Knowledge base-oriented interactive causal inference support method and system

Abstract

According to the knowledge base-oriented interactive causal inference support method and system provided by the invention, a user causal analysis intention is precisely limited through interactive dialogue, a multi-dimensional intention is precisely captured, a fuzzy causal problem is converted into a computable analysis parameter, and ambiguity is remarkably reduced; multiple strategies such as an explicit causal declaration mode, time sequence correlation, statistics of significant correlation and comparative difference analysis are integrated to collect causal evidence, the four strategies are executed in parallel, multiple evidence sources of causal inference are covered, and limitation of a single method is avoided; and the query efficiency is optimized in combination with a special index and a graph algorithm, and finally evidence is presented in a structured and interactive mode and confidence is clearly distinguished, so that the response efficiency is greatly improved. The system solves the problems of fuzzy intention, single evidence and low efficiency in traditional causal inference, is suitable for the fields of industrial fault analysis, medical diagnosis, social phenomenon research and the like, and provides scientific and transparent causal hypothesis generation support for users.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence processing technology, and more specifically, to a knowledge base-oriented interactive causal inference support method and system. Background Art

[0002] A knowledge graph (KG) is a structured semantic knowledge base that describes concepts and their relationships in the physical world in symbolic form. Its basic building blocks are entity-relationship-entity triples and the entity's associated attribute-value pairs. Entities are connected through relationships, forming a network-like knowledge graph.

[0003] The technology closest to the existing one is publication number: CN114328661A, event processing, knowledge base processing method, device, equipment and storage medium, which discloses a specific implementation scheme: obtaining a first event and a first reasoning type, the first reasoning type including a cause type and / or a result type; searching for the first event in a knowledge base, the knowledge base including multiple events and causal relationships between the multiple events; in response to finding the first event, determining a target event in the knowledge base based on the first event and the first reasoning type, the target event including: a cause event and / or a result event of the first event.

[0004] The existing processing method has the following defects:

[0005] (1) Single reasoning strategy, relying on explicit causal claim chains: Patents rely solely on explicit causal claim relationships predefined in the knowledge base (e.g., causal event pairs extracted from text). If direct or indirect causal chains do not exist in the knowledge base, reasoning cannot be performed.

[0006] (2) Insufficient interactivity and poor parameter flexibility: User input is limited to “event + reasoning type” (e.g., cause / result), and it is not possible to customize the time window, domain range, or control group. Parameter generation lacks guided interaction, which limits its applicability to simple queries.

[0007] (3) The confidence rules are simple and do not integrate multi-source evidence: The confidence calculation is based only on the confidence of the evidence text (such as source reliability), without stratifying or integrating different types of evidence (such as statistically significant correlation, temporal changes), and lacks the comprehensive use of explicit causal statement knowledge, time series data, and comparative evidence.

[0008] (4) Lack of efficiency optimization: No special optimization is performed for time attributes and graph structure queries in causal analysis, resulting in a significant performance degradation in large-scale knowledge bases.

[0009] In view of this, there is an urgent need for an interactive causal inference support method and system for knowledge bases that is interaction-centric, multi-strategy evidence-driven, and efficient. Summary of the Invention

[0010] The purpose of the present invention is to propose an interactive causal inference support method and system for knowledge bases that is centered on interaction, driven by multi-strategy evidence, and takes efficiency into consideration.

[0011] To achieve the above object, the present invention provides a knowledge base-oriented interactive causal inference support method, characterized by comprising the following steps:

[0012] Step 1: Obtain the user's causal analysis intention through interactive dialogue, and collect effect description, time range, and domain preference parameters;

[0013] Step 2: Based on the parameters collected in Step 1, execute at least two strategies from the knowledge base in parallel: explicit causal claim pattern inference, time series association analysis, statistically significant correlation analysis, and comparative evidence discovery to generate multiple types of causal evidence;

[0014] Step 3: Create a dedicated index for the time attributes and event sequences in the knowledge base, and use a dedicated graph algorithm to optimize the efficiency of causal queries;

[0015] Step 4: Classify and summarize multi-strategy evidence by type, evaluate the confidence level, and visualize it through an interactive interface to assist users in generating causal hypotheses.

[0016] In some embodiments, in Step 1, the interactive dialogue is implemented through guided questions based on predefined templates to collect at least one of the following: effect clarification: the effect phenomenon that the user is concerned about and the time of occurrence; time range: the time window before the effect occurs; relationship type: direct cause, indirect factor or necessary condition; field limitation: environment, technology or organizational behavior field; evidence preference: explicit causal statement, statistically significant correlation or case comparison.

[0017] In some embodiments, in Step 1: a visualization tool is used to assist the user in performing the following steps: marking the effect occurrence point and the focus window on the timeline; checking the potential cause domain through the domain ontology tree; and inputting known intervention events.

[0018] In some embodiments, Step 1 also includes iterative range adjustment. After collecting preliminary evidence, it is determined whether to adjust the range. If the range does not need to be adjusted, the process proceeds to Step 2. If the range needs to be adjusted, the process returns to Step 1 for iterative adjustment.

[0019] In some embodiments, in Step 2: explicit causal statement pattern inference includes: starting from the effect entity, tracing back the direct or indirect cause along the caused relationship in the knowledge base; pruning the causal path according to the maximum chain length ≤ 3 steps and the domain set set by the user; and outputting a causal path diagram with confidence and time constraints.

[0020] Furthermore, time series association analysis includes: extracting the state change events of the target domain entities within the time window before the effect occurs; if there is an intervention event, comparing the change values ​​of the effect indicators before and after the intervention.

[0021] In some implementations, in Step 3, the dedicated index and graph algorithm includes:

[0022] Establish a B+ tree index for time attributes; build a prefix tree Trie for event sequences; use reverse breadth-first search with domain constraints to perform common cause discovery; calculate confidence transfer based on independent or cumulative relationships for multi-step causal chains; cache high-frequency subquery results.

[0023] In some embodiments, in Step 4, the evidence classification and summary includes: classifying the evidence by explicit causal claim, temporally related events, statistically significant correlation, and comparative difference; wherein the confidence ranking is arranged as follows: explicit causal claim > statistically significant correlation > temporally related events > comparative difference, specifically as follows:

[0024] Furthermore, the same-level confidence refinement rules are as follows:

[0025] Explicit causal statements: sorted by confidence in the relationship × chain length correction factor (chain length correction factor 1.0 for 1 step, 0.9 for 2 steps, 0.8 for ≥3 steps);

[0026] Statistically significant correlation: industrial scenarios prioritize r values, medical scenarios prioritize OR values;

[0027] Time-related events: sorted by Δt proximity (time proximity), with smaller Δt appearing first;

[0028] Comparing differences: Sort by ascending p-value and descending effect size (e.g., p=0.001 and OR=5.3 > p=0.01 and OR=3.2).

[0029] Furthermore, when the domains are different, the confidence ranking is adaptively adjusted across domains.

[0030] In the industrial field: explicit causal statement > statistically significant correlation > temporally related events > comparative differential classification evidence;

[0031] In the medical field: comparative difference classification evidence (between-group studies) > explicit causal statement (medical guidelines) > time series > statistically significant association (univariate analysis).

[0032] In addition, to achieve the above-mentioned object, the present invention further provides a knowledge base-oriented interactive causal inference support system, which is characterized by comprising four modules connected in sequence:

[0033] The interactive causal query intent understanding and scope limitation module has an input connected to the user terminal to receive the original query and visualization operations, and an output connected to the input of the multi-strategy causal evidence collection and generation module to transmit structured analysis parameters;

[0034] The multi-strategy causal evidence collection and generation module has its input connected to the output of the interactive causal query intention understanding and scope limitation module, receiving analysis parameters, and its output connected to the input of the causal evidence synthesis and explanatory presentation module, transmitting four types of evidence sets;

[0035] The causal query execution optimization module has its input connected to the query request interface of the multi-strategy causal evidence collection and generation module.

[0036] Its output end is connected to the result receiving interface of the multi-strategy causal evidence collection and generation module to return the optimization result;

[0037] The causal evidence synthesis and explanatory presentation module has its input end connected to the output end of the multi-strategy causal evidence collection and generation module, receives the original evidence set, and its output end is connected to the user terminal and returns to the visual interactive interface.

[0038] In some embodiments, the interactive causal query intent understanding and scope limitation module includes a template engine and a visualization tool. The template engine is used to load the guidance question template after detecting causal keywords, and the visualization tool is used to provide timeline annotation, domain ontology tree checking, and intervention event input.

[0039] In some embodiments, the multi-strategy causal evidence collection and generation module includes an explicit causal statement pattern inference unit, a time series association analysis unit, a statistically significant correlation analysis unit, and a comparative evidence discovery unit.

[0040] Furthermore, the four units included in the multi-strategy causal evidence collection and generation module adopt a parallel working mode: the input end of the explicit causal declaration pattern inference unit is connected to the domain set parameter interface; the input end of the inter-sequence association analysis unit is connected to the time window parameter interface; the input end of the statistically significant correlation analysis unit is connected to the evidence type preference parameter interface; and the input end of the comparative evidence discovery unit is connected to the intervention event parameter interface.

[0041] In some embodiments, the causal query execution optimization module includes a time attribute index unit, an event sequence index unit, a graph algorithm execution unit, and a result cache unit. The time attribute index unit is used to manage the B+ tree index, the event sequence index unit is used to manage the prefix tree Trie, the graph algorithm execution unit is used to perform reverse breadth-first search, and the result cache unit is used to store high-frequency subquery results.

[0042] In some embodiments, the causal evidence synthesis and explanatory presentation module includes an evidence classification and sorting unit and an interactive visualization unit. The evidence classification and sorting unit is used to classify by type and sort by confidence, and the interactive visualization unit is used to provide a drill-down function and a timeline display.

[0043] Beneficial effects of the present invention: The present invention proposes a knowledge base-oriented interactive causal inference support method and system, which accurately defines the user's causal analysis intention through interactive dialogue, accurately captures multi-dimensional intentions, converts fuzzy causal problems into computable analysis parameters, and significantly reduces ambiguity; integrates multiple strategies such as explicit causal declaration patterns, time series associations, statistically significant correlations, and comparative difference analysis to collect causal evidence, and four types of strategies are executed in parallel to cover multiple sources of evidence for causal inference, avoiding the limitations of a single method; and combines dedicated indexes with graph algorithms to optimize query efficiency, ultimately presenting evidence in a structured, interactive manner and clearly distinguishing confidence levels, greatly improving response efficiency. This system solves the problems of fuzzy intentions, single evidence, and low efficiency in traditional causal inference, and is suitable for industrial failure analysis, medical diagnosis, social phenomenon research, and other fields, providing users with scientific and transparent support for generating causal hypotheses. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a structural diagram of a knowledge base-oriented interactive causal inference support method of this application.

[0045] Figure 2 This is a schematic diagram of the structure of a knowledge base-oriented interactive causal inference support system of this application.

[0046] Figure 3 This is a structural diagram of an interactive causal query intention understanding and scope limitation module of a knowledge base-oriented interactive causal inference support system of the present application.

[0047] Figure 4 This is a structural diagram of an interactive multi-strategy causal evidence collection and generation module of a knowledge base-oriented interactive causal inference support system of this application.

[0048] Figure 5 This is a structural diagram of an interactive causal query execution optimization module of a knowledge base-oriented interactive causal inference support system of the present application.

[0049] Figure 6 This is a structural diagram of an interactive causal evidence synthesis and explanatory presentation module of a knowledge base-oriented interactive causal inference support system of this application.

[0050] Figure 7 This is a schematic diagram of the Step 1 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application.

[0051] Figure 8 This is a schematic diagram of the Step 2 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application.

[0052] Figure 9 This is a schematic diagram of the Step 3 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application.

[0053] Figure 10 This is a schematic diagram of the Step 4 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application.

[0054] The following specific embodiments will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0055] The following examples are described to assist in understanding the present application, and the examples are not and should not be interpreted in any way as limiting the scope of protection of the present application.

[0056] In the following description, those skilled in the art will recognize that throughout this discussion, components may be described as separate functional units (which may include sub-units), but those skilled in the art will recognize that various components or portions thereof may be divided into separate components or may be integrated together (including within a single system or component).

[0057] At the same time, the connections between components or systems are not intended to be limited to direct connections. Instead, data between these components may be modified, reformatted, or otherwise changed by intermediate components. In addition, additional or fewer connections may be used. It should also be noted that the terms "coupled," "connected," or "input" should be understood to include direct connections, indirect connections through one or more intermediate devices, and wireless connections. Example 1:

[0058] like Figure 1 As shown in FIG, a schematic diagram of the structure of a knowledge base-oriented interactive causal inference support method of the present application is shown; Figure 7 As shown, this is a schematic diagram of the Step 1 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application; Figure 8As shown, this is a schematic diagram of the Step 2 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application; Figure 9 As shown, this is a schematic diagram of the Step 3 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application; Figure 10 As shown, it is a schematic diagram of the Step 4 local amplification process structure of an interactive causal inference support method for a knowledge base of this application.

[0059] A knowledge base-oriented interactive causal inference support method, characterized by comprising the following steps:

[0060] Step 1: Obtain the user's causal analysis intention through interactive dialogue, and collect effect description, time range, and domain preference parameters;

[0061] Step 2: Based on the parameters collected in Step 1, execute at least two strategies from the knowledge base in parallel: explicit causal claim pattern inference, time series association analysis, statistically significant correlation analysis, and comparative evidence discovery to generate multiple types of causal evidence;

[0062] Step 3: Create a dedicated index for the time attributes and event sequences in the knowledge base, and use a dedicated graph algorithm to optimize the efficiency of causal queries;

[0063] Step 4: Classify and summarize multi-strategy evidence by type, evaluate the confidence level, and visualize it through an interactive interface to assist users in generating causal hypotheses.

[0064] In Step 1, the interactive dialogue is implemented through guided questions based on predefined templates to collect at least one of the following: effect clarification: the effect phenomenon that the user is concerned about and the time of occurrence; time range: the time window before the effect occurs; relationship type: direct cause, indirect factor or necessary condition; field limitation: environment, technology or organizational behavior field; evidence preference: explicit causal statement, statistically significant correlation or case comparison.

[0065] In Step 1: Use visualization tools to assist users in performing the following steps: mark the effect occurrence point and the focus window on the timeline; select the potential cause domain through the domain ontology tree; and input known intervention events.

[0066] Step 1 also includes iterative range adjustment. After collecting preliminary evidence, it is confirmed whether to adjust the range. If the range does not need to be adjusted, it goes to Step 2. If the range needs to be adjusted, it returns to Step 1 for iterative adjustment.

[0067] In Step 2: Explicit causal statement pattern inference includes: starting from the effect entity, tracing back the direct or indirect cause along the caused relationship in the knowledge base; pruning the causal path according to the user-set maximum chain length ≤ 3 steps and domain set; and outputting a causal path diagram with confidence and time constraints.

[0068] Time series association analysis includes: extracting the state change events of the target domain entities within the time window before the effect occurs; if there is an intervention event, comparing the change values ​​of the effect indicators before and after the intervention.

[0069] In Step 3, the dedicated index and graph algorithm includes:

[0070] Establish a B+ tree index for time attributes; build a prefix tree Trie for event sequences; use reverse breadth-first search with domain constraints to perform common cause discovery; calculate confidence transfer based on independent or cumulative relationships for multi-step causal chains; cache high-frequency subquery results.

[0071] In Step 4, the evidence classification and summary includes: classifying evidence by explicit causal claim, temporally related events, statistically significant correlation, and comparative difference; wherein, the confidence ranking is arranged as follows: explicit causal claim > statistically significant correlation > temporally related events > comparative difference, as follows:

[0072] When it comes to industrial failure analysis, Step 1: The specific implementation steps are as follows:

[0073] Effect clarification: The user inputs "The gearbox of wind turbine #3 shuts down abnormally on June 15, 2024." The system prompts "Was the shutdown accompanied by abnormal noise / temperature?" The user confirms "Monitoring showed that the gearbox oil temperature exceeded 90°C 1 hour before the shutdown."

[0074] Time range: The user marks the effect occurrence time as 14:30 through the timeline tool and selects the potential cause time window as "24 hours before the effect occurs" (14:30 June - 15:30 June).

[0075] Domain limitation: Check "Mechanical transmission → Gearbox → Bearing assembly" through the domain ontology tree and exclude "Electrical system".

[0076] Evidence preference: Select "Explicit causal claims first, statistically significant associations second".

[0077] Step 2: The specific implementation steps are as follows:

[0078] Explicit causal statement pattern inference: Starting from "gearbox abnormal shutdown", trace back the knowledge base causal chain:

[0079] The abnormal gearbox shutdown was traced back to bearing overheating (confidence level 0.92, time constraint [June 15, 13:00-14:30]) and then to lubricant contamination (confidence level 0.85, time constraint [June 14, 10:00-June 15, 13:00]). The chain length was 2 steps, meeting the user-set requirement of ≤ 3 steps.

[0080] Statistically significant correlation analysis: Gearbox related variables within the extracted time window: lubricating oil iron content (V1), bearing vibration amplitude (V2), oil temperature (V3). Calculation:

[0081] The Pearson correlation coefficient between V3 and shutdown status was r=0.89, p<0.001;

[0082] The chi-square test between V1 and the shutdown state was p=0.03, and the effect size was Cramer's V=0.65.

[0083] Time series correlation analysis: The "lubricating oil filter blockage alarm" event was detected at 12:45 on June 15 (belonging to the user-defined bearing component area). Comparison found that the oil temperature suddenly increased from 75°C to 92°C after the alarm, which is consistent with the analysis of indicator changes before and after the intervention.

[0084] Step 3: The specific implementation steps are as follows:

[0085] Time attribute index quickly locates all device events from June 14th to 15th;

[0086] Cache the results of "Gearbox Oil Temperature Historical Data Query" to speed up subsequent statistical analysis.

[0087] Step 4: The specific implementation steps are as follows:

[0088] Evidence Classification:

[0089] Explicit causal statement: Lubricant contamination → Bearing overheating → Shutdown (Confidence level 0.92 × 0.85 = 0.78);

[0090] Statistically significant correlation: oil temperature and shutdown (r=0.89, p<0.001);

[0091] Time-related events: The temporal proximity (Δt=1.75h) between the filter blockage alarm (12:45 on June 15) and the shutdown (14:30).

[0092] Confidence ranking: explicit causal chain (0.78) > statistically significant correlation (r=0.89) > temporally related events (Δt<2h).

[0093] like Figure 2 , which is a schematic diagram of the structure of a knowledge base-oriented interactive causal inference support system of the present application;

[0094] An interactive causal inference support system for a knowledge base, characterized by comprising four modules connected in sequence:

[0095] The interactive causal query intent understanding and scope limitation module has an input connected to the user terminal to receive the original query and visualization operations, and an output connected to the input of the multi-strategy causal evidence collection and generation module to transmit structured analysis parameters;

[0096] The multi-strategy causal evidence collection and generation module has its input connected to the output of the interactive causal query intention understanding and scope limitation module, receiving analysis parameters, and its output connected to the input of the causal evidence synthesis and explanatory presentation module, transmitting four types of evidence sets;

[0097] The causal query execution optimization module has its input connected to the query request interface of the multi-strategy causal evidence collection and generation module.

[0098] Its output end is connected to the result receiving interface of the multi-strategy causal evidence collection and generation module to return the optimization result;

[0099] The causal evidence synthesis and explanatory presentation module has its input end connected to the output end of the multi-strategy causal evidence collection and generation module, receives the original evidence set, and its output end is connected to the user terminal and returns to the visual interactive interface.

[0100] like Figure 3 As shown, this is a structural diagram of an interactive causal query intention understanding and scope limitation module of an interactive causal inference support system for a knowledge base of the present application. The interactive causal query intention understanding and scope limitation module includes a template engine and a visualization tool. The template engine is used to load the guidance question template after detecting causal keywords, and the visualization tool is used to provide timeline annotation, domain ontology tree checking and intervention event input.

[0101] like Figure 4 As shown, this is a structural diagram of an interactive multi-strategy causal evidence collection and generation module of a knowledge base-oriented interactive causal inference support system of the present application. The multi-strategy causal evidence collection and generation module includes an explicit causal statement pattern inference unit, a time series association analysis unit, a statistically significant correlation analysis unit, and a comparative evidence discovery unit.

[0102] The multi-strategy causal evidence collection and generation module includes four units that work in parallel: the input end of the explicit causal statement pattern inference unit is connected to the domain set parameter interface; the input end of the inter-sequence association analysis unit is connected to the time window parameter interface; the input end of the statistically significant correlation analysis unit is connected to the evidence type preference parameter interface; and the input end of the comparative evidence discovery unit is connected to the intervention event parameter interface.

[0103] like Figure 5 As shown, this is a structural diagram of an interactive causal query execution optimization module of an interactive causal inference support system for a knowledge base of the present application. The causal query execution optimization module includes a time attribute index unit, an event sequence index unit, a graph algorithm execution unit, and a result cache unit. The time attribute index unit is used to manage the B+ tree index, the event sequence index unit is used to manage the prefix tree Trie, the graph algorithm execution unit is used to perform a reverse breadth-first search, and the result cache unit is used to store high-frequency subquery results.

[0104] like Figure 6 As shown, this is a structural diagram of an interactive causal evidence synthesis and explanatory presentation module of a knowledge base-oriented interactive causal inference support system of the present application. The causal evidence synthesis and explanatory presentation module includes an evidence classification and sorting unit and an interactive visualization unit. The evidence classification and sorting unit is used to classify by type and sort by confidence, and the interactive visualization unit is used to provide a drill-down function and a timeline display. Example 2:

[0105] like Figure 1 As shown in FIG, a schematic diagram of the structure of a knowledge base-oriented interactive causal inference support method of the present application is shown; Figure 7 As shown, this is a schematic diagram of the Step 1 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application; Figure 8 As shown, this is a schematic diagram of the Step 2 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application; Figure 9 As shown, this is a schematic diagram of the Step 3 local amplification process structure of a knowledge base-oriented interactive causal inference support method of this application; Figure 10 As shown, it is a schematic diagram of the Step 4 local amplification process structure of an interactive causal inference support method for a knowledge base of this application.

[0106] A knowledge base-oriented interactive causal inference support method, characterized by comprising the following steps:

[0107] Step 1: Obtain the user's causal analysis intention through interactive dialogue, and collect effect description, time range, and domain preference parameters;

[0108] Step 2: Based on the parameters collected in Step 1, execute at least two strategies from the knowledge base in parallel: explicit causal claim pattern inference, time series association analysis, statistically significant correlation analysis, and comparative evidence discovery to generate multiple types of causal evidence;

[0109] Step 3: Create a dedicated index for the time attributes and event sequences in the knowledge base, and use a dedicated graph algorithm to optimize the efficiency of causal queries;

[0110] Step 4: Classify and summarize multi-strategy evidence by type, evaluate the confidence level, and visualize it through an interactive interface to assist users in generating causal hypotheses.

[0111] In Step 1, the interactive dialogue is implemented through guided questions based on predefined templates to collect at least one of the following: effect clarification: the effect phenomenon that the user is concerned about and the time of occurrence; time range: the time window before the effect occurs; relationship type: direct cause, indirect factor or necessary condition; field limitation: environment, technology or organizational behavior field; evidence preference: explicit causal statement, statistically significant correlation or case comparison.

[0112] In Step 1: Use visualization tools to assist users in performing the following steps: mark the effect occurrence point and the focus window on the timeline; select the potential cause domain through the domain ontology tree; and input known intervention events.

[0113] Step 1 also includes iterative range adjustment. After collecting preliminary evidence, it is confirmed whether to adjust the range. If the range does not need to be adjusted, it goes to Step 2. If the range needs to be adjusted, it returns to Step 1 for iterative adjustment.

[0114] In Step 2: Explicit causal statement pattern inference includes: starting from the effect entity, tracing back the direct or indirect cause along the caused relationship in the knowledge base; pruning the causal path according to the user-set maximum chain length ≤ 3 steps and domain set; and outputting a causal path diagram with confidence and time constraints.

[0115] Time series association analysis includes: extracting the state change events of the target domain entities within the time window before the effect occurs; if there is an intervention event, comparing the change values ​​of the effect indicators before and after the intervention.

[0116] In Step 3, the dedicated index and graph algorithm includes:

[0117] Establish a B+ tree index for time attributes; build a prefix tree Trie for event sequences; use reverse breadth-first search with domain constraints to perform common cause discovery; calculate confidence transfer based on independent or cumulative relationships for multi-step causal chains; cache high-frequency subquery results.

[0118] In Step 4, the evidence classification and summary includes: classifying evidence by explicit causal claim, temporally related events, statistically significant correlation, and comparative difference; wherein, the confidence ranking is arranged as follows: explicit causal claim > statistically significant correlation > temporally related events > comparative difference, as follows:

[0119] When it comes to medical analysis, Step 1: The specific implementation steps are as follows:

[0120] Effect clarification: The user inputs "55-year-old male patient diagnosed with type 2 diabetes", and the system prompts to add "Glycated hemoglobin HbA1c=8.5%, BMI=32.5kg / m² at the time of diagnosis".

[0121] Time range: The user selects "5 years before diagnosis" as the potential cause time window (2019-2024).

[0122] Field limitation: Check "Metabolic factors → Insulin resistance → Obesity related" and exclude "Genetic factors".

[0123] Evidence preference: Select "Comparative difference analysis first, time series association second".

[0124] Step 2: The specific implementation steps are as follows:

[0125] Comparative evidence: An effect group (n=100, patients with type 2 diabetes) and a control group (n=100, healthy individuals) were defined, and characteristics such as BMI, waist circumference, and fasting blood glucose were extracted. Analysis of differences showed that the proportion of patients with a BMI ≥ 28 kg / m² in the effect group was 75%, while that in the control group was 25% (Chi-square test p<0.001, OR=5.3). The proportion of patients exercising less than three times per week in the effect group was 82%, while that in the control group was 35% (Fisher's exact test p=0.002).

[0126] Time series association analysis: Physical examination data of the patient in the five years before diagnosis were extracted, and it was found that BMI increased year by year from 27kg / m² to 32.5kg / m² from 2022, accompanied by an increase in fasting blood glucose from 5.8mmol / L to 7.2mmol / L (exceeding the diagnostic threshold for diabetes).

[0127] Explicit causal statement pattern inference: The explicit causal statement chain in the knowledge base is: obesity → insulin resistance (confidence 0.9) → type 2 diabetes (confidence 0.88), with a chain length of 2 steps, which meets the domain restrictions.

[0128] Step 3: The specific implementation steps are as follows:

[0129] The event sequence index can quickly retrieve the patient's physical examination events within 5 years;

[0130] Graph algorithms perform common cause discovery, excluding nodes related to genetic factors.

[0131] Step 4: The specific implementation steps are as follows:

[0132] Comparative differences: BMI ≥ 28 kg / m² (p < 0.001, OR = 5.3), insufficient exercise (p = 0.002);

[0133] Explicit causal statement: Obesity → Insulin Resistance → Diabetes (confidence 0.9 × 0.88 = 0.79);

[0134] Time series: The annual increasing trend of BMI and blood sugar (time span 3 years).

[0135] Confidence ranking: comparative difference (OR=5.3) > explicit causal statement (0.79) > time series (continuous change over 3 years).

[0136] like Figure 2 , which is a schematic diagram of the structure of a knowledge base-oriented interactive causal inference support system of the present application;

[0137] An interactive causal inference support system for a knowledge base, characterized by comprising four modules connected in sequence:

[0138] The interactive causal query intent understanding and scope limitation module has an input connected to the user terminal to receive the original query and visualization operations, and an output connected to the input of the multi-strategy causal evidence collection and generation module to transmit structured analysis parameters;

[0139] The multi-strategy causal evidence collection and generation module has its input connected to the output of the interactive causal query intention understanding and scope limitation module, receiving analysis parameters, and its output connected to the input of the causal evidence synthesis and explanatory presentation module, transmitting four types of evidence sets;

[0140] The causal query execution optimization module has its input connected to the query request interface of the multi-strategy causal evidence collection and generation module.

[0141] Its output end is connected to the result receiving interface of the multi-strategy causal evidence collection and generation module to return the optimization result;

[0142] The causal evidence synthesis and explanatory presentation module has its input end connected to the output end of the multi-strategy causal evidence collection and generation module, receives the original evidence set, and its output end is connected to the user terminal and returns to the visual interactive interface.

[0143] As shown in Figure 3 Fig. 4 is a structural schematic diagram of an interactive causal query intention understanding and scope limiting module of an interactive causal inference support system facing a knowledge base according to the present application, the interactive causal query intention understanding and scope limiting module comprising a template engine and a visualization tool, the template engine being used to load a guide question template after detecting a causal keyword, and the visualization tool being used to provide time axis labeling, domain ontology tree checking and intervention event input.

[0144] As shown in Figure 4 Fig. 5 is a structural schematic diagram of an interactive multi-strategy causal evidence collection and generation module of an interactive causal inference support system facing a knowledge base according to the present application, the multi-strategy causal evidence collection and generation module comprising an explicit causal declaration pattern inference unit, a time sequence correlation analysis unit, a statistical significant correlation analysis unit and a comparative evidence discovery unit.

[0145] The four units of the multi-strategy causal evidence collection and generation module work in parallel: the input end of the explicit causal declaration pattern inference unit is connected with a domain set parameter interface; the input end of the time sequence correlation analysis unit is connected with a time window parameter interface; the input end of the statistical significant correlation analysis unit is connected with an evidence type preference parameter interface; and the input end of the comparative evidence discovery unit is connected with an intervention event parameter interface.

[0146] As shown in Figure 5 Fig. 6 is a structural schematic diagram of an interactive causal query execution optimization module of an interactive causal inference support system facing a knowledge base according to the present application, the causal query execution optimization module comprising a time attribute index unit, an event sequence index unit, a graph algorithm execution unit and a result cache unit, the time attribute index unit being used to manage a B+ tree index, the event sequence index unit being used to manage a prefix tree Trie, the graph algorithm execution unit being used to execute a reverse breadth-first search, and the result cache unit being used to store high-frequency subquery results.

[0147] As shown in Figure 6 Fig. 7 is a structural schematic diagram of an interactive causal evidence synthesis and interpretative presentation module of an interactive causal inference support system facing a knowledge base according to the present application, the causal evidence synthesis and interpretative presentation module comprising an evidence classification and sorting unit and an interactive visualization unit, the evidence classification and sorting unit being used to classify by type and sort by confidence, and the interactive visualization unit being used to provide a drill-down function and a time axis display.

[0148] Although several aspects and embodiments have been disclosed in this application, other aspects and embodiments will be apparent to those skilled in the art from the disclosure herein. The aspects and embodiments disclosed in this application are intended to be illustrative only and the true scope of the application will be determined by the following claims appropriately interpreted in accordance with the Patent Statutes, including equivalents.

Claims

1. A knowledge base-oriented interactive causal inference support method, characterized in that: The following steps are involved: Step 1: Obtain the user's causal analysis intention through interactive dialogue and collect preliminary evidence, including effect description, time range, and domain preference parameters; Step 2: Based on the parameters collected in Step 1, execute at least two strategies from the knowledge base in parallel: explicit causal claim pattern inference, time series association analysis, statistically significant correlation analysis, and comparative evidence discovery to generate multiple types of causal evidence; Step 3: Create a dedicated index for the time attributes and event sequences in the knowledge base, and use a dedicated graph algorithm to optimize the efficiency of causal queries; Step 4: Classify and summarize multi-strategy evidence by type, evaluate the confidence level, and visualize it through an interactive interface to assist users in generating causal hypotheses.

2. The method according to claim 1, wherein: In Step 1, the interactive dialogue is implemented through guided questions based on predefined templates to collect at least one of the following: effect clarification: the effect phenomenon that the user is concerned about and the time of occurrence; time range: the time window before the effect occurs; relationship type: direct cause, indirect factor or necessary condition; field limitation: environment, technology or organizational behavior field; evidence preference: explicit causal statement, statistically significant correlation or case comparison.

3. The method according to claim 1 or 2, characterized in that: In Step 1: Use visualization tools to assist users in performing the following steps: mark the effect occurrence point and the focus window on the timeline; select the potential cause domain through the domain ontology tree; and input known intervention events.

4. The method according to claim 1 or 2, characterized in that: Step 1 also includes iterative range adjustment. After collecting preliminary evidence, it is confirmed whether to adjust the range. If the range does not need to be adjusted, it goes to Step 2. If the range needs to be adjusted, it returns to Step 1 for iterative adjustment.

5. The method according to claim 1, wherein: In Step 2: Explicit causal statement pattern inference includes: starting from the effect entity, tracing back the direct or indirect cause along the caused relationship in the knowledge base; pruning the causal path according to the user-set maximum chain length ≤ 3 steps and domain set; and outputting a causal path diagram with confidence and time constraints.

6. The method according to claim 1, wherein: Time series association analysis includes: extracting the state change events of the target domain entities within the time window before the effect occurs; if there is an intervention event, comparing the change values ​​of the effect indicators before and after the intervention.

7. The method according to claim 1, wherein: In Step 3, the dedicated index and graph algorithm includes: Establish a B+ tree index for time attributes; build a prefix tree Trie for event sequences; use reverse breadth-first search with domain constraints to perform common cause discovery; calculate confidence transfer based on independent or cumulative relationships for multi-step causal chains; cache high-frequency subquery results.

8. The method according to claim 1, wherein: In Step 4, the evidence classification summary includes: classifying evidence by explicit causal statement, temporally related events, statistically significant correlation, and comparative difference; wherein, the confidence ranking is arranged as follows: explicit causal statement > statistically significant correlation > temporally related events > comparative difference.

9. An interactive causal inference support system for a knowledge base using the method according to any one of claims 1 to 8, characterized in that: It consists of four modules connected in sequence: The interactive causal query intent understanding and scope limitation module has an input connected to the user terminal to receive the original query and visualization operations, and an output connected to the input of the multi-strategy causal evidence collection and generation module to transmit structured analysis parameters; The multi-strategy causal evidence collection and generation module has its input connected to the output of the interactive causal query intention understanding and scope limitation module, receiving analysis parameters, and its output connected to the input of the causal evidence synthesis and explanatory presentation module, transmitting four types of evidence sets; The causal query execution optimization module has an input end connected to the query request interface of the multi-strategy causal evidence collection and generation module, and an output end connected to the result receiving interface of the multi-strategy causal evidence collection and generation module, and returns the optimization result; The causal evidence synthesis and explanatory presentation module has its input end connected to the output end of the multi-strategy causal evidence collection and generation module to receive the original evidence set, and its output end is connected to the user terminal and the multi-strategy causal evidence collection and generation module respectively for structural feedback.

Images