engineering cost knowledge graph based on the historical engineering cost sample data; Specifically, we first obtain historical engineering cost sample data. These sample data usually include various information of completed engineering projects, such as project scale, construction technology, materials and equipment, labor costs, construction environment, etc., as well as corresponding engineering cost data. By collecting and collating a large amount of historical engineering cost sample data, we can provide a rich and reliable data foundation for the construction of the knowledge graph.

[0019] After obtaining historical engineering cost sample data, the present invention constructs an engineering cost knowledge graph based on these data. Specifically, the sample data is first preprocessed, such as data cleaning, format conversion, etc., to improve data quality and consistency. Then, natural language processing, data mining and other technologies are used to extract key information related to engineering cost from the sample data, such as engineering characteristics, cost indicators, influencing factors, etc. Then, based on the logical relationship and semantic association between these key information, the basic architecture of the engineering cost knowledge graph is constructed.

[0020] Based on the above embodiment, as an optional embodiment, the construction of the engineering cost knowledge graph based on the historical engineering cost sample data includes: S201, preprocessing the historical engineering cost sample data to obtain preprocessed historical engineering cost sample data; Specifically, the preprocessing process of historical engineering cost sample data is as follows: First, the original sample data is cleaned. This step is mainly to detect and correct quality problems such as errors, inconsistencies and missing values ​​in the data. By defining a series of data quality rules and constraints, such as numerical range, format specifications, logical consistency, etc., each sample data is checked and verified one by one. For data that does not meet the quality requirements, strategies such as deletion, correction, and interpolation can be adopted to ensure the accuracy and completeness of the data. Secondly, the cleaned sample data is converted and standardized. Since the engineering cost data from different sources have differences in indicator system, measurement unit, time scale, etc., they need to be converted to a unified representation for comparison and analysis. Specifically, mapping tables, conversion functions and other methods can be used to map indicators of different standards to a unified indicator system, unify values ​​of different units into standard units, and unify data at different time points to the same time base. Through data conversion and standardization, the representation differences between data from different sources can be eliminated, and the comparability and interoperability of data can be improved.

[0021] Finally, the converted sample data is integrated and fused. Engineering cost data is usually scattered in different business systems and databases, and they need to be integrated into a unified data set for global analysis and mining. Specifically, data warehouses, data lakes and other technologies can be used to integrate scattered data according to a unified data model and architecture, and duplicate and redundant data can be identified and merged to form a high-quality and comprehensive engineering cost sample data set. Through data integration and fusion, the data barriers between different systems and departments can be broken down, and the full life cycle management and value mining of engineering cost data can be realized.

[0022] S202, extracting relationship entities related to the construction cost from the preprocessed historical construction cost sample data; Specifically, the process of extracting relational entities from the preprocessed historical engineering cost sample data is as follows: First, based on the knowledge and business experience in the engineering cost field, a standardized relational entity pattern library is defined. The pattern library contains common concepts, entity types and their attribute characteristics in the engineering cost field, such as engineering projects, cost indicators, influencing factors, management systems, etc., as well as the relationships between them, such as inclusion, influence, reference, etc. The relational entity pattern library provides a unified representation framework and specification for entity extraction, which helps to improve the consistency and interpretability of the extraction results.

[0023] Secondly, natural language processing, data mining and other technologies are used to automatically extract relational entities from preprocessed historical engineering cost sample data. Specifically, rule-based, statistical-based, and deep learning-based entity recognition methods can be used, such as regular expression matching, conditional random fields, and convolutional neural networks. According to the predefined relational entity pattern library, entity mentions and attribute values ​​in the sample data can be identified. At the same time, named entity disambiguation and coreference resolution can be used to normalize and link multiple mentions of the same entity to improve the accuracy and completeness of entity recognition.

[0024] S203, performing fusion processing on the relationship entities to construct a consistent entity relationship pedigree; Specifically, the process of fusing relational entities to construct a consistent entity relationship genealogy is as follows: First, based on knowledge organization frameworks such as ontology and concept hierarchy, the type system of relational entities is standardized and defined. Through the systematic combing and abstraction of knowledge in the field of engineering cost, a hierarchical and structured entity type genealogy is formed, and the semantic connotations, attribute characteristics, inheritance relationships, etc. of various entities are clearly defined. On this basis, with reference to the ontology description languages ​​and specifications commonly used in the industry, such as OWL and RDFS, the entity type genealogy is formally represented to form an ontology in the field of engineering cost. This domain ontology provides a unified semantic framework and specification for the fusion of relational entities, which helps to improve the consistency and semantic accuracy of entity fusion.

[0025] Secondly, natural language processing, knowledge representation and other technologies are used to perform semantic comparison and similarity calculation on the extracted relational entities. Specifically, entity features can be extracted from multiple dimensions such as entity name, attribute value, context, etc., and the similarity between different entities can be calculated through literal matching, semantic matching, structural matching and other methods. For entities with high semantic similarity and referring to the same concept, such as "project cost" and "project cost", entity equivalence links are performed to map them to the standardized entity types in the domain ontology to eliminate the differences and redundancy of entity representation. For entities with high semantic relevance and specific relationships, such as "bill of quantities" and "tender control price", entity relationship links are performed to clarify their semantic associations in the ontology. Through entity linking, the originally scattered and independent relational entities are integrated into a unified semantic framework to form a consistent and standardized entity genealogy.

[0026] S204, constructing a triple entity relationship representation based on the entity relationship pedigree, and constructing the engineering cost knowledge graph based on the triple entity relationship representation.

[0027] Specifically, the process of constructing triple entity relationship representation based on entity relationship genealogy and constructing engineering cost knowledge graph based on triple entity relationship representation is as follows: First, perform structural analysis and semantic conversion on the consistent entity relationship genealogy. According to the entity type, attribute, relationship and other elements in the genealogy, the entity relationship genealogy is converted into a semantic network represented by triples using ontology description languages ​​such as RDF and OWL. Among them, entities, as the subject and object of triples, represent the core concepts and objects in the field of engineering cost; relationships, as the predicates of triples, represent the semantic associations between entities. Through ontology mapping, the hierarchical and structured entity relationship genealogy is converted into a flat and chained triple sequence, forming a formal knowledge representation that is easy for computers to understand and process.

[0028] Then, natural language processing, pattern matching and other technologies are used to automatically extract and construct triple knowledge units from the text information such as the definition and description of the entity relationship spectrum. For example, from the description "building area is an important factor affecting the project cost", the triple "building area, impact, project cost" can be extracted. Through semantic analysis and knowledge extraction of the entity spectrum, the entity semantic relationship implicit in the unstructured text is portrayed in the form of explicit triples, further enriching the triple knowledge base. At the same time, for the extracted triples, conflict resolution and quality assessment are required to identify and filter erroneous and redundant triples to ensure the accuracy and consistency of triple knowledge.

[0029] S102, obtaining engineering information of the project to be estimated, and inputting the engineering information into the engineering cost knowledge graph, and outputting a structured representation of the project to be estimated; Specifically, there are many ways to obtain engineering information, such as extracting it from engineering design documents, construction plans, contract documents, etc., or directly obtaining it through on-site inspections, expert interviews, etc. In the acquisition process, attention should be paid to the standardization and normalization of engineering information to ensure that it is compatible with the conceptual system in the knowledge graph. If necessary, engineering information needs to be preprocessed, such as format conversion, data cleaning, etc., to improve its quality and usability.

[0030] After obtaining the engineering information of the project to be estimated, the present invention inputs this engineering information into the constructed engineering cost knowledge graph to achieve a structured representation of the engineering project. Specifically, first match the key elements in the engineering information with the entities, attributes, etc. in the knowledge graph to find their corresponding nodes in the knowledge graph. Then, based on the relationship between these nodes and the semantic logic in the engineering information, a structured subgraph representing the engineering project is constructed on the knowledge graph. This subgraph not only contains the nodes representing the various elements of the engineering project, but also the association relationship between these nodes, forming a complete and coherent representation of the engineering project.

[0031] The process of inputting engineering information into the knowledge graph and outputting a structured representation is actually using the knowledge graph to perform semantic understanding and association analysis on unstructured engineering information. Through this process, various elements and features of the engineering project are mapped to the predefined concept system in the knowledge graph, giving it clear semantic types and logical relationships. This structured representation of engineering projects not only standardizes the organization of engineering information, but also provides semantic support for subsequent cost analysis and estimation.

[0032] S103, mapping the structured expression to the semantic representation in the engineering cost knowledge graph to obtain the association pattern of the engineering project to be estimated; Specifically, the present invention uses the ontology framework and semantic rules predefined in the knowledge graph to map each node and relationship in the structured representation to the corresponding semantic categories and attributes. For example, the node representing the engineering project is mapped to the "engineering project" category, the node representing the construction process is mapped to the "construction process" category, and the relationship between them is mapped to the "adoption" attribute, etc. Through this mapping, the originally independent nodes and relationships in the structured representation are given clear semantic meanings, forming a semantically coherent and logically consistent engineering project representation.

[0033] In the mapping process, it is also necessary to use the semantic reasoning rules in the knowledge graph to further analyze and complete the structured representation. For example, using the reasoning rules "steel structure engineering uses welding connections" and "welding connections require welding materials and welders", the implicit relationship "steel structure engineering requires welding materials and welders" can be supplemented in the structured representation. Through semantic reasoning, the nodes and relationships in the structured representation are further associated and expanded, forming a more complete and rich semantic network of engineering projects.

[0034] The present invention maps the structured representation to the engineering project association pattern obtained in the semantic representation, which is actually a local subgraph highly related to the project extracted from the knowledge graph. This association pattern not only contains the various attributes and characteristics of the project, but also reveals the intrinsic connection and influence mechanism between these attributes and characteristics. For example, by analyzing the semantic path in the association pattern, it can be found that the influence transmission chain of "construction technology → adoption → materials and equipment → price → project cost" is revealed, revealing the impact of construction technology selection on project cost.

[0035] Based on the above embodiment, as an optional embodiment, mapping the structured expression to the semantic representation in the engineering cost knowledge graph to obtain the association pattern of the engineering project to be estimated includes: S301, parsing the structured expression into an entity relationship representation, and matching the entity relationship representation with entities and relationships in the engineering cost knowledge graph to obtain an association relationship entity representation; Specifically, the process of mapping the structured expression to the semantic expression in the engineering cost knowledge graph and obtaining the association pattern of the engineering project to be estimated is as follows: First, the natural language processing technology is used to perform semantic analysis on the structured expression of the engineering project to be estimated. Through the steps of word segmentation, part-of-speech tagging, named entity recognition, syntactic analysis, etc., the key entities (such as project name, construction scale, building type, etc.) and semantic relationships (such as location, land occupation, building area, etc.) in the engineering description are extracted, and the logical priority and semantic dependency between entities are clarified through methods such as reference resolution and semantic role labeling. On this basis, referring to the ontology definition and semantic specification in the knowledge graph, the extracted entities and relationships are converted into standardized entity relationship representations. For example, "the project is located in Beijing, covers an area of ​​2,000 square meters, and has a building area of ​​5,000 square meters" is parsed into triples such as "project-located-Beijing", "project-land occupation-2,000 square meters", and "project-building area-5,000 square meters". Through semantic parsing and standardized representation, the user's engineering description is converted into a structured form that is easy for computers to understand and process, facilitating subsequent semantic association.

[0036] Then, the entity relationship representation obtained by parsing the structured expression is matched with the entities and relationships in the engineering cost knowledge graph. The semantic similarity calculation, graph matching and other technologies are used to calculate the similarity between the project to be estimated and the existing project cases in the knowledge graph at the entity and relationship levels. Through attribute matching, the consistency of the key attributes such as construction scale, building type, and location between the project to be estimated and the project in the graph is identified; through structural matching, the similarity between the project to be estimated and the project in the graph in entity composition and topological structure is examined; through semantic matching, the degree of proximity between the project to be estimated and the project in the graph in semantic connotation is evaluated. Based on the comprehensive analysis of entity attributes, relationship structure, and semantic similarity, the entity in the knowledge graph that is closest to the project to be estimated in semantic representation is identified, and the association relationship between the entity and the project to be estimated entity is obtained to form an association relationship entity representation.

[0037] For example, suppose the description of the project to be estimated is "an office building project is located in Beijing, with an area of ​​2,000 square meters, a building area of ​​5,000 square meters, 20 floors above ground and 1 floor underground". After semantic analysis and normalization, the entity relationship representation of the project is obtained as "office building project-located in Beijing", "office building project-area-2,000 square meters", "office building project-building area-5,000 square meters", "office building project-ground floors-20 floors", "office building project-underground floors-1 floor", etc. The entity relationship representation is matched with the knowledge graph, and it is found that it is highly similar to the "Beijing A Office Building Project" in the graph in terms of construction scale, number of floors, and location, while the similarity with "Beijing B Factory Project" and "Beijing C Residential Project" is relatively low. Based on this, it is identified that the project to be estimated has the highest correlation with "Beijing A Office Building Project", and the association relationship entity representation "Project to be estimated-association-Beijing A Office Building Project" is generated, indicating that the two have similar positions and connections in the semantic space of the knowledge graph. Through semantic matching and association analysis, a semantic mapping is formed between the project to be estimated and the knowledge graph. With the help of the cost knowledge and experience of related projects in the graph, the cost level of this project can be estimated and inferred by analogy.

[0038] S302, performing pattern mining on the engineering cost knowledge graph based on the association relationship entity representation to obtain the association pattern of the engineering project to be estimated.

[0039] Specifically, the process of pattern mining in the engineering cost knowledge graph based on the entity representation of the association relationship to obtain the association pattern of the engineering project to be estimated is as follows: First, the association entities of the engineering project to be estimated are expanded in the knowledge graph from the two dimensions of entity and relationship. On the one hand, starting from the entity representation of the association relationship, the first-order entities directly connected to the engineering project to be estimated are extracted, such as the construction unit, design unit, construction unit, raw material supplier, etc. of the project, to reveal the association network of the main body of the engineering project; on the other hand, starting from the first-order entity, continue to expand outward, extract the second-order entities directly connected to it, such as the policies and regulations, market prices, climate characteristics, etc. of the location of the project, to reveal the macro semantic environment of the engineering project. At the same time, the semantic relationships between the first-order and second-order entities are mined, such as supply and demand relationships, subordinate relationships, reference relationships, etc., to characterize the interaction patterns of different entities in the entire life cycle of the project. Through entity association expansion and relationship mining, the engineering project to be estimated is expanded from an isolated node to a three-dimensional association subgraph, covering the entities and relationships that reflect the core characteristics of the project.

[0040] Then, in the associated subgraph of the project to be estimated, algorithms such as association rule learning and frequent subgraph mining are used to discover universal cost patterns from massive engineering case data. Through association rule learning, the intrinsic relationship between project scale, material, process and cost can be obtained, such as "the larger the building area, the lower the unit cost"; through frequent subgraph mining, typical cost combinations under specific external conditions can be obtained, such as "Beijing municipal office buildings mostly use glass curtain walls, and steel structures mostly use welding connections". Matching the associated subgraph of the project to be estimated with these universal cost knowledge can obtain the association pattern specifically for the project. For example, if the project to be estimated is a government office building in Beijing, the matched association pattern may include "glass curtain wall, welded steel structure, and custom furniture are typical material combinations", "the unit price of glass curtain wall is about 1,500 yuan / square meter", "the unit cost of office building is usually around 5,000 yuan / square meter", etc. The association pattern reflects the universal laws of the project to be estimated in terms of material selection, price level, cost composition, etc., and is the key basis for conducting accurate cost analysis of the project.

[0041] S104, managing and integrating the project information based on the association pattern to obtain project integration information, and extracting features from the project integration information to obtain a feature description of project influencing factors; Specifically, firstly, the semantic paths and influence transmission chains in the association pattern are used to identify the key nodes and relationships that have direct or indirect influence on the project cost. For example, the influence transmission chain of "construction technology → adoption → materials and equipment → price → project cost" is identified from the association pattern, and "construction technology" and "materials and equipment" are two key influencing factors. Then, structured and standardized feature attributes are extracted from these key nodes and relationships using ontology mapping, attribute inheritance and other methods. For example, feature attributes such as "process name", "labor requirements", and "machinery requirements" are extracted from the "construction technology" node, and feature attributes such as "material name", "material price", and "equipment model" are extracted from the "materials and equipment" node. Finally, the extracted feature attributes are organized into a structured feature vector or feature table to form a feature description of the project influencing factors.

[0042] Based on the above embodiment, as an optional embodiment, the engineering information is managed and integrated based on the association mode to obtain engineering integration information, including: S401, converting the engineering information into structured engineering information, and matching the structured engineering information with the association pattern to obtain adapted engineering information; Specifically, the process of managing and integrating engineering information based on the association pattern to obtain engineering integration information is as follows: First, natural language processing and information extraction technology are used to convert unstructured engineering information into structured semantic representation. Information extraction is performed on engineering documents to extract key entities (such as personnel, materials, equipment, etc.) and semantic relationships (such as participation, procurement, use, etc.) in the documents to form entity relationship triples; semantic recognition is performed on engineering drawings to extract information such as components, space, and dimensions in the drawings to form design parameter triples; content analysis is performed on engineering contracts to extract elements such as terms, responsibilities, and processes in the contract to form contract logic triples. At the same time, the semantic types of triples are identified and standardized to eliminate ambiguities caused by synonyms and antonyms. On this basis, ontology languages ​​such as RDF and OWL are used, and triples are used as basic units to convert various types of engineering information into structured semantic network representations to form an engineering knowledge element library with clear semantics and standardized structures. For example, the sentence "the steel specification used in a steel structure project is Q345B, and the unit price is 5,000 yuan / ton" is converted into structured triples such as "steel structure project-used-steel", "steel-specification-Q345B", and "steel-unit price-5,000 yuan / ton". Then, the structured engineering information is semantically matched with the association pattern to screen out the combination of engineering elements that meet the specific cost model and obtain the adapted engineering information. Based on semantic similarity calculation, ontology mapping and other technologies, the degree of adaptation between engineering information and association patterns in terms of entity type, relationship structure, semantic constraints, etc. is measured. First, the core entities of engineering information and association patterns are matched, such as matching "steel structure engineering" in engineering information with "steel structure office building engineering" in the association pattern to determine their semantic correspondence in entity type. Then, the key attributes of the two are matched, such as matching "steel specification" with "common steel specifications" in the association pattern, and matching "geographic location" with "city where the project is located" in the association pattern. On the basis of entity and attribute matching, we further examine whether the semantic relationship between information can be adapted to the relational constraints in the pattern, such as whether the "adoption" relationship can correspond to the "engineering-material" relationship in the association pattern. Through semantic matching, we can filter out a subset of engineering information that is highly consistent with the association pattern in terms of entity type, attribute value, and relational structure, and remove redundant and irrelevant information noise. Taking steel structure engineering as an example, highly compatible information may include: "geographic location-Beijing", "material specification-Q345B", "material unit price-5000 yuan / ton", "construction unit-Company A", "project manager-Zhang San", etc., while information such as "environmental impact assessment report" and "garden design" may be filtered out. This matching process can be automatically implemented based on the graph isomorphism algorithm without manual intervention.

[0043] Through the structuring and adaptation of engineering information, scattered engineering data can be aggregated into an engineering integrated information knowledge base with clear semantics, standardized structure, and high adaptability to industry knowledge. The original engineering information is converted into a structured triple representation through semantic extraction and ontology mapping, which can effectively eliminate the heterogeneity of data from different sources and in different forms and reveal its inherent semantic connection; then, the structured information is adapted and organized with reference to the domain knowledge model, which can filter out irrelevant data, highlight value-intensive areas, and enable engineering information to form precise matching and deep collaboration with industry knowledge in the semantic dimension.

[0044] S402, optimizing and completing the adaptation project information to obtain target adaptation project information, and integrating the target adaptation project information according to a preset format to obtain the project integration information.

[0045] Specifically, the process of converting engineering information into structured engineering information and matching it with the associated pattern to obtain the adapted engineering information is as follows: First, natural language processing and computer vision technologies are used to perform structured analysis on the original engineering information. For engineering information in text form, key entities and their attributes in the text are extracted through word segmentation, part-of-speech tagging, named entity recognition, syntactic analysis and other methods, and the semantic relationship between entities is clarified, and finally the unstructured text is converted into structured triples. For example, the text description of "The project is located in Dongcheng District, Beijing, covers an area of ​​2,000 square meters, has 20 floors above ground and 1 floor underground" is parsed into structured triples such as "The project is located in Dongcheng District, Beijing", "The project covers an area of ​​2,000 square meters", "The project has 20 floors above ground", and "The project has 1 floor underground". For non-text information such as graphics and images, computer vision technologies such as target detection and semantic segmentation are used to identify key information such as components, space, and text in the image, and the internal relationship between the elements in the image is revealed through image understanding and knowledge reasoning, and finally an interpretable semantic network is formed. Through information extraction and semantic representation, the original engineering information is uniformly converted into a structured form that is easy for computers to understand and process, facilitating semantic association and logical fusion.

[0046] Then, the structured engineering information is matched with the association pattern to obtain the adaptation engineering information. The semantic similarity calculation, graph isomorphism matching and other technologies are used to evaluate the consistency between the structured engineering information and the association pattern at the semantic and structural levels. On the one hand, the proximity of the entities, attributes and relationships of the two in the semantic space is compared to identify semantically equivalent or similar information elements; on the other hand, the similarities and differences in the topological structures and connection methods of the two are compared to identify subgraphs or paths that can be mapped to each other. On the basis of comprehensive semantic similarity and structural isomorphism, information fragments that are highly consistent with the association pattern are extracted from the structured engineering information as adaptation engineering information. The adaptation information inherits the semantic structure and representation of the association pattern, and can be seamlessly connected with the association pattern to achieve a smooth transition from the original information to the pattern information. For example, if the association pattern defines pattern fragments such as "project-located in-administrative district", "project-floor area-value", "project-ground floors-value", and "project-underground floors-value", then triples such as "the project-located in-Dongcheng District, Beijing", "the project-floor area-2,000 square meters", "the project-ground floors-20 floors", and "the project-underground floors-1 floor" can be identified as suitable adapted project information.

[0047] Through structured transformation and pattern matching, unstructured and heterogeneous original engineering information is abstracted and refined into structured and standardized adaptive information, breaking through the barriers of information from different sources and in different forms, and realizing the correlation and integration of multi-source heterogeneous information.

[0048] Based on the above embodiment, as an optional embodiment, the feature extraction of the engineering integration information to obtain the feature description of the engineering influencing factors includes: S501, identifying influencing factors of the engineering integration information according to preset expert knowledge to obtain engineering influencing factors; Specifically, the process of identifying the influencing factors of engineering integration information based on the preset expert knowledge and obtaining the engineering influencing factors is as follows: First, with the help of the theoretical knowledge and practical experience of experts in the field, summarize the standard specifications such as engineering cost budget quotas and budgetary indicators, and summarize the universal engineering cost influencing factors. Through the systematic sorting of the professional quotas and budgetary indicators at the national and local levels, it can be found that although the cost estimation calibers of different engineering professions and different construction stages are different, from a macro perspective, factors such as building area, number of floors, height, structure type, exterior wall materials, decoration standards, construction technology, and equipment brands are generally recognized cost-sensitive attributes. These attributes can reflect the investment scale, technical complexity, functional requirements, etc. of construction projects from different angles and different granularities, and are the focus of engineering cost management. Therefore, the above-mentioned main influencing factors are extracted through qualitative analysis, which can provide theoretical guidance and business rule support for the identification of influencing factors of engineering integration information.

[0049] Then, under the guidance of expert knowledge, the influencing factors of the engineering integration information are automatically identified. Since the influencing factors of engineering projects exist in two forms, explicit and implicit, two identification strategies are adopted: rule-based and statistical-based. For explicit influencing factors such as building area and number of floors that have clear numerical descriptions in the engineering integration information, by formulating influencing factor identification rules, regular expressions, dictionary matching and other methods are used to automatically extract relevant attribute values ​​from structured data items. For example, from the structured description of "the total building area of ​​the project is 12,000 square meters, of which the above-ground building area is 10,000 square meters and the underground building area is 2,000 square meters", by matching keywords such as "total building area", "above-ground building area" and "underground building area", the influencing factor of building area can be identified and the corresponding numerical value can be extracted. For implicit influencing factors such as exterior wall materials and structural types that are not clearly defined in the engineering integration information but have semantic associations with other information, statistical methods such as word frequency statistics, co-occurrence analysis and association rule mining are used to discover potential influencing factors from descriptive texts. For example, if keywords such as "glass curtain wall", "stone", and "aluminum plate" frequently appear in the text describing the building facade design, and these keywords are highly co-occurring with concepts such as "exterior wall" and "facade", it can be inferred that the exterior wall of the project may use materials such as glass curtain wall, stone, and aluminum plate. Through explicit rule matching and implicit statistical analysis, engineering features that have a significant impact on construction costs can be automatically and efficiently identified from massive engineering integration information.

[0050] S502, designing a quantitative standard for each of the project influencing factors to obtain a quantitative standard for each of the project influencing factors, and extracting a characteristic value corresponding to each of the project influencing factors from the project integration information based on the quantitative standard; S503, performing feature representation on the feature value according to a preset feature representation format to obtain the feature representation of the engineering influencing factor.

[0051] Specifically, a quantitative standard is designed for each engineering influencing factor, and the quantitative standard of each engineering influencing factor is obtained. The process of extracting the corresponding characteristic value of each engineering influencing factor from the engineering integration information based on the quantitative standard is as follows: First, the data characteristics of each type of influencing factor are sorted out, and its semantics, data type, dimensional unit, etc. are systematically summarized. For example, the building area is usually a positive integer with square meters as the unit; the exterior wall material is usually composed of several types of building materials; the structure type is usually selected from several options such as frame structure, shear wall structure, frame-shear wall structure, etc. Identifying the data characteristics of the influencing factors is the basis for carrying out quantitative design. Then, considering the data characteristics and business semantics of the influencing factors, a targeted quantitative scheme is designed so that the extracted characteristic values ​​can reasonably reflect the cost attributes of the influencing factors. For numerical influencing factors, according to their dimensional units and numerical ranges, appropriate normalization functions are designed to map the original data to a unified scale. Commonly used normalization functions include maximum and minimum value normalization, Z-score normalization, etc.; for categorical influencing factors, One-Hot encoding or ordinal encoding is used to convert discrete category values ​​into equal-length binary codes or unique digital IDs to reflect the logical relationship between different categories; for textual influencing factors, unstructured text is embedded into the vector space by building domain dictionaries and corpora, and the semantic features of the text are represented by word frequency vectors, TF-IDF vectors, and topic vectors. When designing, full reference is made to industry standards and specifications to make the quantification scheme conform to the business characteristics of the engineering cost field. For example, for structural types, a quantification scheme based on the national standard "Architectural Structure Type and Foundation Design Code" can be designed to map frame structures, shear wall structures, frame-shear structures, etc. to unique digital codes. Through factor type-driven quantitative design, a unified, standardized, and computable feature representation can be obtained.

[0052] Finally, based on the quantitative standard, the characteristic value corresponding to each influencing factor is extracted from the engineering integration information. Since the engineering integration information contains a large amount of heterogeneous data, its feature extraction needs to consider the multi-source and diversity of the data. On the one hand, the influencing factors of different data sources differ in description granularity and semantic standardization, and differentiated extraction strategies need to be adopted. For data sources with a high degree of structure, the target field value can be directly obtained through field matching, regular parsing, etc.; for data sources with a low degree of structure, such as natural language text, it is necessary to first use word segmentation, semantic analysis and other technologies to extract key information, and then map it to a standardized feature form, such as extracting semantic features such as project scale and construction technology from the construction organization design description through technologies such as topic models and named entity recognition.

[0053] S105, weighted fusion of the characteristic expressions of the engineering influencing factors is performed through an attention mechanism to generate a representation vector of the project cost influencing factors; Specifically, first, the characteristic representation of the engineering influencing factors is input into the attention network. The attention network usually consists of three parts: query, key, and value, which correspond to the query vector, key vector, and value vector of the influencing factors respectively. Then, the query vector and key vector are used to calculate the correlation weights between the influencing factors. Specifically, methods such as dot product and cosine similarity can be used to calculate the similarity between the query vector and the key vector to obtain a weight matrix representing the correlation of the influencing factors. Next, the value vector is weighted and fused using the correlation weight. Specifically, the weight matrix can be multiplied by the value vector to obtain a weighted influencing factor representation vector. Finally, the representation vectors of each influencing factor are connected in series to form a project cost influencing factor representation vector of the entire project cost influencing factor.

[0054] Based on the above embodiment, as an optional embodiment, the feature representation of the engineering influencing factors is weightedly integrated through the attention mechanism to generate a representation vector of the project cost influencing factors, including: The characteristic expressions of the engineering influencing factors are weighted to obtain the weights of the characteristic expressions of the engineering influencing factors, and the characteristic expressions of the engineering influencing factors are weightedly fused based on the weights to obtain the representation vector of the project cost influencing factors.

[0055] Specifically, the process of obtaining the characteristic expression of the engineering influencing factors by expressing the characteristic values ​​according to the preset characteristic expression format is as follows: First, in combination with the characteristics of the engineering cost business, a characteristic expression format with unified form and complete semantics is designed. Since the factors affecting the engineering cost are complex and numerous, mathematical forms such as vectors or matrices are used to characterize the characteristic attributes of the engineering project. A typical characteristic expression format is a characteristic vector, where each dimension represents an influencing factor, the dimension value is the quantitative characteristic value of the corresponding influencing factor, and the vector as a whole represents the investment attribute portrait of the project. The mathematical form of the characteristic vector is concise and standardized, which is convenient for calculation and analysis, and is widely used in statistical learning, data mining and other fields. For example, the following characteristic vector format can be designed: {building area: 0.6, exterior wall material: [0, 1, 1, 1, 0], structure type: 1, number of floors: 0.8, exterior decoration standard: 2, ...}. The characteristic vector contains multiple influencing factors such as building area, exterior wall material, and structure type. The name of the influencing factor is before each colon, and the corresponding quantitative characteristic value is after the colon. The entire vector comprehensively characterizes the cost characteristics of the engineering project. Another feature representation format is the feature matrix, where each row represents a project, each column represents an influencing factor, and the intersection of rows and columns is the quantitative feature value of the influencing factor. The feature matrix is ​​suitable for representing the feature set of multiple projects, which is convenient for conducting inter-group statistical analysis. Regardless of the feature representation format used, it is necessary to systematically sort out the influencing factors and their feature values, and clarify the semantic connotation, numerical range, prior constraints and other meta-information of each feature dimension to ensure the standardization and completeness of the feature representation. Then, based on the preset format, the quantitative feature values ​​of each influencing factor are filled into the corresponding feature dimension to form a structured feature representation. Since the data types and numerical ranges of the influencing factors are different, appropriate data transformations are required during filling, such as numerical normalization, unique hot encoding, etc., so that the feature values ​​of different sources and scales are mapped to a unified coordinate system, eliminating the differences in numerical magnitudes, and meeting the constraints of the feature representation format. For numerical features, the original values ​​are mapped to the interval [0, 1] through a normalization function to eliminate the dimension effect; for categorical features, discrete categories are mapped to binary vectors of equal length through one-hot encoding, and each category corresponds to a unique code; for text features, text is mapped to a semantic vector of fixed length through word embedding, topic model and other methods. During the transformation, the dimension order, data type and other requirements of the feature representation format are followed to ensure the consistency of the feature values ​​of each dimension. After the filling is completed, the normalized influencing factor features can be obtained. For example, the feature vector of a certain engineering project can be expressed as (0.6, 0, 1, 1, 1, 0, 1, 0.8, 2, ...), where each value corresponds to the quantized value of a feature dimension in turn. The entire vector vividly depicts the cost characteristics of the engineering project and is the digital input for cost estimation.

[0056] Finally, the influencing factor features of each project are serialized and stored to obtain a standardized and structured engineering cost feature library. Using unified storage media such as databases, data warehouses, and file systems, a unique feature file is established for each project. The file content is the feature vector or feature matrix of the project, and a mapping association is established with the original engineering information of the project to form a semantically rich and traceable engineering big data resource. Through the standardized storage of influencing factor features, the structured representation of engineering cost knowledge is realized, and the efficient retrieval and integrated calculation of engineering cost data are supported, providing solid data support for data analysis applications in the field of engineering cost. For example, a project retrieval engine based on feature vectors can be designed. The user enters the target feature conditions, the system parses the conditions to generate structured queries, and performs similarity matching with the feature vectors in the feature library, quickly retrieving similar engineering projects with target feature attributes, and helping engineering cost professionals to quickly find reference cases. For another example, the engineering cost feature vectors of different professions and different owners can be integrated in a unified format, and data warehouses, federated learning and other technologies can be used to realize cross-domain calculation of feature data, breaking through the limitations of data islands and realizing the integrated application of industry big data.

[0057] S106, inputting the engineering information into a preset engineering cost standard model, and outputting the engineering cost standard; Specifically, firstly, a standard model for engineering cost is trained and established based on the engineering information and actual cost data of a large number of completed projects. This model can use machine learning algorithms such as multivariate linear regression, random forest, neural network, etc., and establish an estimation function from engineering attribute features to cost standards by learning the mapping relationship between engineering information and actual cost. During the model training process, it is necessary to perform feature engineering on the engineering information to extract key features that have a greater impact on the cost, such as building area, structure type, geographical location, etc. At the same time, feature screening and optimization are also required to remove redundant and noise features and improve the generalization performance of the model.

[0058] After obtaining the trained engineering cost standard model, the engineering information of the engineering project to be estimated is input into the model to quickly estimate the cost standard of the project. Specifically, the input engineering information is first preprocessed to extract the feature representation consistent with the model training. Then, the feature representation is input into the cost standard model, and the estimated cost standard value of the project is obtained through the feature combination and weight calculation within the model. Finally, the estimated cost standard is combined with the influencing factor representation vector of the project to form a complete project cost description information for subsequent cost optimization and management.

[0059] Based on the above embodiment, as an optional embodiment, the feature input matrix is ​​constructed based on the project cost influencing factor characterization vector, the engineering information and the engineering cost standard, including: S601, integrating the project cost influencing factor representation vector, the engineering information feature and the engineering cost standard to form a unified feature input matrix; Specifically, the process of constructing a feature input matrix based on the representation vector of project cost influencing factors, engineering information and engineering cost standards is as follows: First, the representation vector of project cost influencing factors, engineering information features and engineering cost standards are integrated to form a unified feature input matrix. Although these three types of data come from different sources and in various forms, they all contain feature information related to engineering cost in semantics. In order to explore the inherent connection of these feature information and support the correlation analysis of engineering cost, it is necessary to achieve seamless integration in a unified data space. A direct integration method is to splice the three types of data into a large matrix, the rows of the matrix represent different engineering projects, the columns represent various features, and the intersection of the rows and columns is the feature value of the corresponding project. For example, it can be spliced ​​in the order of [engineering category, building area, number of floors, structure type, decoration standard, cost index], where the first four items are the influencing factor features, the middle item is the engineering information feature, and the last item is the cost standard feature. The entire sequence represents the investment attribute portrait of the engineering project. In particular, for the representation vector of the influencing factors, since it is already a normalized numerical sequence, it can be directly filled into the corresponding feature column. For engineering information and cost standard features, certain data transformations may be required depending on their original forms. For example, for text-based engineering information, it can be represented as a fixed-length topic vector through a topic model; for engineering information such as CAD drawings, it can be represented as a structured feature through image feature extraction methods; for cost indicators, it can be mapped to a fixed numerical range through operations such as interpolation and normalization. During the transformation, the constraints of the matrix structure are followed to ensure the consistency of the dimension and numerical range. Through the above data transformation and alignment process, the project cost influencing factor representation vector, engineering information features, and engineering cost standards can be seamlessly spliced ​​into a unified feature input matrix.

[0060] After completing the feature integration of multi-source heterogeneous data, a feature input matrix with comprehensive content and standardized structure is obtained. From a vertical perspective, each row of the matrix is ​​a complete feature description of a project, covering multi-dimensional information such as influencing factors, project ontology, and cost standards, and comprehensively describing the investment attributes of the project; from a horizontal perspective, each column of the matrix is ​​a sequence of values ​​of a specific feature on different projects, reflecting the distribution characteristics of the feature. This matrix-style feature organization not only takes into account the integrity of information, but also facilitates data correlation analysis. For example, the correlation analysis method can be used to reveal the degree of correlation between various features and cost indicators, and identify key features that have a significant impact on project costs; clustering algorithms can be used to discover typical patterns of project costs and summarize the cost distribution laws under different feature combinations; machine learning methods such as classification and regression can be used to establish a mapping model between feature input matrices and project costs, and use it for cost prediction and deviation analysis of new projects. By integrating the multi-dimensional feature information of engineering projects in the form of a matrix, the application value of engineering data can be maximized, the breadth and depth of analysis of engineering cost business can be expanded, and comprehensive and high-quality data input can be provided for the engineering cost prediction model, laying a solid data foundation for engineering cost management.

[0061] S602: construct the feature input matrix based on the feature input matrix.

[0062] Specifically, the process of constructing the feature input matrix based on the feature input matrix is ​​as follows: First, feature engineering is performed on the feature input matrix to screen out the key features with the largest amount of information and the strongest discrimination. Common feature selection methods include filtering method, wrapping method, embedding method, etc. The filtering method mainly adopts the feature scoring criterion to independently evaluate the importance of each feature, and then selects several features with the highest score to form a new feature subset. The scoring criterion can use statistical indicators, such as Pearson correlation coefficient, mutual information, etc., to measure the correlation between each feature and the project cost from a statistical perspective, which is used to characterize the importance of the feature. Taking mutual information as an example, mutual information measures the interdependence between feature variables and target variables. The larger the value, the stronger the ability of the feature to distinguish different levels of project costs, and it should be included in the new feature matrix first. The wrapping method regards feature selection as a feature subset search problem, and automatically searches for the optimal feature subset through the optimization algorithm so that the subset can obtain the best project cost prediction performance. If a genetic algorithm is used, each individual is encoded as a feature selection scheme, and the population is continuously optimized through evolutionary iteration, and finally the global optimal feature combination is searched. The embedding method combines feature selection with the training process of the engineering cost prediction model, and dynamically adjusts the features included in the model based on the evaluation indicators of the model output, such as feature coefficients and feature importance. For example, the linear regression model with L1 regularization automatically compresses the weights of redundant features through the sparsification effect of the coefficients, and the features that finally work are the key features selected. Through feature selection, the original high-dimensional feature matrix is ​​condensed into the most representative low-dimensional feature matrix. Each feature in the matrix is ​​an indispensable influencing factor in the process of engineering cost formation, carrying the most essential and richest cost information.

[0063] Then, the necessary feature transformation is carried out on the selected key features, and the numerical representation of the features is optimized to make them more standardized and normalized. Common transformation operations include data normalization, data standardization, logarithmic transformation, etc. Data normalization maps the eigenvalue to the interval [0, 1] to eliminate the imbalance caused by the difference in the dimensions of different features. The calculation formula is: x'= (x - min) / (max - min), where x is the original eigenvalue, x' is the normalized eigenvalue, and max and min are the maximum and minimum values ​​of the feature on all samples, respectively. Data standardization centers the original eigenvalue and normalizes the variance, so that the eigenvalue obeys a standard normal distribution with a mean of 0 and a variance of 1. The calculation formula is: x'= (x - μ) / σ, where μ is the mean of the feature and σ is the standard deviation of the feature. Logarithmic transformation takes the logarithm of the original eigenvalue, which can transform the skewed distribution into a normal distribution and also widen the gap between eigenvalues ​​of different orders of magnitude. The calculation formula is: x'= log(x). The transformed feature matrix is ​​easier to calculate and has more stable values, which is conducive to the training and solution of the project cost prediction model. For example, when calculating the similarity between samples, normalization and standardization can ensure the balanced contribution of different features; when using gradient optimization algorithms to train models, standardization of feature values ​​can help accelerate the convergence of loss functions.

[0064] S107, constructing a project cost model based on the project cost influencing factor representation vector, the project information and the project cost standard, and inputting the project information into the project cost model to obtain a project cost result.

[0065] Specifically, the construction process of the engineering cost model is as follows: First, the representation vector of the factors affecting the project cost is used as the main feature input of the model. This representation vector comprehensively depicts the various influencing factors of the project, and through the weighted fusion of the attention mechanism, it effectively captures the interactive effects between the factors, and is the key information carrier for engineering cost estimation. Secondly, the engineering cost standard is used as the reference input of the model. The cost standard reflects the reasonable cost level of the engineering project under normal conditions, which is of great significance for the calibration and evaluation of the estimation results. Finally, combined with other attribute features in the engineering information, such as engineering quantity, progress, risk, etc., a multi-dimensional feature space of influencing factors, cost standards, and engineering attributes is jointly constructed. In this feature space, high-performance learning models such as deep neural networks and random forests are used to establish an end-to-end mapping function from features to costs, forming a complete engineering cost estimation model.

[0066] In the process of training the engineering cost model, a large number of historical engineering cases and actual cost data are needed. By inputting the feature representation of the engineering case into the model, the cost result is estimated, and compared with the actual cost, the estimated loss is calculated, thereby guiding the iterative optimization of the model parameters. At the same time, strategies such as cross-validation and ensemble learning can also be used to improve the generalization performance and robustness of the model. In the process of model training, domain knowledge and expert experience can also be introduced to carry out targeted design of the model structure and loss function to make it more in line with the business characteristics of engineering cost.

[0067] On the other hand, the present invention also provides a system for estimating construction cost. Figure 2 The system comprises: The knowledge graph construction module 1 is used to obtain historical engineering cost sample data and construct an engineering cost knowledge graph based on the historical engineering cost sample data; Preprocessing module 2, used to obtain engineering information of the project to be estimated, input the engineering information into the engineering cost knowledge graph, and output a structured representation of the project to be estimated; The association pattern determination module 3 is used to map the structured expression into the semantic representation in the engineering cost knowledge graph to obtain the association pattern of the engineering project to be estimated; An integration and extraction module 4 is used to manage and integrate the engineering information based on the association pattern to obtain engineering integration information, and to extract features from the engineering integration information to obtain a feature description of engineering influencing factors; An attention mechanism module 5 is used to perform weighted fusion on the characteristic representation of the engineering influencing factors through an attention mechanism to generate a representation vector of the project cost influencing factors; The standard selection module 6 is used to input the engineering information into a preset engineering cost standard model and output the engineering cost standard; The cost module 7 is used to construct a project cost model based on the project cost influencing factor characterization vector, the project information and the project cost standard, and input the project information into the project cost model to obtain the project cost result.

[0068] Please refer to Figure 3 The application also discloses an electronic device. Figure 3 The electronic device 300 may include: at least one processor 301 , at least one network interface 304 , a user interface 303 , a memory 305 , and at least one communication bus 302 .

[0069] The communication bus 302 is used to realize the connection and communication between these components.

[0070] The user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0071] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0072] Among them, the processor 301 may include one or more processing cores. The processor 301 uses various interfaces and lines to connect various parts in the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory 305, and calling data stored in the memory 305. Optionally, the processor 301 can be implemented in at least one hardware form of digital signal processing (Digital Signal Processing, DSP), field programmable gate array (Field~Programmable Gate Array, FPGA), and programmable logic array (Programmable Logic Array, PLA). The processor 301 can integrate one or a combination of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU) and a modem. Among them, the CPU mainly processes the operating system, user interface and application programs; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 301, and it can be implemented separately through a chip.

[0073] Among them, the memory 305 may include a random access memory (Random Access Memory, RAM) and may also include a read-only memory (Read~Only Memory). Optionally, the memory 305 includes a non-transitory computer-readable medium. The memory 305 can be used to store instructions, programs, codes, code sets or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as a touch function, a sound playback function, an image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data involved in the above-mentioned various method embodiments, etc. The memory 305 may optionally be at least one storage system located away from the aforementioned processor 301. Refer to Figure 3The memory 305 as a computer storage medium may include an operating system, a network communication module, a user interface module, and an application program for a flocculant addition analysis method.

[0074] exist Figure 3 In the electronic device 300 shown, the user interface 303 is mainly used to provide an input interface for the user and obtain the data input by the user; and the processor 301 can be used to call the application program of the engineering cost estimation method stored in the memory 305. When executed by one or more processors 301, the electronic device 300 executes one or more methods in the above-mentioned embodiments. It should be noted that for the aforementioned method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present application is not limited to the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required for the present application. In the above embodiments, the descriptions of each embodiment have their own emphasis. For the parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0075] In the several implementations provided in this application, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are only schematic, such as the division of units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some service interfaces, and the indirect coupling or communication connection of the system or unit can be electrical or other forms.

[0076] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0077] The present application also provides a computer storage medium that can store multiple instructions, which are suitable for being loaded and executed by a processor as described above. Figure 1 The road assessment method of the embodiment shown in the figure can be found in the specific implementation process. Figure 1 The specific description of the illustrated embodiment will not be repeated here.

[0078] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0079] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a memory, including a number of instructions for a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned memory includes: various media that can store program codes, such as USB flash drives, mobile hard drives, magnetic disks or optical disks.

[0080] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0081] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for estimating construction cost, characterized in that: The method comprises: Acquire historical engineering cost sample data, and construct an engineering cost knowledge graph based on the historical engineering cost sample data; Acquire engineering information of the project to be estimated, input the engineering information into the engineering cost knowledge graph, and output a structured representation of the project to be estimated; Mapping the structured expression to the semantic representation in the engineering cost knowledge graph to obtain the association pattern of the engineering project to be estimated; Based on the association pattern, the engineering information is managed and integrated to obtain engineering integration information, and features are extracted from the engineering integration information to obtain feature descriptions of engineering influencing factors; The characteristic expressions of the engineering influencing factors are weighted and fused through the attention mechanism to generate a representation vector of the project cost influencing factors; Inputting the engineering information into a preset engineering cost standard model and outputting the engineering cost standard; A project cost model is constructed based on the project cost influencing factor characterization vector, the project information and the project cost standard, and the project information is input into the project cost model to obtain a project cost result.

2. The method according to claim 1, characterized in that The step of constructing a construction cost knowledge graph based on the historical construction cost sample data includes: Preprocessing the historical engineering cost sample data to obtain preprocessed historical engineering cost sample data; Extracting relational entities related to construction cost from the preprocessed historical construction cost sample data; Performing fusion processing on the relationship entities to build a consistent entity relationship pedigree; A triple entity relationship representation is constructed based on the entity relationship pedigree, and the engineering cost knowledge graph is constructed based on the triple entity relationship representation.

3. The method according to claim 1, characterized in that The mapping of the structured expression to the semantic representation in the engineering cost knowledge graph to obtain the association pattern of the engineering project to be estimated includes: Parsing the structured expression into an entity relationship representation, and matching the entity relationship representation with entities and relationships in the engineering cost knowledge graph to obtain an association relationship entity representation; Based on the association relationship entity representation, pattern mining is performed in the engineering cost knowledge graph to obtain the association pattern of the engineering project to be estimated.

4. The method according to claim 1, characterized in that The step of managing and integrating the engineering information based on the association mode to obtain engineering integration information includes: Converting the engineering information into structured engineering information, and matching the structured engineering information with the association pattern to obtain adapted engineering information; The adaptation project information is optimized and completed to obtain target adaptation project information, and the target adaptation project information is integrated according to a preset format to obtain the project integration information.

5. The method according to claim 1, characterized in that The feature extraction of the engineering integration information to obtain the feature description of the engineering influencing factors includes: Identifying influencing factors of the engineering integration information according to preset expert knowledge to obtain engineering influencing factors; Designing a quantitative standard for each of the engineering influencing factors to obtain a quantitative standard for each of the engineering influencing factors, and extracting a characteristic value corresponding to each of the engineering influencing factors from the engineering integration information based on the quantitative standard; The characteristic value is characterized according to a preset characteristic representation format to obtain the characteristic representation of the engineering influencing factor.

6. The method according to claim 1, characterized in that The weighted fusion of the characteristic expressions of the engineering influencing factors is performed through the attention mechanism to generate a representation vector of the project cost influencing factors, including: The characteristic expressions of the engineering influencing factors are weighted to obtain the weights of the characteristic expressions of the engineering influencing factors, and the characteristic expressions of the engineering influencing factors are weightedly fused based on the weights to obtain the representation vector of the project cost influencing factors.

7. The method according to claim 1, characterized in that The constructing of a feature input matrix based on the project cost influencing factor characterization vector, the engineering information and the engineering cost standard includes: Integrate the project cost influencing factor representation vector, the engineering information feature and the engineering cost standard to form a unified feature input matrix; The feature input matrix is ​​constructed based on the feature input matrix.

8. A construction cost estimation system, characterized in that: include: A knowledge graph construction module, used to obtain historical engineering cost sample data and construct an engineering cost knowledge graph based on the historical engineering cost sample data; A preprocessing module, used to obtain engineering information of the project to be estimated, input the engineering information into the engineering cost knowledge graph, and output a structured representation of the project to be estimated; An association pattern determination module is used to map the structured expression into the semantic representation in the engineering cost knowledge graph to obtain an association pattern of the engineering project to be estimated; An integration and extraction module, used to manage and integrate the engineering information based on the association pattern to obtain engineering integration information, and to extract features from the engineering integration information to obtain a feature description of engineering influencing factors; An attention mechanism module, used for weighted fusion of the characteristic expressions of the engineering influencing factors through the attention mechanism to generate a representation vector of the project cost influencing factors; A standard selection module, used to input the engineering information into a preset engineering cost standard model and output the engineering cost standard; The cost module is used to construct a project cost model based on the project cost influencing factor representation vector, the project information and the project cost standard, and input the project information into the project cost model to obtain the project cost result.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes a processor, a memory and a transceiver, the memory is used to store instructions, the transceiver is used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method according to any one of claims 1 to 7.

Images