Railway green performance data knowledge graph construction method and system

Abstract

The invention provides a railway green performance data knowledge graph construction method and system, and the method comprises the steps: obtaining railway green performance data, cleaning the data, and obtaining the cleaned data; obtaining cleaned unstructured data, performing extraction of association information of quantitative characteristics on the unstructured data, and establishing an association relationship between an entity and a corresponding attribute; performing knowledge extraction on the associated data to obtain a plurality of triples; the triple is stored in the graph database, and construction of the railway knowledge graph is completed along with long-term accumulation and expansion of mass green performance related data in the whole life cycle of the railway and updating of the triple in the knowledge graph, so that the problems of difficulty in data management and high labor cost are solved; according to the method, a large amount of railway green performance data is integrated, the incidence relation between entities is established, the entities, attributes and the relation are extracted from mass data in an automatic mode, the burden of manual processing is relieved, and the cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph construction technology, and in particular to a method and system for constructing a knowledge graph of railway green performance data. Background Technology

[0002] The green performance-related data resources involved in railway engineering design, construction and operation are numerous and the data types are complex (such as design drawings, water and electricity consumption data, etc.). The task of collecting, sorting and storing key data is relatively difficult, including the following aspects: Railway engineering design and construction involve multi-source data from different departments and systems. These data may exist in different formats, structures and standards, so there are difficulties in data management.

[0003] Furthermore, railway engineering involves a large amount of data, including design drawings, monitoring data, and construction records. This data is massive and complex, and effectively storing data with inconsistent formats requires significant manpower.

[0004] Therefore, there is an urgent need for a method and system for constructing a knowledge graph of railway green performance data. Summary of the Invention

[0005] This invention provides a method and system for constructing a knowledge graph of railway green performance data, in order to solve the shortcomings of existing technologies such as difficult data management and high labor costs.

[0006] On the one hand, this invention provides a method for constructing a knowledge graph of railway green performance data, including:

[0007] Obtain railway green performance data, clean the data, and obtain cleaned data;

[0008] The process involves acquiring cleaned unstructured data, extracting quantifiable features from it, and then associating and matching the extracted quantifiable features with entities.

[0009] Knowledge extraction is performed on the associated data to obtain several triples;

[0010] By storing triples in a graph database, and expanding and updating the triples in the knowledge graph as massive amounts of green performance-related data throughout the railway's entire lifecycle are accumulated over a long period, the railway knowledge graph can be completed.

[0011] According to the present invention, a method for constructing a knowledge graph of railway green performance data includes, in the step of acquiring railway green performance data and cleaning the data to obtain cleaned data, the following steps are included:

[0012] The railway green performance data is divided into multiple external buckets according to data type; each external bucket is further divided into multiple internal buckets containing the same amount of data based on a preset partitioning mechanism.

[0013] The partitioning mechanism includes:

[0014] Continuously monitor the data stream and assess changes in data volume, including the rate at which new data is added and the frequency of data deletion or updates;

[0015] Set a capacity threshold for the internal buckets to ensure that the amount of data in each internal bucket is equal;

[0016] Using time series segmentation algorithms, geographical region division algorithms, and route classification algorithms, data is automatically allocated to appropriate internal buckets;

[0017] When the amount of data increases or decreases, automatically create new internal buckets or merge existing buckets to maintain a uniform distribution of data;

[0018] Record logs of each partition and adjustment, and detect and report any anomalies;

[0019] The data in the internal buckets is cleaned separately using a step-by-step processing method, including: cleaning the data in each internal bucket and performing cluster analysis on the cleaned data; wherein, the data cleaning includes identifying and removing abnormal data, including duplicate data and incomplete data; the data cleaning also includes removing noise, interference and non-signal data from the data;

[0020] The cleaned data in each inner bucket is integrated, including: merging the clustering results of data in all inner buckets divided by the same outer bucket to obtain the clustering results of each outer bucket; and integrating the clustering results of all outer buckets to obtain the clustering analysis results of all data.

[0021] According to the method for constructing a knowledge graph of railway green performance data provided by the present invention, the step of performing cluster analysis on the cleaned data includes:

[0022] Select k data points from each inner bucket as cluster centers;

[0023] Calculate the clustering of each data point to the k cluster centers;

[0024] Each data point is assigned to the nearest cluster center, forming k clusters;

[0025] Recalculate the centroid of each cluster until the position of the centroid no longer changes or the set number of iterations is reached;

[0026] By following the steps described above, each data point is assigned to a different cluster center, resulting in the clustering results for each internal bucket.

[0027] According to the present invention, a method for constructing a knowledge graph of railway green performance data includes acquiring cleaned unstructured data, extracting quantifiable features from it, and matching the extracted quantifiable features with entities, comprising:

[0028] The cleaned unstructured data is obtained, and multiple entity candidates and multiple attribute candidates are extracted from the target corpus based on multiple entity seeds in the unstructured data entity seed set. The entity seed set consists of multiple entity seeds belonging to the target category.

[0029] Based on multiple entities and multiple attribute candidates, an association relationship between entities and attributes is established, wherein the multiple entities include multiple entity seeds and multiple entity candidates.

[0030] According to the present invention, a method for constructing a knowledge graph of railway green performance data establishes the association between entities and attributes based on multiple entities and multiple attribute candidates, including:

[0031] Based on the association coefficient between each entity and each attribute in the association relationship between the entities and attributes, each entity among the multiple entities and each attribute candidate among the multiple attribute candidates are scored;

[0032] Based on the scoring results of the plurality of entities, at least one target entity is determined from the plurality of entities;

[0033] Based on the scoring results of the multiple attribute candidates, at least one target attribute is determined from the multiple attribute candidates.

[0034] According to the method for constructing a knowledge graph of railway green performance data provided by the present invention, knowledge extraction is performed on the associated data to obtain several triples, including:

[0035] Knowledge extraction is performed on the associated data. Since the data type is unstructured data, entities, entity relationships, and entity attributes are extracted from the knowledge data respectively.

[0036] Based on the extracted entities and attributes, and the data resource description framework of the knowledge graph, triples required for knowledge graph storage are formed.

[0037] According to the present invention, a method for constructing a railway green performance data knowledge graph is provided, which, based on the data resource description framework of the knowledge graph, forms the triples required for knowledge graph storage, including:

[0038] Choose an entity as the subject;

[0039] Determine the type of relationship between the subject and object, and choose an appropriate predicate to describe this relationship;

[0040] Choose the value or object associated with the subject as the object;

[0041] The subject, predicate, and object are organized together to form a complete triple;

[0042] The generated triples are stored in the knowledge graph database to form the triples required for knowledge graph storage.

[0043] On the other hand, the present invention also provides a railway green performance data knowledge graph construction system, comprising:

[0044] Data cleaning module: Acquires railway green performance data, cleans the data, and obtains cleaned data;

[0045] Attribute association module: acquires cleaned unstructured data, extracts association information from its quantitative features, and establishes association relationships between entities and their corresponding attributes;

[0046] Data extraction module: Extracts knowledge from the associated data to obtain several triples;

[0047] Graph Construction Module: Stores triples in the graph database. With the long-term accumulation of massive green performance-related data throughout the railway's entire life cycle, the triples in the knowledge graph are expanded and updated to complete the construction of the railway knowledge graph.

[0048] This invention provides a method and system for constructing a knowledge graph of railway green performance data. The method includes: acquiring railway green performance data, cleaning the data to obtain cleaned data; acquiring the cleaned unstructured data, extracting association information from its quantitative features, and establishing association relationships between entities and their corresponding attributes; extracting knowledge from the associated data to obtain several triples; storing the triples in a graph database, and expanding and updating the triples in the knowledge graph as massive amounts of green performance-related data throughout the railway's lifecycle accumulate over a long period, thus completing the construction of the railway knowledge graph. This method solves the problems of difficult data management and high labor costs, integrates a large amount of railway green performance data, establishes association relationships between entities, and extracts entities, attributes, and relationships from massive amounts of data in an automated manner, reducing the burden of manual processing, lowering costs, and achieving beneficial effects. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0050] Figure 1 This is a flowchart illustrating the method for constructing a knowledge graph of railway green performance data provided in an embodiment of the present invention;

[0051] Figure 2 This is a schematic diagram of the railway green performance data knowledge graph construction system provided in this embodiment of the invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0053] Figure 1 This is a flowchart illustrating the method for constructing a knowledge graph of railway green performance data provided in an embodiment of the present invention.

[0054] like Figure 1 As shown in the figure, an embodiment of the present invention provides a method for constructing a knowledge graph of railway green performance data. The method mainly includes the following steps:

[0055] 101. Obtain railway green performance data, clean the data, and obtain cleaned data.

[0056] Obtain railway green performance data, clean the data, and obtain the cleaned data, including:

[0057] The railway green performance data is divided into multiple external buckets according to data type; each external bucket is further divided into multiple internal buckets containing the same amount of data based on a preset partitioning mechanism.

[0058] The partitioning mechanism includes:

[0059] Continuously monitor the data stream and assess changes in data volume, including the rate at which new data is added and the frequency of data deletion or updates;

[0060] Set a capacity threshold for the internal buckets to ensure that the amount of data in each internal bucket is equal;

[0061] Using time series segmentation algorithms, geographical region division algorithms, and route classification algorithms, data is automatically allocated to appropriate internal buckets;

[0062] When the amount of data increases or decreases, automatically create new internal buckets or merge existing buckets to maintain a uniform distribution of data;

[0063] Record logs of each partition and adjustment, and detect and report any anomalies;

[0064] Among them, the time series segmentation algorithm, the geographical region division algorithm, and the route classification algorithm can be implemented through the following algorithms:

[0065] I. Time Series Segmentation Algorithms

[0066] 1. Sliding window method: By setting a fixed-length window and moving it over the time series, the data is divided into multiple subsequences; suitable for scenarios with large amounts of data that require real-time or near-real-time analysis;

[0067] 2. STL decomposition: Decomposes a time series into three components: trend, seasonality, and residuals; by identifying seasonal patterns, time series can be segmented more accurately, and it is suitable for data with obvious periodicity.

[0068] 3. ARIMA: A statistical model used for time series forecasting, which predicts future values ​​by fitting historical data; when segmenting data, the ARIMA model can be used to identify the trend and periodicity of the data first, and then segmentation can be performed based on this information;

[0069] 4. Breakpoint detection: By identifying breakpoints in a time series, i.e. points where the data distribution or trend changes significantly, the time series can be segmented into multiple subsequences with different characteristics; commonly used algorithms include Binary Segmentation and PELT.

[0070] II. Geographic Region Division Algorithm

[0071] 1. K-Means: Clustering is performed by minimizing the mean squared error of samples within each cluster; in geographical region division, geographical coordinates can be used as features to cluster data points, thereby dividing geographical regions.

[0072] 2. DBSCAN: A density-based clustering algorithm that can identify clusters of arbitrary shapes and is suitable for processing data with complex geographical distributions;

[0073] 3. Hierarchical clustering: By constructing a dendrogram to display the similarity between data points, it can be used to identify geographically adjacent areas;

[0074] III. Line Classification Algorithm

[0075] 1. Decision Tree: Used for classification tasks; by setting a series of feature-based decision rules, a decision tree can classify routes into different categories;

[0076] 2. Random Forest: A random forest is a collection of decision trees that makes the final classification decision through majority voting or average predictions; it is more stable and has stronger generalization ability than a single decision tree.

[0077] 3. SVM: SVM is a binary classification model that maximizes the margin between different classes by finding the optimal hyperplane; for multi-class problems, it can be extended using one-to-one or one-to-many strategies.

[0078] 4. Neural networks: In particular, deep learning models such as convolutional neural networks (CNN) and long short-term memory networks (LSTM) can handle complex nonlinear relationships and are very suitable for handling line classification problems with a large number of features;

[0079] The above algorithms can be selected and combined according to specific needs and data characteristics to achieve efficient data segmentation of railway green performance.

[0080] In some embodiments, it is assumed that we are processing green performance data of China's high-speed railway network. Data sources include, but are not limited to, train operation status, energy consumption, noise pollution monitoring, carbon emission records, etc. This data is generated over time and with changes in geographical areas, forming a huge time series and geospatial dataset.

[0081] I. Time Series Segmentation Algorithm: Using the sliding window method combined with seasonally adjusted time series analysis, the algorithm automatically identifies the trend and periodicity of the data, and then evenly segments the data into internal buckets according to time.

[0082] 1. Data preprocessing: Clean the raw data, remove outliers and missing values, and ensure the continuity and accuracy of timestamps;

[0083] 2. Trend and Cycle Detection: Using the STL method, identify long-term trends and short-term cyclical patterns in the data;

[0084] 3. Sliding window partitioning: Based on the detected periodicity, set the length of the sliding window (e.g., one day, one week, or one month) and evenly divide the data into multiple internal buckets in chronological order;

[0085] 4. Dynamic adjustment: Regularly (e.g., quarterly) reassess data trends and periodicity, and adjust the sliding window length and internal bucket division strategy as needed;

[0086] Assuming that the monitored train energy consumption data has obvious daily and annual periodicity, the data can be divided into an internal bucket for each day, and the data can be compared and analyzed at the same time each year to assess the seasonal changes in green performance.

[0087] II. Geographic Region Division Algorithm: Combining Geographic Information System (GIS) technology and cluster analysis, the algorithm automatically identifies regional boundaries based on the geographic location information of the data and divides data with similar geographic features into the same internal bucket;

[0088] 1. Geographic coordinate extraction: Extract the latitude and longitude information of each record from the dataset;

[0089] 2. Cluster analysis: Using clustering algorithms such as K-means or DBSCAN, the data is clustered according to geographic coordinates to identify naturally formed geographic regions;

[0090] 3. Determining regional boundaries: Using methods such as convex hull or Delaunay triangulation, boundaries are determined for each cluster to form geographical regions;

[0091] 4. Data allocation: Each data point is assigned to the geographical region closest to its location, forming internal buckets;

[0092] Cluster analysis can identify multiple ecologically sensitive areas along the high-speed rail line, such as national parks and river basins. Data from these areas can be divided into separate internal buckets to facilitate environmental impact assessments specifically targeting these areas.

[0093] III. Route Classification Algorithm: A rule-based classification method is adopted, which combines historical data and expert knowledge to automatically identify the characteristics of different routes and classify data with similar operating modes and green performance into the same internal bucket.

[0094] 1. Route Feature Extraction: Extract features such as route operation frequency, train type, and passenger flow from historical data;

[0095] 2. Rule Engine Configuration: Based on line characteristics and green performance indicators, set classification rules, such as "high-speed rail lines with average energy consumption below the threshold";

[0096] 3. Route Classification: Based on the rule engine, routes are automatically classified into predefined categories;

[0097] 4. Data allocation: Line data belonging to the same category are allocated to the same internal bucket to facilitate subsequent performance analysis and optimization;

[0098] It can identify two categories: "busy city routes" and "long-distance high-speed routes". The former may be more concerned about noise pollution and energy efficiency, while the latter may be more concerned about carbon emissions and ecological impact. Storing the data of these two types of routes separately can help to optimize green performance in a targeted manner.

[0099] By implementing the above algorithms, intelligent and automated classification of railway green performance data can be achieved, providing an efficient and accurate data foundation for subsequent data analysis and decision support.

[0100] The various algorithms provided in the above embodiments are existing technologies and will not be elaborated further.

[0101] The data in each internal bucket is cleaned separately using a step-by-step processing method. Specifically, this includes cleaning the data in each internal bucket and performing cluster analysis on the cleaned data. The data cleaning includes identifying and removing abnormal data, including duplicate and incomplete data. The data cleaning also includes removing noise, interference, and non-signal data from the data.

[0102] The cleaned data in each inner bucket is integrated. Specifically, the clustering results of the data in all inner buckets divided by the same outer bucket are merged to obtain the clustering results of each outer bucket; the clustering results of all outer buckets are then integrated to obtain the clustering analysis results of all data.

[0103] Based on data type, railway green performance data can be divided into different external buckets. For example, different types of data, such as operational data, equipment monitoring data, and energy consumption data, can be placed into different external buckets. This allows for classification and management based on data characteristics and uses. Each external bucket can be further divided into multiple internal buckets containing the same amount of data. This division can be determined based on actual needs and data volume, such as by time period, region, or line. Each internal bucket can contain a certain amount of data, making data retrieval and processing more efficient.

[0104] In this way, railway green performance data can be organized and managed hierarchically. The division of external buckets allows for the classification and summarization of data types, facilitating subsequent data retrieval and analysis; while the division of internal buckets improves data storage and processing efficiency, making data access more convenient and faster.

[0105] The identification and removal of outlier data includes: comparing data using unique identifiers or key attributes to identify data with the same characteristics or that recur within the same time period; handling data lacking necessary information by filling in missing values ​​or deleting data containing missing values; if there are too many missing values ​​or they cannot be recovered, deleting data containing missing values ​​may be considered; using statistical methods to identify and remove outliers; and correcting data that exceeds a predefined valid range, or verifying and adjusting it based on domain knowledge. For example, if the current value in green performance data exceeds the maximum value specified by the equipment, the current value can be corrected.

[0106] Non-signal data refers to data with low relevance to green performance, which may be due to data acquisition errors, invalid measurements, or other reasons. During the cleaning process, it is necessary to identify and remove this non-signal data to maintain data accuracy and consistency.

[0107] Cluster analysis was performed on the cleaned data, including:

[0108] Select k data points from each inner bucket as cluster centers;

[0109] Calculate the clustering of each data point to the k cluster centers;

[0110] Each data point is assigned to the nearest cluster center, forming k clusters;

[0111] Recalculate the centroid of each cluster until the position of the centroid no longer changes or the set number of iterations is reached;

[0112] By following the steps described above, each data point is assigned to a different cluster center, resulting in the clustering results for each internal bucket.

[0113] The cluster centers are adjusted based on the distribution of data points until a stable state is reached. Ultimately, each cluster has a cluster center that represents the characteristics of that cluster.

[0114] 102. Obtain the cleaned unstructured data, extract the correlation information from its quantitative features, and establish the association relationship between entities and their corresponding attributes.

[0115] The cleaned unstructured data is obtained, and multiple entity candidates and multiple attribute candidates are extracted from the target corpus based on multiple entity seeds in the unstructured data entity seed set. The entity seed set consists of multiple entity seeds belonging to the target category.

[0116] Based on multiple entities and multiple attribute candidates, an association relationship between entities and attributes is established, wherein the multiple entities include multiple entity seeds and multiple entity candidates.

[0117] Based on the association coefficient between each entity and each attribute in the association relationship between the entities and attributes, each entity among the multiple entities and each attribute candidate among the multiple attribute candidates are scored;

[0118] Based on the scoring results of the plurality of entities, at least one target entity is determined from the plurality of entities;

[0119] Based on the scoring results of the multiple attribute candidates, at least one target attribute is determined from the multiple attribute candidates.

[0120] The process of establishing entity-attribute relationships based on multiple entities and multiple attribute candidates involves matching each entity candidate with the most relevant attribute candidate using text matching or other similarity calculation methods. For example, metrics such as word overlap, similarity, or distance between entity and attribute candidates can be calculated, and the best match is selected. Based on the matching results between entities and attributes, the relationship between the entity and its corresponding attribute is determined, and these relationships are stored and managed using relational graphs, relational databases, or other data structures. The established entity-attribute relationships are then verified and updated by refining or improving the relationships through manual review, feedback from domain experts, and subsequent data processing.

[0121] In one embodiment, a knowledge graph is constructed regarding the domain of "railway green performance." This graph contains entities (such as "high-speed trains," "CO2 emissions," and "energy efficiency") and attributes (such as "speed," "emissions," and "energy consumption"). The following is a specific scoring process combining multiple technologies:

[0122] 1. Data Preparation

[0123] We collected a large number of documents, reports and technical documents related to "green performance of railways".

[0124] Natural language processing techniques, such as named entity recognition (NER) and relation extraction (RE), are used to extract entities and attributes from text, as well as the potential relationships between them.

[0125] 2. Feature Engineering

[0126] For each pair of entities and attributes, construct the following features:

[0127] Co-occurrence frequency: The number of times an entity and attribute co-occur in the same document.

[0128] Text similarity: Calculate the word vector representations of entity and attribute descriptions (using a pre-trained word embedding model such as BERT), and then calculate the cosine similarity between these vectors.

[0129] Semantic relationships: Utilize external knowledge bases (such as WordNet) to measure the semantic similarity between entities and attributes.

[0130] Contextual information: The average TF-IDF values ​​of words surrounding entities and attributes are used to reflect the specificity of the context.

[0131] 3. Training the model

[0132] Build a supervised learning model, such as XGBoost or a deep neural network, to predict the strength of the association between entities and attributes.

[0133] Tag generation: Domain experts label a subset of entity attribute pairs as positive and negative examples. Positive examples refer to entity attribute relationships that experts have confirmed actually exist; negative examples are theoretically non-existent or unreasonable relationships.

[0134] Model training: Use positive and negative examples as training data, take the above features as input, and the correlation as output (e.g., use 1 to represent strong correlation and 0 to represent no correlation) to train the model.

[0135] 4. Entity and Attribute Scoring

[0136] For each pair of entities and attributes in the knowledge graph, a trained model is used to predict their association score.

[0137] This score reflects the strength of the association between entities and attributes; a high score indicates a strong association, while a low score indicates a weak association or no association at all.

[0138] 5. Target entity and attribute selection

[0139] Based on the scores of entities and attributes, the top N entities and attributes with the highest scores are selected as target entities and target attributes for further construction and optimization of the knowledge graph.

[0140] A threshold can be set, and only entities and attributes whose scores exceed a certain threshold will be included in the final knowledge graph.

[0141] 6. Model Iteration and Optimization

[0142] Regularly collect new data and feedback for model iteration and optimization, continuously improving the accuracy of entity and attribute correlation prediction.

[0143] 103. Extract knowledge from the associated data to obtain several triples.

[0144] Knowledge extraction is performed on the associated data. Since the data type is unstructured data, entities, entity relationships, and entity attributes are extracted from the knowledge data respectively.

[0145] Based on the extracted entities and attributes, and the data resource description framework of the knowledge graph, triples required for knowledge graph storage are formed.

[0146] Knowledge extraction is a method for extracting useful information from unstructured data. In correlated data, natural language processing techniques can be used to extract entities, entity relationships, and entity attributes.

[0147] Entity relation extraction refers to extracting the relationships between entities from text, while entity attribute extraction refers to extracting attributes or features related to entities from text.

[0148] After extracting entities, entity relationships, and entity attributes, this knowledge can be represented using a knowledge graph data resource description framework. Knowledge graphs typically use triples (subject, predicate, object) to represent entities and relationships. Each entity can be a node, and each relationship can be an edge; knowledge graphs are constructed by combining these nodes and edges.

[0149] Finally, the extracted entities, attributes, and relationships are stored as triples, which forms the data needed to create a knowledge graph. Such a knowledge graph can then be used for further applications such as knowledge reasoning and intelligent question answering.

[0150] Based on the extracted entities and attributes, and the data resource description framework of the knowledge graph, the triples required for knowledge graph storage are formed, including:

[0151] Select an entity as the subject; determine the type of relationship between the subject and the object, and select an appropriate predicate to describe this relationship; select a value or object associated with the subject as the object; organize the subject, predicate, and object together to form a complete triple; store the generated triple in the knowledge graph database to form the triple required for knowledge graph storage.

[0152] Selecting an entity as the subject means choosing a suitable entity as the subject based on the needs and the goals of the knowledge graph. This entity can be a person, a place, an organization, or any other entity with specific meaning.

[0153] Determining the type of relationship between the subject and the object refers to identifying the type of relationship between the subject and the object.

[0154] Choosing a suitable predicate to describe the relationship means selecting an appropriate predicate to describe the relationship between the subject and the object. The predicate should accurately represent this relationship and should have been defined in the knowledge graph.

[0155] In this context, selecting a value or object associated with the subject as the object means determining the value or object associated with the subject and using it as the object. The object can be a specific entity, such as a location, or an attribute value, such as age or address.

[0156] Organizing a triple refers to combining the subject, verb, and object to form a complete triple. Triples should be arranged in the order of "subject-verb-object" and follow the data resource description framework of a knowledge graph.

[0157] Storing in a knowledge graph database refers to storing the generated triples in the knowledge graph database. Depending on the characteristics and requirements of the database, the triples can be inserted into the database using appropriate APIs or query languages.

[0158] 104. Store the triples in the graph database. As massive amounts of green performance-related data from the entire railway lifecycle accumulate over a long period, expand and update the triples in the knowledge graph to complete the construction of the railway knowledge graph.

[0159] As railway lifecycle data accumulates, the triples in the knowledge graph are continuously expanded and updated. New data can be added to the graph database through continuous extraction and transformation, thereby enriching the railway knowledge graph.

[0160] The query functions provided by the graph database can be used to query and analyze the railway knowledge graph. By writing appropriate query statements, specific entities, relationships, and attributes related to railways can be retrieved, and complex graph analysis and reasoning can be performed.

[0161] Based on the same general inventive concept, this invention also protects a railway green performance data knowledge graph construction system. The railway green performance data knowledge graph construction system provided by this invention will be described below. The railway green performance data knowledge graph construction system described below can be referred to in correspondence with the railway green performance data knowledge graph construction method described above.

[0162] Figure 2 This is a schematic diagram of the railway green performance data knowledge graph construction system provided in this embodiment of the invention.

[0163] like Figure 2 As shown in the figure, an embodiment of the present invention provides a railway green performance data knowledge graph construction system, which mainly includes:

[0164] Data cleaning module: Acquires railway green performance data, cleans the data, and obtains cleaned data;

[0165] Attribute association module: acquires cleaned unstructured data, extracts association information from its quantitative features, and establishes association relationships between entities and their corresponding attributes;

[0166] Data extraction module: Extracts knowledge from the associated data to obtain several triples;

[0167] Graph Construction Module: Stores triples in the graph database. With the long-term accumulation of massive green performance-related data throughout the railway's entire life cycle, the triples in the knowledge graph are expanded and updated to complete the construction of the railway knowledge graph.

[0168] Due to the inconsistent formats, large volume, and diverse types of railway green performance data, it is necessary to clean a large amount of irrelevant data, summarize and classify some useful information, and use clustering analysis techniques to extract corresponding correlation information from this unstructured data. This includes the direct and indirect relationships between data and design, construction, and operation units, as well as the relationships between data and directly related personnel in each unit. Based on the data resource description framework of knowledge graphs, the triples required for knowledge graph storage, namely the "subject-verb-object" structure, are formed. These triples are then stored in a graph database. With the long-term accumulation of massive amounts of green performance-related data throughout the railway's entire lifecycle, a vast entity relationship network can be constructed, thereby completing the construction of the railway knowledge graph.

[0169] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0170] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0171] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for constructing a knowledge graph of railway green performance data, characterized in that, include: Obtain railway green performance data, clean the data, and obtain cleaned data; The process involves acquiring cleaned unstructured data, extracting quantifiable features from it, and then associating and matching the extracted quantifiable features with entities. Knowledge extraction is performed on the associated data to obtain several triples; By storing triples in a graph database, and expanding and updating the triples in the knowledge graph as massive amounts of green performance-related data throughout the railway's entire lifecycle are accumulated over a long period, the railway knowledge graph can be completed.

2. The method for constructing a knowledge graph of railway green performance data according to claim 1, characterized in that, The process of acquiring railway green performance data, cleaning the data, and obtaining cleaned data includes: The railway green performance data is divided into multiple external buckets according to data type; each external bucket is further divided into multiple internal buckets containing the same amount of data based on a preset partitioning mechanism. The partitioning mechanism includes: Continuously monitor the data stream and assess changes in data volume, including the rate at which new data is added and the frequency of data deletion or updates; Set a capacity threshold for the internal buckets to ensure that the amount of data in each internal bucket is equal; Using time series segmentation algorithms, geographical region division algorithms, and route classification algorithms, data is automatically allocated to appropriate internal buckets; When the amount of data increases or decreases, automatically create new internal buckets or merge existing buckets to maintain a uniform distribution of data; Record logs of each partition and adjustment, and detect and report any anomalies; The data in the internal buckets is cleaned separately using a step-by-step processing method, including: cleaning the data in each internal bucket and performing cluster analysis on the cleaned data; wherein, the data cleaning includes identifying and removing abnormal data, including duplicate data and incomplete data; the data cleaning also includes removing noise, interference and non-signal data from the data; The cleaned data in each inner bucket is integrated, including: merging the clustering results of data in all inner buckets divided by the same outer bucket to obtain the clustering results of each outer bucket; and integrating the clustering results of all outer buckets to obtain the clustering analysis results of all data.

3. The method for constructing a knowledge graph of railway green performance data according to claim 2, characterized in that, The cluster analysis of the cleaned data includes: Select k data points from each inner bucket as cluster centers; Calculate the clustering of each data point to the k cluster centers; Each data point is assigned to the nearest cluster center, forming k clusters; Recalculate the centroid of each cluster until the position of the centroid no longer changes or the set number of iterations is reached; By following the steps described above, each data point is assigned to a different cluster center, resulting in the clustering results for each internal bucket.

4. The method for constructing a knowledge graph of railway green performance data according to claim 1, characterized in that, The process of acquiring cleaned unstructured data, extracting quantifiable features from it, and matching the extracted quantifiable features with entities includes: The cleaned unstructured data is obtained, and multiple entity candidates and multiple attribute candidates are extracted from the target corpus based on multiple entity seeds in the unstructured data entity seed set. The entity seed set consists of multiple entity seeds belonging to the target category. Based on multiple entities and multiple attribute candidates, an association relationship between entities and attributes is established, wherein the multiple entities include multiple entity seeds and multiple entity candidates.

5. The method for constructing a knowledge graph of railway green performance data according to claim 4, characterized in that, The step of establishing the association between entities and attributes based on multiple entities and multiple attribute candidates includes: Based on the association coefficient between each entity and each attribute in the association relationship between the entities and attributes, each entity among the multiple entities and each attribute candidate among the multiple attribute candidates are scored; Based on the scoring results of the plurality of entities, at least one target entity is determined from the plurality of entities; Based on the scoring results of the multiple attribute candidates, at least one target attribute is determined from the multiple attribute candidates.

6. The method for constructing a knowledge graph of railway green performance data according to claim 1, characterized in that, The process of extracting knowledge from the associated data yields several triples, including: Knowledge extraction is performed on the associated data. Since the data type is unstructured data, entities, entity relationships, and entity attributes are extracted from the knowledge data respectively. Based on the extracted entities and attributes, and the data resource description framework of the knowledge graph, triples required for knowledge graph storage are formed.

7. The method for constructing a knowledge graph of railway green performance data according to claim 6, characterized in that, The data resource description framework based on the extracted entities and attributes and the knowledge graph forms the triples required for knowledge graph storage, including: Choose an entity as the subject; Determine the type of relationship between the subject and object, and choose an appropriate predicate to describe this relationship; Choose the value or object associated with the subject as the object; The subject, predicate, and object are organized together to form a complete triple; The generated triples are stored in the knowledge graph database to form the triples required for knowledge graph storage.

8. A system for constructing a knowledge graph of railway green performance data, characterized in that, include: Data cleaning module: Acquires railway green performance data, cleans the data, and obtains cleaned data; Attribute association module: acquires cleaned unstructured data, extracts association information from its quantitative features, and establishes association relationships between entities and their corresponding attributes; Data extraction module: Extracts knowledge from the associated data to obtain several triples; Graph Construction Module: Stores triples in the graph database. With the long-term accumulation of massive green performance-related data throughout the railway's entire life cycle, the triples in the knowledge graph are expanded and updated to complete the construction of the railway knowledge graph.

Images