Event development trend prediction method, device, equipment, medium and program product

Abstract

This application provides a method, apparatus, device, medium, and program product for predicting event development trends, relating to the field of artificial intelligence technology, and is used to improve the accuracy of event development trend prediction. The specific technical solution is as follows: Data related to the target event is input into an event development trend prediction model; a multi-source data access standardization module performs data processing operations on the data related to the target event to obtain a standardized data source; a temporal causal knowledge modeling graph module extracts primary causal feature information from the standardized knowledge data source and constructs a causal knowledge graph based on the primary causal feature information; a neural-symbolic fusion prediction module processes the standardized temporal prediction data source and the causal knowledge graph to obtain the development trend prediction result information of the target event, and outputs it. This application is applied to digital trend prediction and analysis scenarios.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device, medium, and program product for predicting the development trend of events. Background Technology

[0002] With the development of internet technology and the continuous increase in information, data analysis technology has also seen new developments and breakthroughs.

[0003] In related technologies, it is possible for humans to make predictions based on statistical methods, such as data analysts, using historical data of the event to be analyzed (such as historical data of stock price fluctuations in the financial field, or historical data of the increase in patent application volume, etc.) to obtain prediction results for a certain period of time in the future (such as whether stocks will rise or fall, the increase in patent application volume, the increase in production volume, etc.).

[0004] However, because the prediction results of manual statistics are easily affected by subjective factors, the accuracy of the prediction results of the event development trend is low. Summary of the Invention

[0005] This application provides a method, apparatus, device, medium, and program product for predicting event development trends, which can improve the accuracy of event development trend prediction.

[0006] In a first aspect, embodiments of this application provide a method for predicting the development trend of an event. The method includes: inputting data related to a target event into an event development trend prediction model, the event development trend prediction model including: a multi-source data access standardization module, a temporal causal knowledge modeling graph module, and a neural-symbolic fusion prediction module; performing data processing operations on the data related to the target event through the multi-source data access standardization module to obtain a standardized data source, the standardized data source including: a standardized knowledge construction data source and a standardized time-series prediction data source; and extracting first causal feature information from the standardized knowledge construction data source through the temporal causal knowledge modeling graph module, and based on the first causal feature... The system constructs a causal knowledge graph from the information of the first causal feature; the first causal feature is used to characterize the causal information that influences the development trend of the event; through the neural-symbolic fusion prediction module, the standardized time series prediction data source and the causal knowledge graph are processed to obtain the development trend prediction result information of the target event, and output it; wherein, the first causal feature includes: the first keyword, the causal relationship feature information, and the event causal feature information; the first keyword is used to characterize the specific content or type of the event; the causal relationship feature includes at least one of the following: driving relationship, inhibiting relationship, triggering relationship; the event causal feature includes: the cause element of the event, the result element of the event, and the time relationship element of the event.

[0007] The technical solution provided in this application brings at least the following beneficial effects: by using an event development trend prediction model and combining relevant event data information to predict the event development trend, the process fully considers the causal relationship of event development, thereby improving the accuracy of event development trend prediction.

[0008] One possible implementation is that the above-mentioned multi-source data access standardization module performs data processing operations on the data related to the target event to obtain a standardized data source, including: performing data cleaning, format normalization, time sequence alignment and feature extraction processing operations on the data related to the target event through the multi-source data access standardization module to obtain a standard data source;

[0009] The data related to the aforementioned target event includes at least one of the following types of data:

[0010] Structured data;

[0011] Semi-structured data;

[0012] Unstructured data;

[0013] Real-time time series data.

[0014] Another possible implementation involves using the temporal causal knowledge modeling graph module to extract first causal feature information from the standardized knowledge construction data source. This includes: extracting the first keyword from the standardized knowledge construction data source based on a pre-set dictionary and contextual semantic analysis technology; extracting causal relationship feature information from the standardized knowledge construction data source based on syntactic analysis relation identification and semantic role labeling technology; and extracting event causal feature information from the standardized knowledge construction data source based on phrase extraction and temporal information extraction technology.

[0015] Another possible implementation involves processing the standardized temporal prediction data source and causal knowledge graph through the neural-symbolic fusion prediction module to obtain the prediction result information of the development trend of the target event. This includes: generating preliminary event development trend prediction result information through the neural network in the neural-symbolic fusion prediction module; correcting the preliminary event development trend prediction result information using a temporal causal constraint prediction correction algorithm; and outputting the prediction result information of the development trend of the target event.

[0016] Another possible implementation is that the above-mentioned event development trend prediction model also includes a temporal causal attribution module. After processing the standardized time-series prediction data source and causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event, the method further includes: constructing a baseline scenario and a counterfactual scenario based on the development trend prediction result information of the target event and the causal knowledge graph through the temporal causal attribution module; calculating the contribution of each influencing factor corresponding to the development trend prediction result information of the target event, generating a structured temporal causal attribution report, and outputting it.

[0017] Secondly, embodiments of this application provide an event development trend prediction device, including: an input module, a processing module, and an output module; the input module is used to input data related to a target event into an event development trend prediction model, the event development trend prediction model including: a multi-source data access standardization module, a temporal causal knowledge modeling graph module, and a neural-symbolic fusion prediction module; the processing module is used to perform data processing operations on the data related to the target event through the multi-source data access standardization module to obtain a standardized data source, the standardized data source including: a standardized knowledge construction data source and a standardized time series prediction data source; the processing module is also used to extract first causal feature information from the standardized knowledge construction data source through the temporal causal knowledge modeling graph module, and based on... A causal knowledge graph is constructed based on the first causal feature information. The first causal feature information is used to characterize the causal information that influences the development trend of the event. This processing module is also used to process the standardized time-series prediction data source and the causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event. This output module is used to output the development trend prediction result information of the target event. The first causal feature information includes: a first keyword, causal relationship feature information, and event causal feature information. The first keyword is used to characterize the specific content or type of the event. The causal relationship feature information includes at least one of the following: driving relationship, inhibiting relationship, and triggering relationship. The event causal feature information includes: event cause elements, event result elements, and event time relationship elements.

[0018] The technical solution provided in this application brings at least the following beneficial effects: by using an event development trend prediction model and combining relevant event data information to predict the event development trend, the process fully considers the causal relationship of event development, thereby improving the accuracy of event development trend prediction.

[0019] One possible implementation is that the aforementioned processing module is specifically used to perform data cleaning, format normalization, time sequence alignment, and feature extraction processing on the data related to the target event through the multi-source data access standardization module to obtain a standard data source;

[0020] The data related to the aforementioned target event includes at least one of the following types of data:

[0021] Structured data;

[0022] Semi-structured data;

[0023] Unstructured data;

[0024] Real-time time series data.

[0025] Another possible implementation, the aforementioned processing module, is specifically used to: extract the first keyword from the standardized knowledge construction data source based on a preset dictionary and contextual semantic analysis technology; extract causal relationship feature information from the standardized knowledge construction data source based on syntactic analysis relation identification and semantic role labeling technology; and extract event causal feature information from the standardized knowledge construction data source based on phrase extraction and temporal information extraction technology.

[0026] Another possible implementation is that the above processing module is specifically used to: generate preliminary event development trend prediction results through the neural network in the neural-symbolic fusion prediction module; correct the preliminary event development trend prediction results using a temporal causal constraint prediction correction algorithm; and output the development trend prediction results of the target event.

[0027] Another possible implementation is that the aforementioned event development trend prediction model also includes a temporal causal attribution module; the aforementioned processing module is further used to, after processing the standardized time-series prediction data source and causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event, construct a baseline scenario and a counterfactual scenario based on the development trend prediction result information of the target event and the causal knowledge graph through the temporal causal attribution module; the aforementioned processing module is further used to calculate the contribution of each influencing factor corresponding to the development trend prediction result information of the target event and generate a structured temporal causal attribution report; the aforementioned output module is further used to output the structured temporal causal attribution report.

[0028] Thirdly, this application provides an electronic device comprising: a processor and a memory; the memory stores a program or instructions executable on the processor, wherein the program or instructions, when executed by the processor, implement the method of the first aspect described above.

[0029] Fourthly, this application provides a readable storage medium on which a program or instructions are stored, which, when executed by a computer, implement the method of the first aspect described above.

[0030] Fifthly, this application provides a computer program product stored in a storage medium, which, when executed by a computer, implements the method described in the first aspect.

[0031] In a sixth aspect, embodiments of this application provide a chip including a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the method described in the first aspect.

[0032] The beneficial effects of the second to sixth aspects mentioned above are described in the corresponding description of the first aspect and will not be repeated here. Attached Figure Description

[0033] Figure 1 A schematic diagram of the core architecture of an intelligent data production method based on graph neural networks and adaptive learning provided in this application embodiment;

[0034] Figure 2 A flowchart of a GNN semantic relation mining method provided in this application embodiment;

[0035] Figure 3 This is a schematic diagram illustrating an iterative optimization of an adaptive learning mechanism provided in an embodiment of this application.

[0036] Figure 4 A flowchart illustrating a financial industry trend prediction case provided in this application embodiment;

[0037] Figure 5 A flowchart illustrating a multivariate time series analysis method based on the fusion of multiple visibility graphs and Transformer, provided for embodiments of this application;

[0038] Figure 6 A schematic diagram of the network architecture for an event development trend prediction method provided in this application embodiment;

[0039] Figure 7 A flowchart illustrating an event development trend prediction method provided in an embodiment of this application;

[0040] Figure 8 A flowchart illustrating an event development trend prediction method provided in an embodiment of this application;

[0041] Figure 9 A flowchart illustrating an event development trend prediction method provided in an embodiment of this application;

[0042] Figure 10 A flowchart illustrating a neural-symbol fusion prediction module provided in an embodiment of this application;

[0043] Figure 11A flowchart illustrating an event development trend prediction method provided in an embodiment of this application;

[0044] Figure 12 A flowchart illustrating an event development trend prediction method provided in an embodiment of this application;

[0045] Figure 13 A flowchart illustrating the implementation process of an event development trend prediction method provided in this application embodiment;

[0046] Figure 14 A schematic diagram of the structure of an event development trend prediction device provided in an embodiment of this application;

[0047] Figure 15 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0048] The following will describe in detail the event development trend prediction method, apparatus, equipment, medium and program products provided in this application, with reference to the accompanying drawings.

[0049] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0050] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0051] The terms "at least one," "at least one of," etc., used in the specification and claims of this application refer to any one, any two, or a combination of two or more of the included items. For example, at least one of a, b, and c can mean: "a," "b," "c," "a and b," "a and c," "b and c," and "a, b, and c," where a, b, and c can be single or multiple. Similarly, "at least two" refers to two or more items, and its meaning is similar to that of "at least one."

[0052] In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0053] This application provides a method, apparatus, device, medium, and program product for predicting event development trends, which can be applied to digital trend prediction and analysis scenarios. Specifically, the application scenarios of this embodiment can cover multiple industry sectors such as industrial manufacturing, e-commerce retail, financial transactions, government services, and energy supply. It is primarily applicable to scenarios requiring accurate trend prediction, identification of influencing factors, and dynamic optimization of decision-making strategies based on multi-source heterogeneous data. Whether it's production scheduling, sales forecasting, and risk management at the enterprise level, or regional economic trend analysis and public service resource allocation at the government level, as long as there is a need for "multi-source data integration - causal relationship analysis - trend prediction - attribution optimization," the technical solution provided in this embodiment can improve prediction accuracy, decision interpretability, and long-term adaptability.

[0054] In related technology 1, an intelligent data production method and system based on graph neural networks and adaptive learning is described. This method collects various types of data through a distributed crawler cluster and a multi-protocol adaptation engine, extracts entity features using a large language model, constructs a dynamic knowledge graph using graph neural networks, optimizes detection thresholds using reinforcement learning, predicts data trends using a spatiotemporal graph convolutional network, and outputs an analysis report. Its core architecture is as follows: Figure 1 As shown, it covers data collection, knowledge modeling, quality inspection, and trend prediction, focusing on multi-source data fusion and adaptive learning mechanisms, such as... Figure 2 The diagram shows a flowchart of semantic relation mining using a graph neural network (GNN) in related technology 1, as follows: Figure 3 The diagram shown is an iterative optimization illustration of the adaptive learning mechanism in related technology 1, as follows: Figure 4 The diagram shown is a flowchart of a financial industry trend prediction case in related technology 1.

[0055] In related technology 2, such as Figure 5 As shown, this paper presents a multivariate time series analysis method based on the fusion of multiple visibility graphs and a self-attention mechanism (Transformer). A multiple visibility graph is constructed using a sliding window mechanism, and cross-channel consensus connections are extracted using an AND aggregation mechanism. Node embeddings are iteratively updated using a visibility graph-attention (VG-Attention) mechanism, ultimately adapting to downstream tasks such as prediction and classification. This related technology focuses on solving the problems of global dependency capture and computational complexity in multivariate time series analysis, optimizing the feature extraction and modeling capabilities of time series data.

[0056] In the process of digital transformation, various industries have increasingly urgent needs for the accuracy, interpretability, and dynamic adaptability of event trend predictions. Although related technologies have achieved functions such as multi-source data collection, knowledge graph construction, and time series prediction, there is still room for improvement in areas such as deep causal relationship modeling, neural-symbolic fusion optimization, phased attribution analysis, and closed-loop iterative mechanisms. It is difficult to simultaneously meet the comprehensive needs of prediction accuracy, causal analysis, and dynamic optimization in complex scenarios.

[0057] The shortcomings of the aforementioned related technology 1 are: it focuses on cross-modal entity association modeling and data production quality control, but lacks a refined characterization of temporal causal relationships, does not construct a dynamic causal strength calculation model, and has weak causal interpretability of prediction results; attribution analysis only focuses on overall influencing factors and does not achieve phased contribution quantification, making it difficult to support accurate decision-making.

[0058] The shortcomings of the aforementioned related technology 2 are: it focuses on the extraction of structural features and dependency capture of multivariate time series, without incorporating domain causal rule constraints, and the prediction results are easily affected by data noise; it does not design a closed-loop iterative optimization mechanism, and cannot dynamically adjust model parameters and knowledge graphs according to real-time data updates and prediction performance feedback, resulting in insufficient long-term prediction stability.

[0059] The common shortcomings of both are: they have not achieved a closed-loop design of the entire process of "data standardization - causal knowledge modeling - neural-symbolic fusion prediction - phased attribution - iterative optimization", and there are gaps in the deep integration of causal constraints and neural networks, and the collaborative optimization of dynamic knowledge graphs and prediction models.

[0060] In response, this application provides a big data-based digital trend prediction and analysis system, namely the aforementioned event development trend prediction model. Through multi-source data standardization processing, temporal causal knowledge graph construction, neural-symbolic fusion prediction, phased attribution analysis, and closed-loop iterative optimization, it solves the problems of unclear causal relationships, insufficient prediction accuracy, one-sided attribution analysis, and poor dynamic adaptability in existing technologies, thereby improving the accuracy, interpretability, and long-term stability of digital trend prediction.

[0061] To address the aforementioned technical problems, embodiments of this application provide a method, apparatus, device, medium, and program product for predicting event development trends. By using an event development trend prediction model and combining relevant event data, the development trend of an event is predicted. The process fully considers the causal relationship of event development, thereby improving the accuracy of event development trend prediction.

[0062] The following description, in conjunction with the accompanying drawings, details the event development trend prediction method, apparatus, device, medium, and program products provided in the embodiments of this application.

[0063] Figure 6The diagram illustrates the network architecture of an event development trend prediction method provided in an embodiment of this application. For example... Figure 6 As shown, the network architecture includes an event development trend prediction device 101 and a terminal device 102. The event development trend prediction device 101 and the terminal device 102 are interconnected.

[0064] In some embodiments, the event development trend prediction device 101 may be a server, a computer, or a processor or processing unit within a server or computer. The server may be a single server or a server cluster consisting of multiple servers. It should be noted that the specific device form of the event development trend prediction device 101 is not limited in the embodiments of this application. Figure 6 The example shown is a single server, namely the event development trend prediction device 101.

[0065] In some embodiments, the terminal device may be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, personal computer (PC), ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc., and the embodiments of this application do not specifically limit it. Figure 6 The example shown is a mobile phone, with terminal device 102 as an example.

[0066] In some embodiments, the event development trend prediction device 101 receives an instruction from the terminal device 102 to predict the development trend of a target event. The event development trend prediction device 101 inputs data related to the target event into the event development trend prediction model. The event development trend prediction model includes: a multi-source data access standardization module, a temporal causal knowledge modeling graph module, and a neural-symbolic fusion prediction module. Through the multi-source data access standardization module, data processing operations are performed on the data related to the target event to obtain standardized data sources, including: standardized knowledge construction data sources and standardized temporal prediction data sources. Through the temporal causal knowledge modeling graph module, first causal feature information is extracted from the standardized knowledge construction data sources, and a causal knowledge graph is constructed based on the first causal feature information. The first causal feature information is used to characterize the causal information affecting the event development trend. Through the neural-symbolic fusion prediction module, the standardized temporal prediction data sources and the causal knowledge graph are processed to obtain the target event development trend prediction result information, which is then output. The first causal feature information includes: a first keyword, causal relationship feature information, and event causal feature information. The first keyword is used to characterize the specific content or type of the event. The causal relationship feature information includes at least one of the following: a driving relationship, an inhibiting relationship, and a triggering relationship. The event causal feature information includes: event cause elements, event result elements, and event time-related elements.

[0067] It should be noted that the network architecture described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As network architectures evolve, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0068] See Figure 7 This is a flowchart illustrating an event development trend prediction method provided in an embodiment of this application. Figure 7 As shown, the event development trend prediction method provided in this application embodiment can be implemented by the above-mentioned event development trend prediction device, specifically including the following steps 201 to 204.

[0069] Step 201: The event development trend prediction device inputs the data related to the target event into the event development trend prediction model.

[0070] In some embodiments, the above event development trend prediction model includes: a multi-source data access standardization module, a temporal causal knowledge modeling graph module, and a neural-symbolic fusion prediction module.

[0071] In some embodiments, the multi-source data access standardization module described above is used to process multi-source data into a standardized data source with a unified format.

[0072] In some embodiments, the temporal causal knowledge modeling graph module described above is used to extract and analyze standardized data sources, obtain their causal feature information, and construct a knowledge graph based on the causal feature information.

[0073] In some embodiments, the neural-symbolic fusion prediction module described above is used to fuse and correct causal and temporal feature information to obtain a prediction result of the event's development trend.

[0074] In some embodiments, the target event described above is a dynamically evolving event.

[0075] For example, the aforementioned target events include, but are not limited to: sales volume of goods, production volume of items, stock price trends, or road traffic forecasts.

[0076] In some embodiments, the relevant data of the target event mentioned above are data that are related to and have an impact on the development of the target event.

[0077] In some embodiments, the data related to the target event mentioned above includes at least one of the following types of data:

[0078] Structured data;

[0079] Semi-structured data;

[0080] Unstructured data;

[0081] Real-time time series data.

[0082] For example, the structured data mentioned above can be statistical reports, financial statements, market transaction records, equipment operating parameter tables, production scheduling data tables, etc.

[0083] For example, the aforementioned semi-structured data can include industry research reports, product documentation, user review records, equipment maintenance logs, project progress summaries, etc.

[0084] For example, the aforementioned unstructured data can include industry news articles, social media discussion texts, transcripts of interviews with corporate executives, subtitles for industry conference videos, and chat logs from technical exchanges.

[0085] For example, the aforementioned real-time time series data can be market transaction time-sharing data, e-commerce platform real-time order data, offline store customer flow statistics, industrial production line sensor real-time data collection, and network traffic real-time monitoring data.

[0086] Step 202: The event development trend prediction device performs data processing operations on the data related to the target event through the multi-source data access standardization module to obtain a standardized data source.

[0087] In some embodiments, the standardized data sources mentioned above include: standardized knowledge construction data sources and standardized time series prediction data sources.

[0088] In some embodiments, the aforementioned standard data source refers to a data set or data stream that has been processed by a multi-source data access standardization module, has a unified standard format, and can be accessed through a predetermined interface, and is used as data input for knowledge construction or time series prediction.

[0089] In some embodiments, after receiving data related to the target event, the multi-source data access standardization module performs data cleaning, format normalization, time sequence alignment, and feature extraction on the data to finally obtain a standardized data source.

[0090] For example, the aforementioned multi-source data access standardization module can receive input structured, semi-structured, unstructured, and real-time time-series data within the industry domain, i.e., data related to the aforementioned target event. It performs data cleaning, format normalization, time-series alignment, and feature extraction processing on these data, and outputs standardized knowledge construction data sources and standardized time-series prediction data sources.

[0091] In some embodiments, combined with Figure 7 ,like Figure 8 As shown, step 202 above can be specifically achieved through the following step 202a.

[0092] Step 202a: The event development trend prediction device performs data cleaning, format normalization, time sequence alignment, and feature extraction processing on the data related to the target event through the multi-source data access standardization module to obtain a standard data source.

[0093] For example, the above data cleaning is used to correct the received data. The process involves: performing quality checks and error correction on the received structured, semi-structured, unstructured, and real-time time-series data.

[0094] For example, the above format normalization is used to standardize the format of received data. The process involves converting cleaned data of varying sources and formats into a unified and standardized internal representation.

[0095] For example, the aforementioned time-series alignment is used to align data with time attributes in terms of both time and order. The process involves: for data with time attributes, especially real-time time-series data, a unified and aligned time reference is performed to ensure that data from different sources are comparable and merging on the timeline.

[0096] For example, the feature extraction process described above is used to extract features from the received data that can be used to construct a causal knowledge graph. The process is as follows: from the normalized and time-aligned data, key information and indicators for subsequent knowledge construction and time-series prediction tasks are extracted.

[0097] For example, after the above data cleaning, format normalization, temporal alignment, and feature extraction processes, the original multi-source heterogeneous data is transformed into standardized data with a well-defined structure, reliable quality, consistent time reference, and rich in effective information. This data is then organized and output in two forms: standardized knowledge construction data sources and standardized time series prediction data sources. The standardized knowledge construction data sources focus on providing a well-structured, static or quasi-static dataset with clearly defined entities and relationships for tasks such as knowledge graph construction, rule mining, and pattern discovery. The standardized time series prediction data sources focus on providing a time series dataset with a complete temporal dimension and well-engineered features for tasks such as time series prediction, trend analysis, and anomaly detection.

[0098] Thus, on the one hand, using diversified data sources can improve the accuracy of prediction results; on the other hand, unifying diversified data sources into standardized data sources makes it easier for the model to process them, thereby improving the efficiency of model processing.

[0099] Step 203: The event development trend prediction device extracts the first causal feature information from the standardized knowledge construction data source through the temporal causal knowledge modeling graph module, and constructs a causal knowledge graph based on the first causal feature information.

[0100] In some embodiments, the aforementioned first causal feature information is used to characterize causal information that influences the development trend of an event.

[0101] In some embodiments, the first causal feature information mentioned above includes: a first keyword, causal relationship feature information, and event causal feature information.

[0102] In some embodiments, the first keyword mentioned above is used to characterize the specific content or type of the event.

[0103] For example, the aforementioned first keyword may include, but is not limited to, keywords from at least one of the following types: economic indicators, market entities, time points, equipment types, production factors, transaction categories, regional identifiers, and institution types. Among them, economic indicators cover Gross Domestic Product (GDP), fixed asset investment, and residents' disposable income; market entities cover enterprises, industry associations, and individual businesses; and time points cover the start date of the statistical period, the date of the event, and the date of data collection.

[0104] In some embodiments, the aforementioned causal relationship characteristics include at least one of the following: promoting relationship, inhibiting relationship, and triggering relationship.

[0105] In some embodiments, the above-mentioned event causal feature information includes: event cause element, event result element, and event time correlation element.

[0106] For example, the elements that cause the above-mentioned event include, but are not limited to: the event subject, the event action, and the event occurrence conditions.

[0107] For example, the above-mentioned event outcome elements include, but are not limited to: the affected object, the trend of object change, and the magnitude of change.

[0108] For example, the above-mentioned event time-related elements include, but are not limited to: the time of occurrence of the causal event, the time of occurrence of the result event, and the delay duration of the causal event's impact on the result.

[0109] In some embodiments, the aforementioned causal knowledge graph is a structured knowledge base stored in graph form, containing domain entities (such as events, variables, and states) and the causal relationships between them. Nodes in the graph represent entities, and edges represent causal relationships (such as "cause," "inhibit," or "related to"). Edges may also be appended with attributes such as causal strength, confidence level, and time delay. It serves as the prior knowledge foundation for constructing causal scenarios, used to determine which factors may be causally related to the target event, and to guide the path of counterfactual reasoning.

[0110] In some embodiments, the standardized knowledge construction data source output by the multi-source data access standardization module is used as input. The domain entity, namely the first keyword mentioned above, the causal relationship, namely the causal association relationship feature information mentioned above, and the rule primitive extraction, namely the event causal feature information mentioned above, are executed. A temporal effect function is constructed using a multi-factor temporal coupling causal strength algorithm. A dynamic symbolic knowledge graph is constructed based on the causal rules bound by the event and the temporal effect function. The output is a dynamic symbolic knowledge graph with a temporal effect function, namely the causal knowledge graph mentioned above.

[0111] In some embodiments, combined with Figure 7 ,like Figure 9 As shown, step 203 above can be specifically achieved through steps 203a to 203c.

[0112] Step 203a: The event development trend prediction device extracts the first keyword from the standardized knowledge construction data source based on the preset dictionary and contextual semantic analysis technology.

[0113] In some embodiments, domain entity extraction is performed by combining keyword matching based on an industry-customized dictionary with contextual semantic analysis technology, i.e., the extraction of the first keyword mentioned above.

[0114] Step 203b: The event development trend prediction device extracts causal relationship feature information from standardized knowledge construction data sources based on syntactic analysis relation identification and semantic role labeling technology.

[0115] In some embodiments, causal relationship extraction is performed using relation identification based on dependency parsing combined with semantic role labeling technology.

[0116] For example, the core predicate is determined by parsing the syntactic structure of the text, and the agent and patient corresponding to the predicate are located by combining semantic role labeling, thereby identifying the causal relationship between the agent and patient.

[0117] Step 203c: The event development trend prediction device extracts event causal feature information from a standardized knowledge construction data source based on phrase extraction and time series information extraction technology.

[0118] In some embodiments, phrase extraction based on event trigger word recognition combined with temporal information extraction technology is used to perform rule primitive extraction, that is, the above-mentioned event causal relationship feature information extraction.

[0119] For example, core text fragments of causal events can be located by trigger words, and causal elements, result elements, and time-related elements can be extracted from the fragments.

[0120] Thus, by extracting causal features at different levels, such as keywords, causal relationships, and the causes, results, and timing of events, we can obtain the influencing factors that affect the development trend of events, thereby more accurately predicting the development trend of events.

[0121] Step 204: The event development trend prediction device processes the standardized time series prediction data source and causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event and output it.

[0122] In some embodiments, such as Figure 10 As shown, the event development trend prediction device uses the standardized time-series prediction data source output by the above-mentioned multi-source data access standardization module and the dynamic symbolic knowledge graph output by the temporal causal knowledge modeling graph module. That is, the above-mentioned causal knowledge graph is a dual input. Through a neural network, such as Transformer, a preliminary trend prediction result is generated. Then, the temporal causal constraint prediction correction algorithm is called to correct the preliminary prediction result, and the final dynamic trend prediction result and correction description document are output.

[0123] In some embodiments, combined with Figure 7 ,like Figure 11 As shown, step 204 above can be specifically achieved through steps 204a and 204b.

[0124] Step 204a: The event development trend prediction device generates preliminary event development trend prediction results information through the neural network in the neural-symbolic fusion prediction module.

[0125] In some embodiments, the neural network receives feature-engineered time-series data from a standardized time-series prediction data source as its primary input. It automatically learns complex nonlinear patterns, time dependencies, and latent feature representations from historical time-series data. Based on the learned patterns, it extrapolates the future state, index values, or probability of occurrence of the target event, generating preliminary predictions of the event's development trend.

[0126] For example, the neural network mentioned above can be a recurrent neural network (RNN), a long short-term memory network (LSTM), a gated recurrent unit (GRU), a temporal convolutional network (TCN), or a Transformer based on an attention mechanism.

[0127] Step 204a: The event development trend prediction device uses a temporal causal constraint prediction correction algorithm to correct the preliminary event development trend prediction results and outputs the development trend prediction results of the target event.

[0128] In some embodiments, symbolic logic and causal knowledge are introduced to constrain and correct the above-mentioned data-driven preliminary predictions. After causal logic verification and correction, more reliable and interpretable prediction results of the development trend of the target event are generated that conform to the historical data patterns and satisfy the domain causal logic, and then output.

[0129] In this way, through neural network processing, combined with causal feature information and relevant event data, a preliminary prediction of the event's development trend is first generated. Then, through a correction algorithm, the prediction result is further corrected, thereby improving the accuracy of the prediction result.

[0130] The event development trend prediction method provided in this application uses an event development trend prediction model to predict the event development trend in combination with relevant event data. The process fully considers the causal relationship of event development, thereby improving the accuracy of event development trend prediction.

[0131] In some embodiments, the above event development trend prediction model further includes a temporalized causal attribution module; combined with Figure 7 ,like Figure 12As shown, after step 204 above, the event development trend prediction method provided in this application embodiment may further include the following steps 301 and 302.

[0132] Step 301: The event development trend prediction device constructs a baseline scenario and a counterfactual scenario based on the development trend prediction results information and causal knowledge graph of the target event through the temporal causal attribution module.

[0133] In some embodiments, the aforementioned temporal causal attribution module is used for tracing and quantifying the causes of events. Its core is to combine time dimensions for causal inference, analyzing not only influencing factors but also precisely analyzing the timing, duration, and dynamic changes of these influences. It identifies and quantifies the contribution of each factor to the prediction results within a specific time window by constructing and comparing different scenarios.

[0134] In some embodiments, the aforementioned baseline scenario is also referred to as a "factual scenario." It is a complete scenario that simulates the development path of the real world, constructed based on historical and future data that have been actually observed or predicted. In this scenario, the values ​​and temporal evolution of all relevant variables (especially potential causal variables) are set according to their actual or most likely states, serving as a realistic reference for assessing causal effects.

[0135] In some embodiments, the counterfactual scenario described above is a logically constructed hypothetical scenario used for comparison with a baseline scenario. In this scenario, by selectively intervening in or altering the values ​​or states of one or more antecedent variables considered causally related in the baseline scenario at a specific point in time (e.g., assuming a factor never occurred, occurred with different intensities, or occurred at different times), while simultaneously using a causal knowledge graph to keep the states of other irrelevant variables unchanged, a hypothetical "what if..." event development path is deduced. The counterfactual scenario is a core tool for causal inference; by comparing the differences in the predicted outcomes of the target event under the baseline scenario and the counterfactual scenario, the causal effect of the intervened variable is quantitatively or qualitatively attributed.

[0136] Step 302: The event development trend prediction device calculates the contribution of each influencing factor corresponding to the development trend prediction result information of the target event, generates a structured temporal causal attribution report, and outputs it.

[0137] In some embodiments, the aforementioned influencing factors refer to variables, events, or conditions identified from causal knowledge graphs and data analysis that are considered to have a potential or proven causal relationship with the development trend of the target event. For example, in sales forecasting, price, promotional activities, seasonality, economic indices, etc., can all be considered influencing factors.

[0138] In some embodiments, the contribution of each influencing factor corresponding to the aforementioned trend prediction result information of the target event is a quantitative measure of the extent to which each influencing factor plays a role in forming the final trend prediction result. It indicates the degree to which the presence of the factor or its specific value causes a change in the prediction result (e.g., increase or decrease in value, increase or decrease in probability) compared to a baseline situation (e.g., the absence of the factor or the taking of an average value).

[0139] For example, the above-mentioned contribution is usually calculated based on the difference between the prediction results of the baseline scenario and the counterfactual scenario. Common methods include SHAP value, integral gradient, and counterfactual simulation. The results are usually presented in the form of numerical values ​​(e.g., +5 units), percentages (e.g., contributing 20% ​​of the increase), or in order of importance.

[0140] In some embodiments, the aforementioned structured temporal causal attribution report is an output document or data object that systematically and standardizedly organizes the results of attribution analysis. It has a clear and fixed format (“structured”) that not only describes what the influencing factors are and the magnitude of their contribution, but also precisely links them to the time dimension (“temporal”), clarifying what kind of impact each factor had at different points in time or time periods in the course of the event.

[0141] For example, the core of the structured temporal causal attribution report mentioned above is to reveal the "cause" of changes in things and follow the logical framework of "who-when-how much".

[0142] For example, the typical content of the above-mentioned structured temporal causal attribution report may include, but is not limited to, at least one of the following: description of the target event, prediction time range, description of attribution method, list of influencing factors sorted by contribution, contribution value and trend of each factor at key time points, causal path analysis and summary conclusion.

[0143] Thus, by constructing the baseline scenario and counterfactual scenario through the temporal causal attribution module, the influencing factors affecting the prediction results, i.e. the causal relationship, are calculated, thereby generating an attribution report. Based on this attribution report, the influencing factors affecting the prediction results can be adjusted.

[0144] The event development trend prediction method of this application will be described below through specific embodiments.

[0145] like Figure 13 As shown, taking the event development trend prediction model in this application embodiment as an example, which includes a multi-source data access standardization module, a temporal causal knowledge modeling graph module, a neural-symbolic fusion prediction module, a temporalized causal attribution module, and a knowledge prediction model iterative optimization module, the implementation process of the event development trend prediction method provided in this application embodiment includes the following S1 to S5:

[0146] S1. Receives structured, semi-structured, unstructured, and real-time time-series data from the industry sector through the multi-source data access standardization module, i.e., the relevant data of the aforementioned target events. Performs data cleaning, format normalization, time-series alignment, and feature extraction processing, and outputs standardized knowledge construction data sources and standardized time-series prediction data sources.

[0147] For example, the above-mentioned multi-source data access standardization module adopts a protocol adaptation + channel customization mechanism.

[0148] For example, the structured data mentioned above connects to relational databases via the Java Database Connectivity / Open Database Connectivity (JDBC / ODBC) protocol, and parses Excel / CSV files using Apache POI / OpenCSV tools; semi-structured PDF reports extract text and embedded tables using the Apache PDFBox tool, and policy interpretation documents extract hierarchical structure content using an HTML parser; unstructured text connects to mainstream news platforms via the Hypertext Transfer Protocol Application Programming Interface (HTTP API), and audio-to-text transcripts are obtained through the Automatic Speech Recognition (ASR) tool interface; real-time time-series data is accessed via a message queue.

[0149] For example, the structured data mentioned above includes statistical reports, financial statements, market transaction records, equipment operating parameter tables, and production scheduling data tables. It can be received by connecting to relational databases via JDBC / ODBC protocol, using POI tool to parse Excel format files, or using OpenCSV tool to parse CSV format files.

[0150] For example, the aforementioned semi-structured data includes industry research reports, product manuals, user evaluation records, equipment maintenance logs, and project progress summaries. These can be received by using PDFBox to parse PDF files, HTML parsers to extract hierarchical document content, and JSON and XML parsers to parse the corresponding key-value pair format files respectively.

[0151] For example, the aforementioned unstructured data includes industry news articles, social media discussion texts, transcripts of interviews with corporate executives, subtitles for industry conference videos, and chat logs from technical exchanges. The data is received by connecting to a data publishing platform via HTTP API, accessing the company's internal document library via a local file reading tool, obtaining transcripts via an ASR tool interface, and extracting subtitles via an Optical Character Recognition (OCR) tool interface.

[0152] For example, the aforementioned real-time time-series data includes market transaction time-sharing data, e-commerce platform real-time order data, offline store customer flow statistics, industrial production line sensor real-time data collection data, and network traffic real-time monitoring data. Millisecond-level data access can be achieved through Kafka message queues and Flink streaming computing frameworks.

[0153] S2. The temporal causal knowledge modeling graph module takes the standardized knowledge construction data source output by the multi-source data access standardization module as input, performs domain entity, causal relationship and rule primitive extraction, uses multi-factor temporal coupling causal strength algorithm to construct temporal effect function, constructs dynamic symbolic knowledge graph based on the entity and the causal rules bound to the temporal effect function, and outputs dynamic symbolic knowledge graph with temporal effect function and query API.

[0154] For example, the domain entities, causal relationships and rule primitives extracted above are respectively the first keyword, causal relationship feature information and event causal feature information, which is the first causal feature information mentioned above.

[0155] For example, the above multi-factor temporal coupling causal strength algorithm can be implemented using the following formula (1).

[0156] (1)

[0157] in, For a moment The dynamic causal strength, Based on causal strength, This is a vector of influence from multiple factors. It is a multi-factor weight vector and , For the first Temporal adaptation function for each factor, The cause-and-effect relationship delays the start-up time. For the peak time of causal influence, The dynamic attenuation coefficient, For the improved Sigmoid function and , This represents the slope parameter of the improved Sigmoid function. This represents the input variables of the improved Sigmoid function. For temporal attention weights, This is the historical time series feature matrix.

[0158] For example, the specific process of constructing the dynamic symbolic knowledge graph, i.e., the causal knowledge image, is as follows: The Neo4j graph database is used as the storage medium, and the graph adopts a three-layer structure of entity nodes, temporal causal relationship edges, and attribute information. When constructing entity nodes, a unique identifier following the domain_entity type_unique coding standard is determined, and the corresponding domain entity type and differentiated core attributes are associated. When constructing temporal causal relationship edges, cause entity nodes and result entity nodes are associated, and the attributes include relationship type, multi-factor temporal coupling causal strength algorithm parameter set, dynamic causal strength calculation results at different time nodes, rule confidence, and data source. When constructing attribute information, the metadata of entity nodes and relationship edges is associated. The construction process is as follows: standardized entry of entity nodes, construction of temporal causal relationship edges, attribute information filling, and graph integrity verification. After verification, an index based on the unique identifier of the node, node type, and relationship type is established.

[0159] S3, the neural-symbolic fusion prediction module takes the standardized time-series prediction data source output by the multi-source data access standardization module and the dynamic symbolic knowledge graph output by the temporal causal knowledge modeling graph module as dual inputs. It generates preliminary trend prediction results through the Transformer neural network, calls the temporal causal constraint prediction correction algorithm to correct the preliminary prediction results, and outputs the final dynamic trend prediction results and correction explanation document.

[0160] For example, the Transformer neural network described above employs a 6-layer stacked encoder structure, with each encoder layer containing two core sub-layers: a multi-head self-attention mechanism and a feedforward neural network. The multi-head self-attention mechanism has 8 heads, each with a dimension of 64. It obtains the association weights of temporal features through dot product attention calculation, using a masking mechanism to shield information from future time steps during the calculation process. The feedforward neural network contains two fully connected layers, with the first layer having an output dimension of 2048 and the second layer having an output dimension of 512. The activation function is a Gaussian Error Linear Unit (GELU). The network is configured with a dropout regularization layer with a dropout rate of 0.1, which is embedded in the outputs of the multi-head self-attention mechanism and the feedforward neural network, respectively. The model training uses mean squared error as the loss function, and the optimizer is an Adam with Decoupled Weights optimizer. Decay (AdamW), with an initial learning rate of 1e-4, a weight decay coefficient of 1e-5, and a training batch size of 32; the training process adopts an early stopping strategy, using the prediction error of the validation set as the monitoring indicator. When the validation set error does not decrease for 10 consecutive training rounds, the model training is stopped and the current optimal model parameters are saved.

[0161] For example, the temporal causal constraint prediction correction algorithm described above can be implemented using the following formula (2).

[0162] (2)

[0163] in, For a moment The final corrected prediction value, For a moment The initial predictions from the neural network, For a moment The dynamic causal strength, For causal constraint weighting coefficients, This is a timing deviation adjustment item. For historical causal residual correction coefficients, The time step of the residual effects of historical causality. For the first A historical time step, For the first A historical residual error correction factor.

[0164] S4. The temporal causal attribution module takes the final dynamic trend prediction result output by the neural-symbolic fusion prediction module and the dynamic symbolic knowledge graph output by the temporal causal knowledge modeling graph module as input to construct the baseline scenario and counterfactual scenario, calculate the contribution of influencing factors in stages and throughout the entire cycle, and output a structured temporal causal attribution report.

[0165] For example, the specific contents of the baseline scenario and counterfactual scenario constructed above are as follows: The baseline scenario is constructed based on the final dynamic trend prediction result output by the neural-symbolic fusion prediction module and the dynamic symbolic knowledge graph output by the temporal causal knowledge modeling graph module. In this scenario, all factors affecting the final dynamic trend prediction result play their roles according to the original parameters of the temporal causal rules in the dynamic symbolic knowledge graph. The dynamic causal intensity corresponding to each influencing factor maintains its original calculated value, without any factor removal or parameter adjustment, serving as a benchmark reference for subsequent contribution calculation. The counterfactual scenario includes a full-cycle removal scenario and a phased removal scenario. The full-cycle removal scenario is constructed for a single target influencing factor. The method involves setting the dynamic causal intensity of the target influencing factor to 0 throughout the entire prediction period, while keeping the parameters and operational status of other influencing factors consistent with the baseline scenario. The phased removal scenario is constructed based on the phase division of the temporal effect function parameters in the dynamic symbolic knowledge graph. The phases are divided into a delay period, a peak period, and a decay period. The delay period corresponds to the time range before the delayed start time of the dynamic causal intensity, the peak period corresponds to the time range from the delayed start time to the peak time, and the decay period corresponds to the time range after the peak time. The construction method is to set the dynamic causal intensity of the target influencing factor to 0 in only the above single phase, while keeping the parameters and operational status of other phases and other influencing factors consistent with the baseline scenario.

[0166] For example, the specific process for calculating the phased and full-cycle contribution of the influencing factors is as follows: Phased contribution calculation is based on the prediction results of the baseline scenario and the corresponding phased removal scenario. First, the predicted values ​​for each phase under the baseline scenario and the corresponding phase under the phased removal scenario are extracted. The difference between the two is used to obtain the absolute impact of the target influencing factor on the prediction results in that phase. Then, the absolute impact is compared with the predicted value of the corresponding phase under the baseline scenario to obtain the contribution of the target influencing factor in that phase. The phases are divided into a delay period, a peak period, and a decay period, and the contribution of each phase is calculated independently. Full-cycle contribution calculation is based on the contribution of each phase and the order of the phases. The stage weighting is implemented by determining the stage weight by the ratio of the average dynamic causal intensity of each stage to the average dynamic causal intensity of the whole cycle. First, the average dynamic causal intensity of each stage is calculated, and then the average dynamic causal intensity of the whole cycle is obtained by summing them. The ratio of the average dynamic causal intensity of a single stage to the average dynamic causal intensity of the whole cycle is used as the stage weight. Finally, the contribution of each stage is weighted and summed with the corresponding stage weight to obtain the whole cycle contribution of the target influencing factor. After the calculation is completed, the stage contribution and whole cycle contribution of each influencing factor are sorted, and the influencing factors with an absolute contribution value lower than the preset threshold are removed. The contribution data of the significant influencing factors are retained for subsequent report output.

[0167] S5, the knowledge prediction model iterative optimization module collects real-time updated data from the multi-source data access standardization module, prediction performance data from the neural-symbolic fusion prediction module, attribution feedback data from the temporal causal attribution module, and user feedback information. It evaluates and optimizes the effectiveness of temporal causal rules, algorithm parameters, and prediction models, and feeds back the optimization results to the temporal causal knowledge modeling graph module and the neural-symbolic fusion prediction module.

[0168] Thus, the technical solution provided in this application has the following technical effects:

[0169] 1. The "protocol adaptation + channel customization" mechanism for multi-source data access enables standardized processing of structured, semi-structured, unstructured, and real-time time-series data, ensuring data quality;

[0170] 2. Temporal causal knowledge modeling methods, including a three-order extraction process of domain entities, causal relationships, and rule primitives, a multi-factor temporal coupling causal strength algorithm, and a three-layer dynamic symbolic knowledge graph construction scheme of "entity-relationship-attribute";

[0171] 3. The neural-symbolic fusion prediction architecture captures temporal features through the Transformer neural network and optimizes the prediction results by combining the temporal causal constraint correction algorithm, thereby achieving the synergy of statistical fitting and causal logic;

[0172] 4. A phased and temporal causal attribution mechanism quantifies the contribution of influencing factors at different stages and throughout the entire cycle by comparing baseline scenarios and counterfactual scenarios;

[0173] 5. A closed-loop iterative optimization system dynamically adjusts causal rules, algorithm parameters, and prediction models based on multi-dimensional feedback data to ensure long-term prediction stability.

[0174] It should be noted that the descriptions of each step S1 to S5 in this embodiment can be found in the descriptions in the above embodiments, and will not be repeated here.

[0175] It should be noted that the above-described method embodiments, or the various possible implementations of the method embodiments, can be executed individually, or, provided there is no conflict, they can be combined with each other. The specific implementation can be determined according to actual usage requirements, and this application embodiment does not impose any restrictions on this.

[0176] The following two specific scenarios illustrate the event development trend prediction method provided in the embodiments of this application.

[0177] Scenario 1: Capacity Trend Prediction and Optimization of Smart Manufacturing Production Lines

[0178] The embodiments of this application can be directly applied to the field of intelligent manufacturing, for capacity trend prediction and production scheduling optimization in large-scale production scenarios such as automotive parts and electronic components.

[0179] In this scenario, the event trend prediction device accesses multiple types of data through a "protocol adaptation + channel customization" mechanism: structured data includes equipment operating parameter tables, production scheduling data, and quality inspection statistical reports (connected to the production management database via JDBC protocol and parsed from Excel reports using the POI tool); semi-structured data includes equipment maintenance logs, process improvement minutes, and supplier supply quality reports (parsed from PDF documents using PDFBox and extracted from process documents using an XML parser); unstructured data includes feedback text from frontline operators, transcripts of equipment fault diagnosis recordings, and news about industry technological innovations (connected to the enterprise's internal document library via HTTP API and obtained from the ASR tool); and real-time time-series data includes real-time temperature, pressure, and speed data collected by production line sensors, as well as real-time production cycle statistics (accessed at the millisecond level via Kafka message queue and the Flink framework).

[0180] The event development trend prediction process first cleans, normalizes, and aligns the data to output a standardized data source. Then, through temporal causal knowledge modeling, it extracts causal relationships in areas such as "equipment failure - decreased production efficiency" and "raw material quality fluctuations - increased defect rate," constructing a dynamic symbolic knowledge graph to clarify the impact logic of different factors on production capacity. Subsequently, through neural-symbolic fusion prediction, combined with historical production line capacity data and causal rules, it accurately predicts the production capacity trend for the next 1-3 months, while identifying key influencing factors (such as high-frequency minor faults of certain types of equipment or delays in raw material supply). Through phased attribution analysis, it quantifies the different contributions of these factors in the "problem latency period," "peak impact period," and "attenuation and recovery period," for example, clarifying that equipment failure accounts for 40% of the impact on production capacity during the peak period. Finally, through a closed-loop iterative optimization module, it collects real-time production data, prediction deviation feedback, and adjustment suggestions from workshop schedulers to dynamically optimize causal rules and prediction model parameters, and adjusts equipment maintenance cycles and production scheduling plans in reverse to achieve stable capacity improvement and efficient resource allocation.

[0181] Scenario 2: Regional E-commerce Retail Sales Trend Forecasting and Inventory Allocation

[0182] The embodiments of this application can be widely used in predicting regional sales trends of goods on e-commerce platforms or chain retail enterprises, helping to optimize inventory and make marketing decisions, and are especially suitable for complex market environments such as holiday promotions and seasonal changes.

[0183] In this scenario, the data types accessed by the event trend prediction device include: structured data covering regional historical sales data, product inventory reports, and user consumption profile data (connected to e-commerce transaction databases via ODBC protocol and parsed from CSV format inventory files using OpenCSV); semi-structured data including product detail documentation, user review records, and regional marketing campaign plans (extracted from product detail page content using an HTML parser and processed from user review data using a JSON parser); unstructured data covering product discussion texts on social media, interview captions for regional consumption trend analysis, and news of competitor promotional activities (connected to social media platforms via HTTP API and extracted from interview video captions using OCR tools); and real-time time-series data including regional real-time order data, logistics delivery timeliness data, and real-time statistics on offline store customer traffic (accessed via Kafka to real-time transaction streams).

[0184] After standardization, the event trend prediction device constructs a temporal causal knowledge graph, extracting causal relationships such as "promotional activities - sales growth," "weather changes - seasonal product demand fluctuations," and "logistics delays - decreased repurchase rates," clarifying the temporal characteristics of the impact of different factors. Utilizing neural-symbolic fusion prediction, combined with historical sales patterns and causal constraints, it accurately predicts regional product sales trends for the next 7-30 days, such as predicting the peak sales of a certain home appliance during the National Day promotion. Through phased attribution analysis, it quantifies the sales contribution of promotional activities during the "pre-sale period," "peak period," and "closing period," as well as the auxiliary influence of factors such as weather and logistics. Finally, based on the prediction results and attribution feedback, it optimizes inventory allocation strategies (such as pre-stocking regions with significant sales growth forecasts) and the intensity of promotional activities (such as adjusting advertising channels during the pre-sale period when the contribution is low), while dynamically updating causal rules (such as adding a causal relationship between "live-streaming sales" and sales), effectively reducing inventory backlog and stockout risks, and improving regional market response efficiency.

[0185] As can be seen, the above mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the embodiments of this application provide corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should readily recognize that, in conjunction with the modules and algorithm steps of the various examples described in the embodiments disclosed herein, the embodiments of this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0186] This application embodiment can divide the event development trend prediction device into functional modules according to the above method example. For example, each function can be divided into its own functional module, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. Optionally, the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0187] In some embodiments, this application also provides an event development trend prediction device. This event development trend prediction device may include one or more functional modules for implementing the event development trend prediction method of the above method embodiments.

[0188] For example, Figure 14 This is a schematic diagram of the structure of an event development trend prediction device provided in an embodiment of this application. Figure 14 As shown, the event development trend prediction device 900 includes: an input module 901, a processing module 902, and an output module 903. The input module 901 is used to input data related to the target event into the event development trend prediction model, which includes: a multi-source data access standardization module, a temporal causal knowledge modeling graph module, and a neural-symbolic fusion prediction module. The processing module 902 is used to perform data processing operations on the data related to the target event through the multi-source data access standardization module to obtain a standardized data source, which includes: a standardized knowledge construction data source and a standardized time-series prediction data source. The processing module 902 is also used to extract first causal feature information from the standardized knowledge construction data source through the temporal causal knowledge modeling graph module, and based on... The first causal feature information is used to construct a causal knowledge graph; the first causal feature information is used to characterize the causal information that influences the development trend of the event; the processing module 902 is also used to process the standardized time series prediction data source and the causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event; the output module 903 is used to output the development trend prediction result information of the target event; wherein, the first causal feature information includes: first keyword, causal relationship feature information, and event causal feature information; the first keyword is used to characterize the specific content or type of the event; the causal relationship feature information includes at least one of the following: driving relationship, inhibiting relationship, triggering relationship; the event causal feature information includes: event cause elements, event result elements, and event time relationship elements.

[0189] The technical solution provided in this application brings at least the following beneficial effects: by using an event development trend prediction model and combining relevant event data information to predict the event development trend, the process fully considers the causal relationship of event development, thereby improving the accuracy of event development trend prediction.

[0190] In some embodiments, the above-mentioned processing module 902 is specifically used to perform data cleaning, format normalization, time sequence alignment and feature extraction processing operations on the data related to the target event through the multi-source data access standardization module to obtain a standard data source;

[0191] The data related to the aforementioned target event includes at least one of the following types of data:

[0192] Structured data;

[0193] Semi-structured data;

[0194] Unstructured data;

[0195] Real-time time series data.

[0196] In other embodiments, the processing module 902 described above is specifically used to: extract the first keyword from the standardized knowledge construction data source based on a preset dictionary and contextual semantic analysis technology; extract causal relationship feature information from the standardized knowledge construction data source based on syntactic analysis relation identification and semantic role labeling technology; and extract event causal feature information from the standardized knowledge construction data source based on phrase extraction and temporal information extraction technology.

[0197] In some other embodiments, the processing module 902 is specifically used to: generate preliminary event development trend prediction results information through the neural network in the neural-symbolic fusion prediction module; correct the preliminary event development trend prediction results information by using a temporal causal constraint prediction correction algorithm; and output the development trend prediction results information of the target event.

[0198] In some embodiments, the above-mentioned event development trend prediction model further includes a temporal causal attribution module; the processing module 902 is further configured to, after processing the standardized time-series prediction data source and causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event, construct a baseline scenario and a counterfactual scenario based on the development trend prediction result information of the target event and the causal knowledge graph through the temporal causal attribution module; the processing module 902 is further configured to calculate the contribution of each influencing factor corresponding to the development trend prediction result information of the target event and generate a structured temporal causal attribution report; the output module 903 is further configured to output the structured temporal causal attribution report.

[0199] It should be noted that the event development trend prediction device can implement all the processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0200] In the case where the functions of the integrated modules described above are implemented in hardware, this application provides a possible structural schematic diagram of the electronic device involved in the above embodiments. For example... Figure 15 As shown, the electronic device 90 includes: a processor 92, a communication interface 93, and a bus 94. Optionally, the electronic device 90 may also include a memory 91.

[0201] Processor 92 may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 92 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0202] Communication interface 93 is used to connect with other devices via a communication network. This communication network can be Ethernet, wireless access network, wireless local area network (WLAN), etc.

[0203] The memory 91 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0204] As one possible implementation, the memory 91 can exist independently of the processor 92. The memory 91 can be connected to the processor 92 via a bus 94 and is used to store instructions or program code. When the processor 92 calls and executes the instructions or program code stored in the memory 91, it can implement the event development trend prediction method provided in the embodiments of this application.

[0205] In another possible implementation, memory 91 can also be integrated with processor 92.

[0206] Bus 94 can be an Extended Industry Standard Architecture (EISA) bus, etc. Bus 94 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 15 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0207] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the service calling device can be divided into different functional modules to complete all or part of the functions described above.

[0208] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described event development trend prediction method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0209] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0210] This application also provides a readable storage medium storing a program or instructions that, when executed by a computer, implement the event trend prediction method provided in the above embodiments. It is understood that all or part of the processes in the above method embodiments can be executed by computer instructions instructing related hardware; the readable storage medium can be any of the foregoing embodiments or memory; the readable storage medium can also be an external storage device of the service invocation device, such as a pluggable hard drive, SmartMedia Card (SMC), Secure Digital (SD) card, flash card, etc., equipped on the service invocation device. Further, the readable storage medium can include both internal storage units of the service invocation device and external storage devices. The readable storage medium is used to store the computer program and other programs and data required by the service invocation device. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0211] This application also provides a computer program product, which is stored in a storage medium and, when executed by a computer, implements the event development trend prediction method provided in the above embodiments.

[0212] It should be noted that, in this document, 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 that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0213] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0214] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for predicting the development trend of an event, characterized in that, The method includes: The data related to the target event are input into the event development trend prediction model, which includes: a multi-source data access standardization module, a temporal causal knowledge modeling graph module, and a neural-symbolic fusion prediction module. The multi-source data access standardization module performs data processing operations on the data related to the target event to obtain a standardized data source, which includes: a standardized knowledge construction data source and a standardized time series prediction data source. The temporal causal knowledge modeling graph module extracts first causal feature information from the standardized knowledge construction data source and constructs a causal knowledge graph based on the first causal feature information; the first causal feature information is used to characterize causal information that influences the development trend of events. The neural-symbolic fusion prediction module processes the standardized time-series prediction data source and the causal knowledge graph to obtain the development trend prediction result information of the target event, and outputs it. The first causal feature information includes: a first keyword, causal relationship feature information, and event causal feature information; the first keyword is used to characterize the specific content or type of the event; the causal relationship feature information includes at least one of the following: promoting relationship, inhibiting relationship, and triggering relationship; the event causal feature information includes: event cause element, event result element, and event time relationship element.

2. The event development trend prediction method according to claim 1, characterized in that, The step of performing data processing operations on the data related to the target event through the multi-source data access standardization module to obtain a standardized data source includes: The multi-source data access standardization module performs data cleaning, format normalization, time sequence alignment, and feature extraction on the data related to the target event to obtain the standard data source. The data related to the target event includes at least one of the following types of data: Structured data; Semi-structured data; Unstructured data; Real-time time series data.

3. The event development trend prediction method according to claim 1, characterized in that, The step of extracting information related to the first causal feature from the standardized knowledge construction data source through the temporal causal knowledge modeling graph module includes: Based on a pre-defined dictionary and contextual semantic analysis technology, the first keyword is extracted from the standardized knowledge construction data source; Based on syntactic analysis-based relation identification and semantic role labeling technology, the causal association feature information is extracted from the standardized knowledge construction data source. Based on phrase extraction and time-series information extraction techniques, the causal feature information of the event is extracted from the standardized knowledge construction data source.

4. The event development trend prediction method according to claim 1, characterized in that, The process of processing the standardized time-series prediction data source and the causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event includes: The neural network in the neural-symbolic fusion prediction module generates preliminary prediction results of event development trends. The preliminary event development trend prediction results are corrected using a temporal causal constraint prediction correction algorithm, and the development trend prediction results of the target event are output.

5. The event development trend prediction method according to any one of claims 1 to 4, characterized in that, The event development trend prediction model further includes a temporalized causal attribution module; after processing the standardized temporal prediction data source and the causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event, the method further includes: The temporal causal attribution module constructs a baseline scenario and a counterfactual scenario based on the development trend prediction results of the target event and the causal knowledge graph. The system calculates the contribution of each influencing factor corresponding to the predicted trend of the target event, generates a structured temporal causal attribution report, and outputs it.

6. An event development trend prediction device, characterized in that, The device includes: an input module, a processing module, and an output module; The input module is used to input data related to the target event into the event development trend prediction model. The event development trend prediction model includes: a multi-source data access standardization module, a temporal causal knowledge modeling graph module, and a neural-symbolic fusion prediction module. The processing module is used to perform data processing operations on the data related to the target event through the multi-source data access standardization module to obtain a standardized data source. The standardized data source includes: a standardized knowledge construction data source and a standardized time series prediction data source. The processing module is further configured to extract first causal feature information from the standardized knowledge construction data source through the temporal causal knowledge modeling graph module, and construct a causal knowledge graph based on the first causal feature information; the first causal feature information is used to characterize causal information that influences the development trend of events; The processing module is further configured to process the standardized time-series prediction data source and the causal knowledge graph through the neural-symbolic fusion prediction module to obtain the development trend prediction result information of the target event; The output module is used to output the prediction result information of the development trend of the target event; The first causal feature information includes: a first keyword, causal relationship feature information, and event causal feature information; the first keyword is used to characterize the specific content or type of the event; the causal relationship feature information includes at least one of the following: promoting relationship, inhibiting relationship, and triggering relationship; the event causal feature information includes: event cause element, event result element, and event time relationship element.

7. The event development trend prediction device according to claim 6, characterized in that, The processing module is specifically used to perform data cleaning, format normalization, time sequence alignment and feature extraction processing operations on the data related to the target event through the multi-source data access standardization module to obtain the standard data source. The data related to the target event includes at least one of the following types of data: Structured data; Semi-structured data; Unstructured data; Real-time time series data.

8. The event development trend prediction device according to claim 6, characterized in that, The processing module is specifically used for: Based on a pre-defined dictionary and contextual semantic analysis technology, the first keyword is extracted from the standardized knowledge construction data source; Based on syntactic analysis-based relation identification and semantic role labeling technology, the causal association feature information is extracted from the standardized knowledge construction data source. Based on phrase extraction and time-series information extraction techniques, the causal feature information of the event is extracted from the standardized knowledge construction data source.

9. The event development trend prediction device according to claim 6, characterized in that, The processing module is specifically used for: The neural network in the neural-symbolic fusion prediction module generates preliminary prediction results of event development trends. The preliminary event development trend prediction results are corrected using a temporal causal constraint prediction correction algorithm, and the development trend prediction results of the target event are output.

10. The event development trend prediction device according to any one of claims 6 to 9, characterized in that, The event development trend prediction model also includes a temporalized causal attribution module; and the neural-symbolic fusion prediction module is used in the process. The processing module is further configured to process the standardized time-series prediction data source and the causal knowledge graph to obtain the development trend prediction result information of the target event, and then construct a baseline scenario and a counterfactual scenario based on the development trend prediction result information of the target event and the causal knowledge graph through the temporalized causal attribution module. The processing module is also used to calculate the contribution of each influencing factor corresponding to the development trend prediction result information of the target event, and generate a structured temporal causal attribution report; The output module is also used to output a structured temporal causal attribution report.

11. An electronic device, characterized in that, It includes a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions being executed by the processor to implement the event development trend prediction method as described in any one of claims 1 to 5.

12. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a computer, implement the event trend prediction method as described in any one of claims 1 to 5.

13. A computer program product, characterized in that, The computer program product is stored in a storage medium, and when executed by a computer, the computer program product implements the event development trend prediction method as described in any one of claims 1 to 5.

Images