Larix artificial forest land quality evaluation knowledge graph construction and service system

Abstract

The invention discloses a knowledge graph construction and service system for artificial larch forest land quality evaluation. The system is composed of a mode layer, a data layer and a knowledge service module. The mode layer and the data layer realize interaction through a bidirectional constraint verification mechanism, cooperatively complete construction of a knowledge graph and provide data support for the knowledge service module; the model layer is based on industry standards, resource systems and expert knowledge to construct a larch site quality evaluation ontology framework, the entity class takes evaluation indexes and site factors as a core top class, and the semantic relationship covers relationships such as hierarchical structures and causal effects. The invention relates to the crossing field of forestry science and information technology. According to the system, term unification and format standardization of multi-source heterogeneous data are achieved through the semantic alignment module, the consistency of 1424 entity nodes and 3152 semantic relationships in the knowledge graph is ensured, and the problems that traditional multi-source data are difficult to integrate and prone to conflicts are solved.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of forestry science and information technology, specifically to a knowledge graph construction and service system for evaluating the site quality of larch plantations. Background Technology

[0002] In the field of larch plantation site quality evaluation, existing methods have significant limitations and are insufficient to meet the practical needs of refined and intelligent forestry management. The core issues are concentrated in four aspects: First, the knowledge system is fragmented, relying on scattered industry standards, scientific research literature and expert experience, lacking a unified structured framework. The conceptual boundaries between site factors and evaluation indicators are blurred and the semantic relationship is unclear. Moreover, the evaluation indicators are mostly limited to 15-22 core items, failing to cover multi-dimensional factors such as climate and biology, resulting in one-sided evaluation results. Second, the data processing efficiency is low. Unstructured text knowledge relies on manual extraction, which is cumbersome. Furthermore, the terminology and format differences among multi-source heterogeneous data (standard documents, scientific papers, field survey data, etc.) are inconsistent, making it difficult to achieve efficient integration. Third, the service application capabilities are insufficient. Information retrieval still relies on keyword matching, making it difficult to achieve multi-dimensional related retrieval. The professional terminology has a high threshold for understanding, making it difficult for non-professional users to use. Factor analysis mainly stays at the statistical level and lacks in-depth decision support. Fourth, the results are unreliable, relying only on a single dataset for verification and failing to cover multiple dimensions such as geographical distribution and environmental factors, resulting in insufficient credibility of the conclusions and affecting the scientific nature and accuracy of forestry production decisions. Summary of the Invention

[0003] To solve the above technical problems, the present invention is implemented through the following technical solution: a knowledge graph construction and service system for larch plantation site quality evaluation, comprising a model layer, a data layer and a knowledge service module, wherein the model layer and the data layer interact with each other through a two-way constraint verification mechanism to jointly complete the construction of the knowledge graph and provide support for knowledge services; The model layer, based on industry standards, resource systems, and expert knowledge, employs Stanford University's seven-step ontology engineering method to construct a larch site quality evaluation ontology framework. This framework includes 195 entity classes, 13 semantic relationship types, and 35 constraint rules. The entity classes are centered around evaluation indicators and site factors, covering four categories of factors: climate, topography, soil, and biology. Semantic relationships encompass hierarchical structures, causal interactions, and other related relationships. The data layer includes modules for data collection, data processing, corpus construction, large language model training, semantic alignment, and storage. The data collection module is used to collect multi-source heterogeneous data. The large language model training module is based on the DeepSeek-Chat model and uses a three-stage progressive prompting engineering to achieve knowledge extraction. The storage module uses the Neo4j graph database to construct a "node-edge" structure for knowledge storage and management. The knowledge service module provides intelligent query, correlation analysis and intelligent question answering functions, and the system performs cross-validation based on three independent datasets, with the accuracy of geographical distribution information reaching over 96%.

[0004] Preferably, the ontology framework of the pattern layer constructs 277 relation instances, and the constraint rules cover logical integrity, semantic consistency, professional accuracy and scientific rationality. The entity classes include four site factor subclasses: climate, terrain, soil and biology.

[0005] Preferably, the multi-source heterogeneous data collected by the data collection module includes national standards and specifications, industry technical documents, academic papers, research reports and field survey data, and the data processing module achieves data consistency verification through data cleaning and quality control.

[0006] Preferably, the three-stage progressive prompting engineering includes requirement training, ontology training, and example training. This strategy can automatically extract entity and relation information from unstructured text, improving extraction efficiency by more than 40% compared to traditional methods.

[0007] Preferably, the semantic alignment module processes the original triples through entity alignment and relation standardization to eliminate terminological ambiguity and format differences between different data sources, thereby ensuring the consistency of the knowledge graph.

[0008] Preferably, the knowledge graph contains 1,424 knowledge entity nodes and 3,152 semantic relationships, and integrates 69 multi-dimensional evaluation indicators, achieving comprehensive coverage of the indicator system compared to the traditional 15-22 core indicators.

[0009] Preferably, the intelligent query function supports entity query, relationship query and compound query, and uses semantic association retrieval technology to achieve fast response, with a query response time of no more than 0.85 seconds and a query accuracy of no less than 89%.

[0010] Preferably, the correlation analysis function is based on graph algorithms to mine complex correlations between site factors, construct a multi-level factor network, realize the evaluation of larch growth suitability, and the accuracy of the analysis results reaches more than 85%.

[0011] Preferably, the cross-validation is based on the larch distribution validation dataset, the site information validation dataset, and the site index validation dataset, with an environmental factor numerical range compliance rate of not less than 84% and an SI site quality interval compliance rate of not less than 88%.

[0012] Preferably, the intelligent question-answering function adopts a hybrid question-answering strategy that prioritizes knowledge graphs. It can receive natural language queries and match them with knowledge graphs, achieving a question-answering accuracy rate of over 87%, and supporting non-professional users to efficiently acquire professional knowledge.

[0013] This invention provides a knowledge graph construction and service system for evaluating the site quality of larch plantations. It has the following beneficial effects: (I) The knowledge graph construction and service system for site quality evaluation of larch plantations is constructed based on the seven-step ontology engineering method of Stanford University. It combines 195 entity classes (including four site factor subclasses such as climate and topography), 13 semantic relations and 35 constraint rules to form a structured system of scattered evaluation knowledge, avoiding the problems of "fuzzy concepts and chaotic relations" in traditional methods.

[0014] (II) The knowledge graph construction and service system for larch plantation site quality assessment eliminates terminological ambiguity and format differences in multi-source heterogeneous data (standard specifications, scientific papers, field data, etc.) through the semantic alignment module (entity alignment, relation standardization), and combines the "node-edge" storage structure of Neo4j graph database to ensure the consistency of 1424 entity nodes and 3152 semantic relations in the knowledge graph, effectively solving the problem of "difficult integration and easy conflict" of traditional multi-source data.

[0015] (III) The knowledge graph construction and service system for larch plantation site quality assessment is cross-validated based on three independent datasets (distribution verification, site information verification, and site index verification). The accuracy of geographical distribution information reaches over 96%, the environmental factor numerical range compliance rate is no less than 84%, and the SI site quality interval compliance rate is no less than 88%. The system's output results are verified from multiple dimensions to demonstrate their scientific validity and reliability, overcoming the shortcomings of traditional methods that are "supported by single data and have low credibility of conclusions".

[0016] (iv) The larch plantation site quality evaluation knowledge graph construction and service system realizes the full-process integration of "modeling-processing-service-verification", which can directly support site quality classification, suitability planting planning and growth potential assessment in forestry production, and provide technical support for the efficient utilization and sustainable management of larch plantation resources, filling the gap in the professional knowledge graph service system in this field. Detailed Implementation

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention provides a technical solution: I. Environmental preparation for implementation; Hardware environment: Server configuration: Two Huawei TaiShan 2280 servers (CPU: Kunpeng 920 64 cores, memory: 256GB DDR4, hard disk: 4TB SSD×4) are used for knowledge graph construction (including data processing and model training) and knowledge service deployment, respectively. Terminal equipment: The equipment used for field surveys includes a DJI Mavic 3 drone (for terrain data acquisition), a Trimble R10 GNSS receiver (positioning accuracy ≤1m), and a soil nutrient rapid tester (detection indicators: organic matter, nitrogen, phosphorus and potassium content).

[0019] Software environment: Operating system: The server uses CentOS 8.5, and the terminal survey devices use Android 13; Core software: Neo4j 5.12 (community edition, supporting storage of 1 billion nodes), Python 3.9 (with Pandas 1.5.3, PyTorch 2.0.1, and Jieba 0.42.1 word segmentation library), DeepSeek-Chat 1.3 (open source large language model, with a parameter scale of 7B), Apache JMeter 5.6 (performance testing tool), and Gephi 0.9.7 (graph visualization and association analysis tool).

[0020] II. Schema Layer: Ontology framework construction and implementation; The ontology construction work was carried out in stages, adopting the seven-step ontology engineering method of Stanford University, combined with forestry industry standards (such as GB / T15776-2016 "Technical Regulations for Afforestation"), "Technical Regulations for the Cultivation of Larch Plantations", and the knowledge of three forestry experts with more than 10 years of experience in site quality assessment research: 1. Requirements Analysis Phase; A special team consisting of two ontology engineers and three forestry experts was formed. Through three rounds of seminars, the core requirements for ontology construction were clarified: to cover the entire process of larch site quality evaluation, support the semantic association of "factor-indicator-evaluation results", and meet the knowledge query and analysis needs of scientific research and production. The "Larch Site Quality Evaluation Ontology Requirements Document" was developed, which clarifies the core scenarios that the ontology needs to cover: including site factor retrieval, evaluation index correlation analysis, and growth suitability prediction. 2. Conceptualization and formalization stage; The core concepts were extracted: "Evaluation Indicators" and "Site Factors" were used as the top-level entity classes, and then subcategories were formed. "Site Factors" included four subcategories: climate factors (such as average annual temperature and annual precipitation), topographic factors (such as slope and altitude), soil factors (such as soil thickness and organic matter content), and biological factors (such as forest cover and microbial abundance). "Evaluation Indicators" included subcategories such as site index (SI), growth suitability level, and biomass. Finally, 195 entity classes were determined (see Appendix 1 for the complete list). 3. Define semantic relations: A total of 13 semantic relation types are defined, including: Hierarchical structure relationships: for example, "site factors - include - climate factors" and "site index - belong to - evaluation indicators"; Causal relationships: for example, "average annual temperature - influences - suitability for larch growth" and "soil thickness - determines - site index"; Spatial distribution relationships: for example, "Chongli District - located in - Zhangjiakou City" and "North China larch - distributed in - Chongli District"; Based on the above types, a total of 277 relationship instances were constructed, such as "Annual precipitation (climate factor) - impact - larch growth suitability (evaluation index)" and "Altitude (topographic factor) - value range - 300~1800m".

[0021] 4. Definition and verification of constraint rules; 35 constraint rules were defined, with specific examples as follows: Logical integrity rule: "Each evaluation indicator must be associated with at least one site factor, and each site factor must correspond to two or more evaluation indicators." Semantic consistency rule: "Terms such as 'average annual temperature' and 'annual average temperature' are uniformly mapped to the entity class 'average annual temperature', and synonymous entities are not allowed to coexist." Professional accuracy rule: "The soil thickness evaluation threshold must comply with the GB / T15776-2016 standard, that is, the soil thickness suitable for larch growth is not less than 60cm." The rules were validated using Protégé 5.6, and violations (such as "evaluation metrics for unrelated site factors") were detected using SPARQL queries until the violation rate was 0.

[0022] III. Data Layer: Knowledge Acquisition and Storage Implementation; 1. Collection of multi-source heterogeneous data; Structured data: Downloaded 12 national standards / industry documents such as GB / T 15776-2016 and LY / T 2241-2014 "Technical Regulations for Forest Site Quality Evaluation" from the official website of the National Forestry and Grassland Administration, and obtained the larch plantation plot data in Northeast and North China from 2018 to 2023 from the China Forest Resources Database; Unstructured data: Obtained 320 academic papers related to "larch site quality" from 2010 to 2023 (210 in Chinese and 110 in English) through the CNKI and Web of Science platforms; Semi-structured data: Conducted on-site data collection from July to September 2023 using an Android terminal APP, covering 6 main larch production areas such as Yichun in Heilongjiang and Changbai Mountain in Jilin, and established a total of 500 plots. The collected data content includes GPS coordinates, soil nutrient test results, and images of forest tree growth, etc.

[0023] 2. Data processing and corpus construction; Data cleaning: Processed the structured data using the Python Pandas library, excluded plot records with a missing rate exceeding 30%, a total of 128 were excluded, and for abnormal data, such as altitudes greater than 3000m, a combination of "mean replacement + expert review" was used for correction; Corpus processing: Preprocessed unstructured texts (papers, reports). First, used the Jieba word segmentation library for Chinese word segmentation (for example, segmented "The site index of larch plantations is affected by soil thickness" into "larch plantation / site index / soil thickness / affect"); then used the NLTK library to remove stop words (such as "of", "in", etc.). Finally, a domain corpus containing approximately 1.2 million word segmentation results was formed.

[0024] 3. Large language model training (three-stage progressive prompting engineering); Based on the DeepSeek-Chat large language model, training was carried out under the PyTorch framework. The model was iterated 200 rounds, and the learning rate was set to 1×10 -5 The implementation details of each stage are as follows: Stage 1: Requirement training (rounds 1-50); The prompting template used was: "Extract entities related to larch site quality evaluation (including site factors, evaluation indicators, indicator values) from the following text, in the format of [Entity type: Entity name]: {text segment}"; The input data was 20,000 text segments randomly selected from the corpus (each 50 to 100 words), and 1000 were manually marked as training samples. Stage 2: Ontology training (rounds 51-120); The prompt template is as follows: "Based on the larch site quality assessment ontology framework, match the extracted entities to the specified entity classes (e.g., 'slope' matches 'topography factor'), and identify the relationships between entities (e.g., 'influence', 'containment'), in the format of [Entity 1 - Relationship - Entity 2]: {List of extracted entities}". The input data consists of 8,000 entity results output from Phase 1, which are then used for matching training in conjunction with the entity class list of the ontology framework.

[0025] Phase 3: Example Training (rounds 121-200); The training example is as follows: "Text: 'In areas with an average annual precipitation of 1000mm, the larch site index can reach 22m' → Extraction result: [Climate factor: average annual precipitation, index value: 1000mm, evaluation index: site index], relationship: [average annual precipitation - impact - site index]".

[0026] The trained model achieved an entity extraction accuracy of 92% and a relation extraction accuracy of 88%, which is about 40% higher than the traditional CRF algorithm (entity accuracy 78%, relation accuracy 72%).

[0027] 4. Semantic alignment and storage; Semantic alignment: Entity alignment: The edit distance algorithm (threshold 0.8) is used to unify synonymous entities. For example, "average annual temperature" and "average annual temperature" are aligned to "average annual temperature", and "soil organic matter content" and "organic matter content" are aligned to "organic matter content". Relationship standardization: unify semantically similar relationships into a standard form, for example, unify "act on", "influence", and "association" into "influence"; 5. Atlas storage: In Neo4j, create node labels: “Site Factor”, “Evaluation Indicator”, “Indicator Value”, and “Evaluation Result”. Example node: (n:Site Factor{Name:"Average Annual Temperature",Unit:"℃",Value Range:"-5~15℃"}); Creating a relation edge: Example relation (a: location factor {name: "average annual temperature"}) - [r: influence {influence level: "significant"}] -> (b: evaluation index {name: "location index"}); The final knowledge graph contains 1,424 entity nodes, 3,152 semantic relationships, and integrates 69 multi-dimensional evaluation indicators, achieving comprehensive coverage of the traditional 15–22 core indicators.

[0028] IV. Implementation of the Knowledge Service Module; 1. Intelligent query function; This module provides intelligent query functionality, supporting three types of queries: entity query, relationship query, and composite query. It employs "semantic parsing + Cypher statement automatic generation" technology to achieve seamless integration between natural language input and graph query. Implementation details: Entity Query: When a user enters "query all terrain factors", the system first performs semantic parsing and then automatically generates a query statement: MATCH(n:site factor{subclass:"terrain factor"})RETURN n.name. After execution, the returned results include terrain factor entities such as "slope, elevation, aspect, and slope position".

[0029] Relationship query: When a user inputs "which site factors affect the site index", the system generates the statement: MATCH(a: site factor)-[r: influence]->(b: evaluation index {name: "site index"})RETURN a.name. The query results include key site factors such as "average annual temperature, soil thickness, and annual precipitation".

[0030] Compound query: The user inputs "query the site index range corresponding to the area with an altitude of 300-800m and a soil thickness ≥60cm", and generates Cypher: MATCH(a: site factor {name: "altitude", value range: "300~800m"})-[r1: impact]->(b: evaluation index {name: "site index"}),(c: site factor {name: "soil thickness", value range: ">=60cm"})-[r2: impact]->(b)RETURN b. value range; Performance testing: Apache JMeter was used to simulate 100 concurrent users for query testing. The average response time was 0.62 seconds (≤0.85 seconds). 1000 query results were randomly selected for manual verification, and the accuracy rate was 91.2% (≥89%).

[0031] 2. Association analysis function; This module uses the Louvain community detection algorithm (deployed on the Gephi platform) to mine the complex relationships between site factors and construct a multi-level association network of "site factors - evaluation indicators - growth suitability".

[0032] Implementation steps: Data import: Export the graph data (containing 1424 nodes and 3152 edges) from Neo4j to CSV format and import it into Gephi for network analysis; Network construction: Taking "growth suitability" as the core node, "evaluation indicators" (such as site index and biomass) are linked layer by layer, and then "site factors" (such as climate, topography and soil factors) are linked to form a three-layer network structure; Suitability assessment: The analytic hierarchy process (AHP) was used to determine the weights of the factors (annual mean temperature weight 0.25, soil thickness weight 0.22, annual precipitation weight 0.18) and a scoring model was constructed: growth suitability score = Σ(factor standardized value × weight), a score ≥80 is "suitable", 60-79 is "relatively suitable", and <60 is "unsuitable". Effect verification: 500 field survey plots (with known actual growth conditions) were selected, and the model evaluation results showed a consistency of 87.3% (≥85%) with the actual situation.

[0033] 3. Intelligent question-and-answer function; Hybrid Question Answering Strategy: A hybrid question answering strategy of "knowledge graph first" is adopted - first, the natural language question is parsed into a graph query. If the query has no results, the DeepSeek-Chat large language model is called to generate an answer. Implementation example: User question: "How can a non-professional determine whether a plot of land is suitable for planting larch?" Semantic analysis: Extract the core entities "plot", "larch", and "suitable", and map them to "site conditions", "larch", and "growth suitability" in the graph; Cypher lookup: MATCH(a:site conditions)-[r:influence]->(b:evaluation result{name:"suitability for larch growth"})RETURN a.key factors, b.judgment criteria; Answer generated: Returns "Key factors: average annual temperature 5-10℃, soil thickness ≥60cm, slope ≤25°; Judgment criteria: meeting 3 of the criteria is suitable, meeting 2 of the criteria is relatively suitable, and meeting less than 2 of the criteria is unsuitable"; Performance testing: 1000 natural language questions were selected (including 300 questions from non-professional users). The question-answering accuracy rate reached 89.5% (≥87%), and the comprehension rate of non-professional users reached 92%, indicating that the module has good usability and intelligent response capabilities.

[0034] V. Cross-validation of system performance; Validation was conducted using three independent datasets. The dataset sources and validation methods are as follows: 1. Larch distribution validation dataset; Data set: Geographic coordinates (latitude, longitude, altitude) and climate data (average annual temperature, annual precipitation) of 112 larch standard sample plots across the country, derived from field survey data; Verification method: Compare the geographical data and measured data of "suitable distribution area of ​​larch" in the system map to calculate the accuracy of geographical distribution; Result: Accuracy rate 97.2% (≥96%). 2. Site information verification dataset; Data set: 536 larch standard plots with measured site factor data (slope, soil thickness, organic matter content, etc.), derived from field survey data; Verification method: Compare the numerical range of each site factor in the system map with the range of measured data, and calculate the consistency rate; Results: The compliance rate of environmental factor values ​​was 86.5% (≥84%). 3. Ground Index Validation Dataset; Dataset: A publicly available dataset of spatial distribution of site index (SI) of four major larch species in China, derived from A gridded 90m dataset of site index for site quality assessment of four larch species in China (V1.0); Verification method: Compare the degree of agreement between the "site index interval" (e.g., suitable interval SI≥18m) in the system map and the measured interval, and calculate the agreement rate; Results: The compliance rate of SI site quality intervals was 90.3% (≥88%). Summary of Implementation Results The larch plantation site quality evaluation knowledge graph construction and service system constructed in this embodiment achieves the following core effects: Comprehensive knowledge coverage: It integrates 69 multi-dimensional evaluation indicators, achieving full coverage compared to traditional indicators, and fully covers the entire chain of "factor-indicator-evaluation" with 1,424 entity nodes and 3,152 semantic relationships; Improved technical efficiency: The three-stage prompting process improves knowledge extraction efficiency by more than 40% compared to traditional methods, and the intelligent query response time is ≤0.62 seconds; The service is highly accurate: the accuracy rate of correlation analysis is 87.3%, the accuracy rate of intelligent question answering is 89.5%, and all cross-validation indicators meet the design requirements, which can support the knowledge needs of different user groups such as forestry researchers and grassroots afforestation personnel.

[0035] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0036] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A knowledge graph construction and service system for larch plantation site quality evaluation, characterized in that: It includes a schema layer, a data layer, and a knowledge service module; The pattern layer and the data layer interact through a two-way constraint verification mechanism to collaboratively complete the construction of the knowledge graph and provide data support for the knowledge service module. The model layer is based on industry standards, resource systems and expert knowledge to construct an ontology framework for evaluating larch site quality. The entity classes are evaluation indicators and site factors as the core top-level classes, and the semantic relationships cover hierarchical structure, causal relationship and other relationships. The data layer includes modules for data collection, data processing, corpus construction, large language model training, semantic alignment, and storage. The data collection module is used to collect multi-source heterogeneous data. The large language model training module adopts a three-stage progressive prompting engineering to automatically extract entity and relation information from unstructured text. The storage module is based on the Neo4j graph database to store and manage knowledge graphs and construct a "node-edge" graph structure. The knowledge service module provides intelligent query, association analysis and intelligent question answering functions. The system uses three independent datasets for cross-validation to ensure the accuracy and reliability of the knowledge graph.

2. The knowledge graph construction and service system for larch plantation site quality evaluation according to claim 1, characterized in that: The ontology framework of the schema layer establishes relational instances, and the entity class includes four site factor subclasses: climate, terrain, soil, and organisms.

3. The knowledge graph construction and service system for larch plantation site quality evaluation according to claim 1, characterized in that: The data collection module collects multi-source heterogeneous data, including national standards and specifications, industry technical documents, academic papers, research reports, and field survey data. The data processing module performs consistency verification of multi-source data through cleaning and quality control.

4. The knowledge graph construction and service system for larch plantation site quality evaluation according to claim 3, characterized in that: The three-stage progressive prompting engineering includes requirement training, ontology training, and example training. This strategy can automatically extract entity and relation information from unstructured text.

5. The knowledge graph construction and service system for larch plantation site quality evaluation according to claim 1, characterized in that: The semantic alignment module eliminates terminological ambiguity and format differences between different data sources by processing the original triples through entity alignment and relation standardization.

6. The knowledge graph construction and service system for larch plantation site quality evaluation according to claim 1, characterized in that: The intelligent query function in the knowledge service module supports entity query, relationship query and compound query, and uses semantic association retrieval technology to achieve fast knowledge retrieval.

7. The knowledge graph construction and service system for larch plantation site quality evaluation according to claim 6, characterized in that: The association analysis function in the knowledge service module is based on graph algorithms to mine complex relationships between site factors and construct a multi-level factor network to generate larch growth suitability evaluation results.

8. The knowledge graph construction and service system for larch plantation site quality evaluation according to claim 7, characterized in that: The intelligent question-answering function adopts a hybrid question-answering strategy that prioritizes knowledge graphs. It can receive natural language queries and match them with knowledge graph information. When relevant information is lacking in the knowledge graph, the system will call the DeepSeek-Chat large language model to generate supplementary answers, thereby achieving a comprehensive and flexible question-answering service.