A method and system for constructing a correction factor network and recommending a correction measure

Abstract

The present application relates to the technical field of psychological correction measure recommendation, and particularly discloses a method and system for constructing a correction factor network and recommending correction measures, wherein the target promotion factor set and the target reduction factor set are obtained, and a pre-constructed correction factor network is queried, the network fusing correction factor nodes, correction measure nodes and the associated edges between the nodes constructed from historical sample data, historical correction data and literature data, the comprehensive influence score of each correction measure node on the target factor set can be calculated, and the correction measure corresponding to the target correction measure node is selected as the recommended result based on the score, the complex relationship between the correction measures and the correction factors is explicitly modeled as a structured network, and the recommended result is interpretable; meanwhile, the network is constructed by fusing multiple data sources, and effective recommendation can be realized in the case of sparse historical interaction data.

Description

Technical Field

[0001] This invention relates to the field of psychological correction measure recommendation technology, and in particular to a method and system for constructing a correction factor network and recommending correction measures. Background Technology

[0002] In the fields of mental health intervention, behavioral modification, and judicial social correction, developing precise corrective measures for abnormal psychological assessment indicators (such as deviations from norms in depression / anxiety scale scores, abnormal cognitive function indicators, and excessive recidivism risk assessment values) is a key step in improving intervention effectiveness and reducing relapse rates. Traditional psychological interventions rely on the intuitive judgment of clinical experts and standardized treatment manuals (such as uniformly using cognitive behavioral therapy (CBT) or drug intervention for depression). However, research shows that mental disorders are highly heterogeneous, with a single treatment regimen typically having a clinical cure rate of less than 30%, and approximately 55% of patients experiencing side effects or treatment resistance.

[0003] In recent years, the Precision Mental Health paradigm has emerged, advocating for data-driven treatment matching based on multidimensional patient characteristics (demographic features, clinical symptoms, biomarkers, psychosocial factors, and historical treatment responses), rather than the traditional trial-and-error approach. Treatment Matching Algorithms have become a core technology, mapping patient characteristics (input) to optimal treatment plans (output) to achieve personalized care. For example, the Personalized Advantage Index (PAI) algorithm generates quantitative treatment selection recommendations by comparing patients' expected responses to different treatments (such as CBT vs. antidepressants).

[0004] Meanwhile, knowledge graph (KG) technology has been introduced into the field of mental health to integrate the complex semantic relationships between mental disorders, symptoms, intervention techniques, mechanisms of action, and prognostic factors. By constructing a graph structure of mental problems, intervention measures, and influencing factors, cross-symptom domain knowledge association and reasoning can be achieved. For example, knowledge graphs can be used to link the symptom of sleep disorders with intervention nodes such as cognitive behavioral therapy (CBT-I) for insomnia and sleep hygiene education, and further link to regulatory factors such as circadian rhythms and anxiety levels.

[0005] However, existing precision mental health technologies primarily focus on a two-dimensional patient-treatment match, using machine learning models (such as random forests and deep neural networks) to learn the statistical correlation between patient feature vectors and treatment responses. While these methods can predict treatment responses, they lack explicit modeling of the complex network of relationships between corrective measures and corrective factors. Specifically, the effectiveness of corrective measures (such as specific psychotherapeutic approaches, intervention modules, and medication types) is influenced by the synergistic or antagonistic effects of multiple corrective factors (such as patient personality traits, social support systems, quality of the treatment alliance, baseline physiological indicators, and environmental factors), and these factors also exhibit complex causal and correlational relationships (e.g., social support affects treatment responses by regulating neuroendocrine function). Existing technologies fail to construct and effectively utilize this multi-level, multi-type network of relationships, resulting in recommendations that lack interpretability and are difficult to adapt to the dynamic evolution of patient states. Summary of the Invention

[0006] To overcome the problems in existing mental health technologies, such as the lack of explicit construction of the complex, multidimensional relationship network between corrective measures and corrective factors, which leads to a lack of interpretability in corrective measure recommendations and an inability to meet the transparency requirements of clinical decision-making, this invention provides a method and system for constructing a corrective factor network and recommending corrective measures.

[0007] In a first aspect, the present invention provides a method for constructing a corrective factor network and recommending corrective measures, the method comprising:

[0008] Obtain a set of target improvement factors and a set of target reduction factors. The set of target improvement factors includes at least one corrective factor for expected numerical improvement, and the set of target reduction factors includes at least one corrective factor for expected numerical reduction. The corrective factors are used to characterize the user's psychological or physiological state. Query the pre-constructed correction factor network, which includes multiple correction measure nodes. Based on the target enhancement factor set and the target reduction factor set, calculate the comprehensive influence score of each correction measure node in the correction factor network on the target enhancement factor set and the target reduction factor set. Based on the comprehensive impact score, at least one target corrective measure node is selected from the plurality of corrective measure nodes, and the corrective measure corresponding to the target corrective measure node is used as the recommendation result; The corrective factor network further includes multiple corrective factor nodes and association edges connecting any two nodes. Each association edge is configured with an association value representing the influence relationship between the two nodes. The construction of the corrective factor network includes: Acquire historical sample data, historical correction data, and literature data; Calculate the first correlation value between any two correction factor nodes based on the historical sample data, calculate the second correlation value between any two correction factor nodes based on the historical correction data, and calculate the third correlation value between any two correction factor nodes based on the literature data. Calculate the first intervention value between any correction measure node and any correction factor node based on the historical correction data, and calculate the second intervention value between any correction measure node and any correction factor node based on the literature data; The association value between any two correction factor nodes is determined based on the first association value, the second association value, and the third association value. The association value between any correction measure node and any correction factor node is determined based on the first intervention value and the second intervention value. The association value between any two correction measure nodes is also determined, generating a correction factor network containing multiple correction factor nodes, multiple correction measure nodes, and inter-node association edges.

[0009] According to one specific implementation, in the above method, the associated edge includes a first associated edge, a second associated edge, and a third associated edge; The first associated edge connects two correction factor nodes, and the associated value of the first associated edge is determined based on at least one of the first associated value, the second associated value, and the third associated value. The second associated edge connects the correction factor node and the correction measure node, and the associated value of the second associated edge is determined based on the first intervention value and the second intervention value; The third association edge connects the two corrective measure nodes, and the association value of the third association edge is determined based on the number of times the two corrective measure nodes co-occur in the historical corrective data or the literature data.

[0010] According to one specific implementation, in the above method, the corrective factor includes at least one of personal basic attribute factors, environmental attribute factors, psychological indicator factors, and physiological indicator factors.

[0011] According to one specific implementation, the corrective measures in the above method include reasoning correction, reading appreciation correction, reading and writing discussion correction, meditation and mind cultivation correction, introspection correction, labor correction, craft activity correction, gardening activity correction, music correction, mindfulness meditation correction, yoga correction, cognitive change correction, employment skills training, and exercise correction.

[0012] According to one specific implementation, in the above method, calculating the first correlation value between any two correction factor nodes based on the historical sample data includes: Obtain the values ​​of each correction factor in the historical sample data, calculate the correlation coefficient between any two correction factor nodes, and use it as the first correlation value; Calculate the second correlation value between any two correction factor nodes based on the historical correction data, including: Obtain the pre-measured value of each correction factor in the historical correction data before the application of correction measures and the post-measured value after the application of correction measures, and calculate the correlation coefficient between the rate of change of any two correction factors as the second correlation value; Calculate the third correlation value between any two correction factor nodes based on the literature data, including: The number of literature articles in which each corrective factor node appears and the number of literature articles in which any two corrective factor nodes appear together are obtained from the literature data. The co-occurrence coefficient is calculated and used as the third association value.

[0013] According to a specific implementation, in the above method, calculating the first intervention value between any corrective measure node and any corrective factor node based on the historical corrective data includes: Obtain the pre-measured value of the target correction factor before the application of the target correction measure and the post-measured value of the target correction factor after the application of the target correction measure from historical correction data. Based on the rate of change of the difference between the pre-measured value and the post-measured value relative to the pre-measured value, determine the first intervention value between the target correction measure node and the target correction factor node.

[0014] According to one specific implementation, in the above method, calculating the second intervention value between any corrective measure node and any corrective factor node based on the literature data includes: A large language model is used to perform semantic analysis on the literature data to extract the intervention direction and effect intensity of the target correction measures on the target correction factors in the literature, and the second intervention value is determined based on the intervention direction and effect intensity.

[0015] According to a specific implementation, in the above method, calculating the comprehensive influence score of each corrective measure node in the corrective factor network on the target enhancement factor set and the target reduction factor set includes: Calculate the sum of the correlation values ​​between each corrective measure node and each corrective factor node in the target improvement factor set, and the sum of the correlation values ​​between each corrective measure node and each corrective factor node in the target reduction factor set. The comprehensive impact score of each corrective measure node is determined based on the sum of the correlation values ​​of each corrective factor node in the target enhancement factor set and the sum of the correlation values ​​of each corrective factor node in the target reduction factor set.

[0016] According to one specific implementation, in the above method, selecting at least one target corrective measure node from the plurality of corrective measure nodes includes: The corrective measure node with the highest comprehensive impact score is determined as the target corrective measure node; Alternatively, all corrective action nodes whose comprehensive impact score exceeds a preset threshold are identified as the target corrective action nodes; Alternatively, the corrective measure nodes can be sorted according to the comprehensive impact score, and the top K corrective measure nodes can be selected from the sorting results as the target corrective measure nodes based on the total number of corrective factors in the target enhancement factor set and the target reduction factor set, where K is a positive integer.

[0017] Secondly, the present invention provides an apparatus for constructing a correction factor network and recommending correction measures, comprising: The acquisition module is used to acquire a target improvement factor set and a target reduction factor set. The target improvement factor set contains at least one corrective factor for expected numerical improvement, and the target reduction factor set contains at least one corrective factor for expected numerical reduction. The corrective factors are used to characterize the user's psychological or physiological state. The calculation module is used to query a pre-constructed correction factor network, which includes multiple correction measure nodes. Based on the target enhancement factor set and the target reduction factor set, the module calculates the comprehensive influence score of each correction measure node in the correction factor network on the target enhancement factor set and the target reduction factor set. The recommendation module is used to select at least one target corrective measure node from the plurality of corrective measure nodes based on the comprehensive impact score, and to use the corrective measure corresponding to the target corrective measure node as the recommendation result; A network construction module is used to construct a correction factor network. The correction factor network includes multiple correction factor nodes and association edges connecting any two nodes. Each association edge is configured with an association value representing the influence relationship between the two nodes. The construction of the correction factor network includes: acquiring historical sample data, historical correction data, and literature data; calculating a first association value between any two correction factor nodes based on the historical sample data; calculating a second association value between any two correction factor nodes based on the historical correction data; calculating a third association value between any two correction factor nodes based on the literature data; and calculating a third association value between any two correction factor nodes based on the historical correction data. The first intervention value between any corrective measure node and any corrective factor node is calculated based on positive data. The second intervention value between any corrective measure node and any corrective factor node is calculated based on the literature data. The correlation value between any two corrective factor nodes is determined based on the first correlation value, the second correlation value, and the third correlation value. The correlation value between any corrective measure node and any corrective factor node is determined based on the first intervention value and the second intervention value. The correlation value between any two corrective measure nodes is also determined, generating a corrective factor network containing multiple corrective factor nodes, multiple corrective measure nodes, and correlation edges between nodes.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention acquires a set of target enhancement factors and a set of target reduction factors, and queries a pre-constructed correction factor network. This network integrates correction factor nodes, correction measure nodes, and inter-node association edges constructed from historical sample data, historical correction data, and literature data. It can calculate the comprehensive influence score of each correction measure node on the target factor set, and select the correction measure corresponding to the target correction measure node as the recommendation result based on the score. This invention can explicitly model the complex relationship between correction measures and correction factors into a structured network, making the recommendation results interpretable. At the same time, by integrating multi-source data to construct the network, it can achieve effective recommendations even when historical interaction data is sparse, solving the problem that black-box prediction models in the prior art lack interpretability and are difficult to deal with cold start scenarios. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating a method for constructing a correction factor network and recommending correction measures, provided in an embodiment of the present invention. Detailed Implementation

[0020] The present invention will now be described in further detail with reference to specific embodiments. However, this should not be construed as limiting the scope of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0021] Unless otherwise specified, the use of terms such as "first," "second," and "third" in the description of specific embodiments of the present invention is merely for distinguishing descriptions of identical or similar components and should not be construed as emphasizing or implying the relative importance of a particular component.

[0022] Furthermore, in the description of the embodiments of the present invention, "several", "more than", and "a number of" represent at least two. The number can be any number, such as two, three, four, five, six, seven, eight, or nine, and can even exceed nine.

[0023] Please refer to Figure 1 The diagram illustrates a flowchart of a method for constructing a correction factor network and recommending correction measures according to an embodiment of the present invention. The method includes: Step 1: Obtain the set of target enhancement factors and the set of target reduction factors; Step 2: Query the pre-constructed correction factor network, which includes multiple correction measure nodes. Based on the target enhancement factor set and the target reduction factor set, calculate the comprehensive influence score of each correction measure node in the correction factor network on the target enhancement factor set and the target reduction factor set. Step 3: Based on the comprehensive impact score, select at least one target corrective measure node from the multiple corrective measure nodes, and use the corrective measure corresponding to the target corrective measure node as the recommendation result.

[0024] Specifically, this embodiment of the invention first constructs a correction factor network. This correction factor network forms the basis for subsequent measure recommendations, and its construction process integrates multi-source data. Specifically, the correction factor network also includes multiple correction factor nodes and association edges connecting any two nodes. Each association edge is configured with an association value representing the influence relationship between the two nodes. Its construction process includes: Data acquisition includes acquiring historical sample data, historical correction data, and literature data.

[0025] Understandably, historical sample data refers to large-scale data unrelated to whether or not correction was performed, with each sample containing specific values ​​for multiple correction factors. For example, in a mental health intervention scenario, historical sample data could include values ​​for correction factors such as age, gender, anxiety score, depression score, sleep quality index, and social support score for a large number of individuals. This data is used to calculate the statistical correlation between correction factors.

[0026] Historical corrective data includes pre-test values ​​for multiple corrective factors before the application of corrective measures, as well as post-test values ​​for multiple corrective factors after the application of corrective measures. For example, historical corrective data may record the anxiety scores and sleep quality indices of individuals receiving mindfulness meditation correction before correction, and the corresponding values ​​after correction. This data is used to calculate the consistency of changes in corrective factors and the intervention effect of corrective measures on the factors.

[0027] The literature data comes from academic papers, clinical guidelines, professional books, etc. Literature searches are conducted using the name of each corrective factor node or corrective measure node, and the literature recalled for all nodes is compiled into a literature set. This data is used to mine co-occurrence relationships between factors, as well as the direction and intensity of intervention on factors.

[0028] Furthermore, the relationships between the correction factor nodes are constructed, including: The first correlation value between any two corrective factor nodes is calculated based on historical sample data. Specifically, in this embodiment of the invention, the values ​​of each corrective factor in the historical sample data are obtained, and the correlation coefficient between any two corrective factor nodes is calculated as the first correlation value. For example, the Pearson correlation coefficient between anxiety score and sleep quality index is calculated. If the correlation coefficient is negative, it indicates that the two factors are negatively correlated; if it is positive, it indicates that they are positively correlated. This correlation value reflects the covariation relationship between factors under natural conditions.

[0029] The system calculates a second correlation value between any two correction factor nodes based on historical correction data. Specifically, it obtains the pre-test values ​​of each correction factor before and after the application of correction measures from historical correction data, and calculates the correlation coefficient between the rates of change of any two correction factors as the second correlation value. For example, for anxiety scores and sleep quality indices, the system calculates the rate of change of anxiety scores and the rate of change of sleep quality indices for each individual, and then calculates the correlation coefficient between these two rate of change sequences. This correlation value reflects the consistency of changes in factors during the correction intervention process; if the improvement trends of the two factors are consistent, the second correlation value is positive; if one improves while the other deteriorates, it is negative.

[0030] The third association value between any two corrected factor nodes is calculated based on literature data. Specifically, the number of literature articles in which each corrected factor node appears and the number of literature articles in which any two corrected factor nodes co-occur are obtained from the literature data, and the co-occurrence coefficient is calculated as the third association value. For example, if the anxiety node appears in 800 articles, the sleep node appears in 600 articles, and both co-occur in 400 articles, then the co-occurrence coefficient is 2 × 400 / (800 + 600) = 0.571. This association value reflects the frequency with which factors are jointly studied in academic literature, reflecting the domain knowledge's understanding of factor correlation.

[0031] Furthermore, the correlation value between any two corrective factor nodes is determined based on the first, second, and third correlation values. For example, methods such as taking the maximum value or weighted averaging can be used for comprehensive analysis. Through the above calculations, the system assigns a correlation value to each pair of corrective factor nodes. This correlation value can be a positive number (indicating positive correlation or consistency) or a negative number (indicating negative correlation or antagonistic relationship), thereby constructing a correlation network between corrective factor nodes.

[0032] Constructing the association between corrective action nodes and corrective factor nodes includes: The first intervention value between any corrective measure node and any corrective factor node is calculated based on historical corrective data. Specifically, the pre-measure value of the target corrective factor before the application of the target corrective measure and the post-measure value of the target corrective factor after the application of the target corrective measure are obtained from historical corrective data. The first intervention value between the target corrective measure node and the target corrective factor node is determined based on the rate of change of the difference between the pre-measure and post-measure values ​​relative to the pre-measure value. This first intervention value reflects the actual intervention effect of the measure on the factor in real corrective practice. If the intervention value is positive, it indicates that the measure has a reinforcing effect on the factor; if it is negative, it indicates a weakening effect; the absolute value indicates the intensity of the intervention effect.

[0033] The system calculates a second intervention value between any corrective measure node and any corrective factor node based on literature data. Specifically, it uses a large language model to perform semantic analysis on the literature data, extracting the intervention direction and effect intensity of the target corrective measure on the target corrective factor. The system can construct a question template, such as: "In this literature, does the [measure] enhance or weaken the [factor]? According to the five-level classification, level 1 is the weakest and level 5 is the strongest. What is the approximate level of the intensity of the [measure]'s enhancement or weakening effect on the [factor]?" After parsing, the large language model outputs the intervention direction and effect intensity. The system calculates the second intervention value based on a preset mapping relationship (e.g., level 1 corresponds to 0.2, level 2 to 0.4, level 3 to 0.6, level 4 to 0.8, and level 5 to 1.0) and the intervention direction (enhancement is +1, weakening is -1). This second intervention value reflects the academic literature's understanding of the relationship between the measure and the factor and can serve as supplementary information when historical data is insufficient.

[0034] In one possible implementation, for the first correlation value, the correlation coefficient between any two corrective factor nodes is calculated by obtaining the values ​​of each corrective factor in the historical sample data. For the second correlation value, the correlation coefficient between the rates of change of any two corrective factors is calculated by obtaining the pre-test values ​​of each corrective factor before the application of corrective measures and the post-test values ​​after the application of corrective measures in the historical corrective data. For the third correlation value, the co-occurrence coefficient is calculated by obtaining the number of documents in which each corrective factor node appears in the literature data and the number of documents in which any two corrective factor nodes co-occur. These calculation methods are specific and operable, enabling those skilled in the art to calculate the correlation values ​​based on actual data.

[0035] Furthermore, by obtaining the pre-measured values ​​of the target correction factor before the application of the target correction measure and the post-measured values ​​of the target correction factor after the application of the target correction measure from historical correction data, the first intervention value is determined based on the rate of change of the difference between the pre-measured and post-measured values ​​relative to the pre-measured value. This calculation method reflects the actual intervention effect of the measure on the factor and has a clear physical meaning.

[0036] Furthermore, by utilizing a large language model to perform semantic analysis on the literature data, the intervention direction and effect intensity of the target correction measures on the target correction factors in the literature are extracted, and a second intervention value is determined based on the intervention direction and effect intensity. This approach uses natural language processing technology to extract knowledge from massive amounts of literature, which can effectively supplement the deficiencies of historical data.

[0037] The correlation value between any corrective measure node and any corrective factor node is determined based on the first and second intervention values. For example, when the product of the two is positive, the maximum value is taken; when the product is negative, the average value is taken. Through the above calculation, the system assigns a correlation value to each pair of corrective measure nodes and corrective factor nodes. The sign of the correlation value indicates whether the measure strengthens or weakens the factor, and the absolute value indicates the intensity of the effect.

[0038] Furthermore, the relationships between the corrective action nodes are constructed, including: The system determines the correlation value between any two corrective measures. Specifically, the system can calculate the co-occurrence of measures based on historical corrective data, i.e., by counting the number of times any two corrective measures are used together in historical data, and calculating the co-occurrence coefficient as the first measure correlation value; or it can calculate the co-occurrence of measures based on literature data, i.e., by counting the number of articles in which any two measures appear together, and calculating the co-occurrence coefficient as the second measure correlation value; then, the maximum value of the two is taken as the final measure correlation value. This correlation value reflects the degree to which measures are frequently used in combination, and is a positive number; the larger the value, the higher the co-occurrence of measures.

[0039] Through the above steps, this embodiment of the invention generates a correction factor network containing multiple correction factor nodes, multiple correction measure nodes, and inter-node association edges. This network is stored in a database in the form of a graph structure, and the association values ​​between nodes quantitatively characterize the mutual influence relationships between the various elements.

[0040] After the correction factor network is constructed, as described in step 1 above, the target improvement factor set and the target reduction factor set are obtained. The target improvement factor set contains at least one correction factor with a desired numerical increase, i.e., a factor for which improvement is desired and the numerical value needs to be increased; the target reduction factor set contains at least one correction factor with a desired numerical decrease, i.e., a factor for which relief is desired and the numerical value needs to be reduced. These factors are used to characterize the user's psychological or physiological state and serve as input conditions for personalized recommendations.

[0041] Further, as described in step 2 above, the pre-constructed correction factor network is queried, and the comprehensive impact score of each correction measure node in the correction factor network on the target improvement factor set and the target reduction factor set is calculated based on the target improvement factor set and the target reduction factor set.

[0042] Specifically, the system iterates through all corrective action nodes in the corrective factor network. For each corrective action node, it calculates the sum of its correlation values ​​with each corrective factor node in the target enhancement factor set, and the sum of its correlation values ​​with each corrective factor node in the target reduction factor set. Since positive values ​​in the correlation values ​​represent reinforcement and negative values ​​represent weakening effects, the system directly adds the correlation values ​​corresponding to the target enhancement factors, and negates the correlation values ​​corresponding to the target reduction factors before adding them, or uses a difference method to calculate the comprehensive impact score.

[0043] Further, as described in step 3 above, based on the comprehensive impact score, at least one target corrective measure node is selected from multiple corrective measure nodes, and the corrective measure corresponding to the target corrective measure node is used as the recommendation result.

[0044] For example, you can choose the measure node with the highest comprehensive impact score as the target, or you can choose all measure nodes with scores exceeding a preset threshold as the target, or you can choose the top K measure nodes as the target based on the number of factors.

[0045] In one possible implementation, the association edges include a first association edge, a second association edge, and a third association edge. The first association edge connects two corrective factor nodes, and its association value is determined based on at least one of the aforementioned first, second, and third association values. This clarifies the source of the association between factor nodes, making network construction more flexible and allowing for the selection of different calculation methods based on data availability. The second association edge connects a corrective factor node and a corrective measure node, and its association value is determined based on a first intervention value and a second intervention value. This clarifies the source of the association between measures and factors, integrating real-world data and literature knowledge, and improving the reliability of the association values. The third association edge connects two corrective measure nodes, and its association value is determined based on the co-occurrence frequency of the two corrective measure nodes in historical corrective data or literature data. This establishes associations between measures, providing a basis for subsequent joint recommendations or multi-measure combinations. This makes the hierarchical structure of the corrective factor network clearer, with different types of association edges performing different functions, facilitating subsequent inference calculations.

[0046] Furthermore, the corrective factors may include at least one of the following: basic personal attribute factors (such as age, gender, education level, marital status, etc.), environmental attribute factors (such as growth history, family environment, economic situation, etc.), psychological indicator factors (such as anxiety score, depression score, antisocial personality tendency, etc.), and physiological indicator factors (such as sleep quality index, heart rate variability, cortisol level, etc.).

[0047] Furthermore, corrective measures can include reasoning correction, reading appreciation correction, reading and writing discussion correction, meditation and self-cultivation correction, introspection correction, labor correction, craft activity correction, gardening activity correction, music correction, mindfulness meditation correction, yoga correction, cognitive change correction, employment skills training, and exercise correction. These measures cover multiple dimensions such as psychological intervention, behavioral correction, and physiological regulation, making the technical solution of this invention applicable to various corrective scenarios.

[0048] In one possible implementation, calculating the comprehensive impact score includes: calculating the sum of the correlation values ​​between each corrective measure node and each corrective factor node in the target enhancement factor set, and the sum of the correlation values ​​between each corrective measure node and each corrective factor node in the target reduction factor set; and determining the comprehensive impact score of each corrective measure node based on the sum of these two items.

[0049] This calculation method allows for a quantitative assessment of the overall impact of each corrective measure on a user's personalized needs (enhancing certain factors or reducing others). Since positive values ​​in the correlation values ​​represent strengthening effects and negative values ​​represent weakening effects, this calculation method accurately reflects the overall regulatory effect of the measures on the target factors. This preferred implementation makes the recommendation criteria more scientific and quantitative.

[0050] As a preferred embodiment of the present invention, the specific strategy for selecting the target correction measure node can include three types: The corrective measure node with the highest overall impact score is identified as the target corrective measure node; Alternatively, all corrective action nodes whose overall impact score exceeds a preset threshold can be identified as target corrective action nodes; Alternatively, the corrective measure nodes can be sorted according to their comprehensive impact scores. Based on the total number of corrective factors in the target enhancement factor set and the target reduction factor set, the top K corrective measure nodes can be selected as target corrective measure nodes from the sorting results, where K is a positive integer.

[0051] Specifically, the first strategy is suitable for scenarios requiring a single core intervention, providing users with the most effective core measures. The second strategy is suitable for scenarios requiring combined interventions, offering users multiple options. The third strategy can control the number of recommended measures while ensuring effectiveness, avoiding overly complex solutions. By providing multiple selection strategies, claim 9 enables the technical solution of the present invention to adapt to different application scenario requirements.

[0052] On the other hand, the present invention also provides an apparatus for constructing a correction factor network and recommending correction measures. The apparatus includes a network construction module, an acquisition module, a calculation module, and a recommendation module.

[0053] The network construction module is used to construct a corrective factor network. This module acquires historical sample data, historical corrective data, and literature data. Based on the historical sample data, it calculates the first correlation value between any two corrective factor nodes; based on the historical corrective data, it calculates the second correlation value; based on the literature data, it calculates the third correlation value; based on the historical corrective data, it calculates the first intervention value between any corrective measure node and any corrective factor node; based on the literature data, it calculates the second intervention value; based on the first, second, and third correlation values, it determines the correlation value between any two corrective factor nodes; based on the first and second intervention values, it determines the correlation value between any corrective measure node and any corrective factor node; and finally, it determines the correlation value between any two corrective measure nodes, generating a corrective factor network containing multiple corrective factor nodes, multiple corrective measure nodes, and inter-node correlation edges.

[0054] The acquisition module is used to acquire the target improvement factor set and the target reduction factor set. The target improvement factor set contains at least one corrective factor for expected numerical improvement, and the target reduction factor set contains at least one corrective factor for expected numerical reduction. The corrective factors are used to characterize the user's psychological or physiological state.

[0055] The calculation module is used to query the pre-built correction factor network and, based on the target enhancement factor set and the target reduction factor set, calculate the comprehensive impact score of each correction measure node in the correction factor network on the target enhancement factor set and the target reduction factor set.

[0056] The recommendation module is used to select at least one target corrective measure node from multiple corrective measure nodes based on the comprehensive impact score, and to use the corrective measure corresponding to the target corrective measure node as the recommendation result.

[0057] The functions of each module of this device correspond to the steps of the above method embodiments, and can be found in the description of the method embodiments for details, which will not be repeated here. Through modular division, the technical solution of this invention can be implemented in software or hardware, facilitating deployment and application.

[0058] Based on the above technical solution, this invention obtains a set of target enhancement factors and a set of target reduction factors, and queries a pre-constructed correction factor network. This network integrates correction factor nodes, correction measure nodes, and inter-node association edges constructed from historical sample data, historical correction data, and literature data. It can calculate the comprehensive influence score of each correction measure node on the target factor set, and select the correction measure corresponding to the target correction measure node as the recommendation result based on the score. This can explicitly model the complex relationship between correction measures and correction factors into a structured network, making the recommendation result interpretable. At the same time, by integrating multi-source data to construct the network, effective recommendations can be achieved even when historical interaction data is sparse, solving the problem that black-box prediction models in the prior art lack interpretability and are difficult to cope with cold start scenarios.

[0059] The technical solution provided by the present invention will be described and explained in detail below with reference to specific embodiments.

[0060] First, construct the correction factor network: Based on multi-source data, it is necessary to construct an undirected factor network graph of correction factors and correction measures. The network includes: correction factor nodes N, correction measure nodes C, and the association relationship between the nodes R. Top-level design of the network: The design of the correction factor node N is such that each type of attribute is a correction factor node, and its scope includes: basic personal attributes (age, gender, education level, marital status, etc.), environmental attributes (growth history, original family environment, remarried family environment, economic situation, etc.), psychological indicators (antisocial personality, depression, anxiety, diagnosis of mental illness, concerns, etc.), and physiological indicators (illness, disability, physical condition, etc.).

[0061] The design of the corrective measure node C includes various corrective measures, each of which constitutes a corrective measure node. The scope of corrective measures includes: reasoning correction, reading appreciation correction, reading and writing discussion correction, meditation and mind cultivation correction, introspection correction, labor correction, craft activity correction, gardening activity correction, music correction, mindfulness meditation correction, yoga correction, cognitive change correction, employment skills training, exercise correction, and other corrective methods.

[0062] The design of the node association relationship R allows for associations between any two nodes; each association relationship contains an association value. , >0 indicates a positive impact. <0 indicates a negative impact; the correlation is not directional; the correlation includes: positive or weak correlation between factor nodes; co-occurrence between measure nodes; and the strengthening or weakening effect of measures on factors between measures and factor nodes.

[0063] Constructing the relationship (NRN) between corrective factors: The correlation between factors is calculated based on the collected factor data: Based on historically collected sample data (regardless of whether correction is applied), each sample contains specific values ​​for multiple factor nodes. The correlation between any two factor nodes is then calculated. The correlation values ​​between The formula is as follows:

[0064] in, .

[0065] Factor consistency is calculated based on historical correction data: Data is extracted from historical correction data, and each factor is considered before each application of measures. Pre-test values After applying the measures, each factor Post-test values Calculate two factor nodes The correlation values ​​between The formula is as follows:

[0066] Factor co-occurrence is calculated based on authoritative literature: Literature is retrieved by the name of each factor node, and a literature set is formed by aggregating all literature recalled by each factor node; the co-occurrence of factors in each literature is calculated. The co-occurrence frequency is used to calculate the two factor nodes. The correlation values ​​between The formula is as follows:

[0067] in, for The number of articles appearing in each category , They represent The number of documents appearing and The number of references mentioned.

[0068] Calculate the correlation values ​​between factor nodes The formula is as follows:

[0069] Constructing the correlation between corrective measures and corrective factors (CRN): The interventional effect of corrective measures on factors is calculated based on literature: literature is retrieved by the name of each factor node and by the name of each measure, and all recalled literature is collected to form a literature set; a large language model is used to interpret the literature, traversing all measures and factor nodes in the literature, and extracting the direction and intensity of intervention of the measures on the factors, promoter, and degree of intervention. The formula is as follows:

[0070] in, These are preset hyperparameters, with a default value of 1; The effect of the measure is indicated by a value of 1 for enhancement and -1 for weakening. The effect strength is 0.2 for level 1, 0.4 for level 2, 0.6 for level 3, 0.8 for level 4, and 1.0 for level 5.

[0071] The question template is as follows: In this document, "measure" right "factor" The effect is to enhance "factor" or weaken "factor" According to the five-level classification, level 1 is the weakest, and level 5 is the strongest. "measure" right "factor" What level of intensity is the increase or decrease effect? The intervention effectiveness of measures on factors is calculated based on historical correction data: Data is extracted based on historical correction data, and before each application of measures, each factor... Pre-test values After applying the measures, each factor Post-test values Calculation measures for factors Intervention value Assuming there are M data points, the formula is as follows:

[0072] This represents the minimum preset value for the Ni index.

[0073] Calculate the intervention values ​​of factor nodes and measure nodes. The formula is as follows:

[0074] Establishing Correctional Relationships (CRCs): Calculating the co-occurrence of measures based on historical correctional data: Data is extracted from historical correctional data to obtain the set of measures applied each time, and the co-occurrence of measures is calculated. The formula is as follows:

[0075] for The number of times they are used together , They represent Number of applications and Number of times it is applied.

[0076] Based on the co-occurrence of measures in the literature: Literature is retrieved by the name of each measure node, and a literature set is formed by aggregating all documents recalled by each measure node; the co-occurrence of measures in each document is calculated. The number of co-occurrences is used to calculate the number of two measure nodes. The correlation values ​​between The formula is as follows:

[0077] for Number of documents appearing together , They represent Number of articles and Number of articles appearing.

[0078] Calculate the correlation values ​​between the nodes of the measure. The formula is as follows:

[0079] Store the correction factor network.

[0080] Recommend corrective measures based on the corrective factor network: given a set of factors that need to be improved. and the set of target reduction factors Then, assuming that the factors that need to be regulated are... Each strategy is recommended based on its specific characteristics.

[0081] Recommended maximum coverage measures :

[0082] The union of measures. Represents the set of factor nodes Related measures nodes, Represents the set of factor nodes Associated factor nodes.

[0083] Recommendation of a single core measure :

[0084] Optimal set : First, calculate the single measure. right and Impact score The formula is:

[0085]

[0086]

[0087]

[0088] This is the maximum traversal depth; it is a hyperparameter and defaults to 4. All of Sort from largest to smallest to form The K largest values ​​are determined according to the following formula. And corresponding measures:

[0089]

[0090] This represents the deviation factor from the standard deviation, with a default value of 2, meaning the (k+1)th value is 2 standard deviations lower. The total number of regulatory factors, i.e. and The sum of quantities.

[0091] Taking the mindfulness meditation node as an example, its correlation with the anxiety factor is -0.24 (reducing anxiety), with the sleep factor is -0.18 (reducing sleep problems), and with the hypothetical social support factor is 0.15 (improving social support). If the target set of factors to be reduced is {anxiety, sleep problems}, and the target set of factors to be improved is empty, then... =0-[(-0.24)+(-0.18)]=0-(-0.42)=0.42. The higher this score, the more effective the measure is in reducing the target factor.

[0092] The following section uses a key individual from a community corrections agency as an example to provide a detailed description of the technical solution of this invention.

[0093] In a real-world correctional setting, this key individual was sentenced to community correction for dangerous driving. Upon entry into community correction, the psychological assessment showed the following: a recidivism risk assessment scale score of 75 (normative safety threshold of 60), a self-rating anxiety scale score of 68 (normative borderline of 50), and a social support rating scale score of 32 (lower norm level of 40). Based on these risk characteristics, personalized correctional measures that reduce recidivism risk and anxiety levels while simultaneously improving social support can be recommended.

[0094] The corrective measures recommendation system first constructs a corrective factor network based on judicial correction data from the past three years within the region.

[0095] (1) Obtaining historical sample data: Collect the file data of 5,000 community correction subjects in the jurisdiction. Each sample includes the values ​​of correction factors such as age, education level, recidivism risk score, anxiety score, depression score, social support score, and family economic status. The system calculates the correlation coefficient between any two correction factor nodes as the first correlation value. For example, the correlation coefficient between recidivism risk and anxiety score is 0.62, indicating that the two are moderately positively correlated; the correlation coefficient between recidivism risk and social support score is -0.55, indicating that the two are negatively correlated.

[0096] (2) Obtain historical correction data: Collect pre-test and post-test data from 1000 subjects who have completed correction. The system calculates the correlation coefficient between the rate of change of any two correction factors as the second correlation value. For example, after receiving psychological counseling, the correlation coefficient between the rate of change of recidivism risk score and the rate of change of anxiety score is 0.58, indicating that the improvement of the two is consistent.

[0097] (3) Obtaining literature data: Academic papers were retrieved using keywords such as "recidivism risk," "anxiety," "social support," "cognitive behavioral therapy," and "vocational skills training" to construct a literature collection. The co-occurrence coefficient was calculated as the third correlation value. For example, the co-occurrence coefficient between recidivism risk and cognitive behavioral therapy was 0.72, indicating that the two were frequently associated in the literature.

[0098] (4) Calculate the intervention value: Based on historical correctional data, calculate the first intervention value of each correctional measure on the correctional factor. For example, the first intervention value of cognitive behavioral therapy on recidivism risk is -0.35 (a negative value indicates a reducing effect), and the first intervention value of vocational skills training on social support is 0.28. Based on literature data, use a large language model to extract the intervention direction and intensity, and calculate the second intervention value. For example, in the literature, the intervention direction of cognitive behavioral therapy on recidivism risk is weakening, and the intensity level is 4, corresponding to a second intervention value of -0.8. Combining the first and second intervention values, the association value between cognitive behavioral therapy and recidivism risk is determined to be -0.5.

[0099] (5) Determine the correlation value between measures: Based on the number of times measures co-occur in the historical correction data, the co-occurrence coefficient of cognitive behavioral therapy and psychological counseling is calculated to be 0.65, which is used as the correlation value between measures.

[0100] Through the above steps, the system generates a correction factor network containing 30 correction factor nodes, 20 correction measure nodes, and associated edges.

[0101] Furthermore, based on the assessment data of this key individual, the system obtains a set of target enhancement factors and a set of target reduction factors. The target reduction factor set consists of {recidivism risk factors, anxiety factors}, while the target enhancement factor set consists of {social support factors}.

[0102] The system traverses all corrective measure nodes in the corrective factor network, calculates the sum of the correlation values ​​between each corrective measure node and each corrective factor node in the target improvement factor set, and the sum of the correlation values ​​between each corrective measure node and each corrective factor node in the target reduction factor set, and then determines the comprehensive impact score of each corrective measure node.

[0103] Taking the cognitive behavioral therapy node as an example: its correlation value with the recidivism risk factor is -0.5 (weakened), its correlation value with the anxiety factor is -0.42 (weakened), and its correlation value with the social support factor is 0.15 (reinforced). Therefore, the comprehensive influence score = (0.15) - [(-0.5) + (-0.42)] = 0.15 - (-0.92) = 1.07.

[0104] Taking vocational skills training as an example: its correlation with recidivism risk factor is -0.18, its correlation with anxiety factor is -0.05, and its correlation with social support factor is 0.35. The overall impact score = 0.35 - [(-0.18) + (-0.05)] = 0.35 - (-0.23) = 0.58.

[0105] Taking the mindfulness meditation node as an example: its correlation with the recidivism risk factor is -0.22, its correlation with the anxiety factor is -0.38, and its correlation with the social support factor is 0.08. The overall impact score = 0.08 - [(-0.22) + (-0.38)] = 0.08 - (-0.60) = 0.68.

[0106] In one possible implementation, the system employs the first selection strategy, identifying the corrective measure node with the highest overall impact score as the target corrective measure node. The calculation results show that cognitive behavioral therapy has the highest overall impact score (1.07), therefore the system recommends cognitive behavioral therapy as the outcome.

[0107] Alternatively, the system can employ a second strategy, recommending all nodes whose overall impact score exceeds a preset threshold (e.g., 0.6). In this case, cognitive behavioral therapy (1.07) and mindfulness meditation (0.68) would both be recommended. Another strategy is to select the top K=2 nodes (cognitive behavioral therapy and mindfulness meditation) based on the total number of target factors M=3.

[0108] In this embodiment, the modules of the corrective measure recommendation device work collaboratively. The acquisition module receives the target factor set of the key personnel; the network construction module has completed the pre-construction of the corrective factor network and stored it in the database; the calculation module queries the network and performs the above-mentioned comprehensive influence score calculation; the recommendation module outputs the recommended solution: "It is recommended that the key personnel prioritize participation in cognitive behavioral therapy correction. This measure has significant effects on reducing the risk of recidivism (correlation value -0.5) and alleviating anxiety (correlation value -0.42), and can also help improve social support (correlation value 0.15). If conditions permit, mindfulness meditation training can be combined to enhance emotional regulation ability." Furthermore, this recommended approach was applied to the community correction of this key individual. After three months of intervention, the individual's recidivism risk assessment score decreased from 75 to 52 (below the safe threshold), the self-rating anxiety scale score decreased from 68 to 45 (below the critical value), and the social support rating scale score increased from 32 to 43. The intervention was effective, validating the effectiveness and practicality of the technical solution of this invention in risk assessment and correctional measure recommendations for key individuals in the judicial field.

[0109] It is understood that those skilled in the art will recognize that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software 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 the invention. In the several embodiments provided by the present invention, it should be understood that the disclosed apparatus, devices, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for example, the division of modules is merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatus or modules may be electrical, mechanical, or other forms.

[0110] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0111] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more units can be integrated into one module.

[0112] In the above embodiments, the functions of each functional module can be implemented entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital video discs, DVDs), or semiconductor media (e.g., solid-state disks, SSDs), etc.

[0113] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0114] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for constructing a corrective factor network and recommending corrective measures, characterized in that, The method includes: Obtain a set of target improvement factors and a set of target reduction factors. The set of target improvement factors includes at least one corrective factor for expected numerical improvement, and the set of target reduction factors includes at least one corrective factor for expected numerical reduction. The corrective factors are used to characterize the user's psychological or physiological state. Query the pre-constructed correction factor network, which includes multiple correction measure nodes. Based on the target enhancement factor set and the target reduction factor set, calculate the comprehensive influence score of each correction measure node in the correction factor network on the target enhancement factor set and the target reduction factor set. Based on the comprehensive impact score, at least one target corrective measure node is selected from the plurality of corrective measure nodes, and the corrective measure corresponding to the target corrective measure node is used as the recommendation result; The corrective factor network further includes multiple corrective factor nodes and association edges connecting any two nodes. Each association edge is configured with an association value representing the influence relationship between the two nodes. The construction of the corrective factor network includes: Acquire historical sample data, historical correction data, and literature data; Calculate the first correlation value between any two correction factor nodes based on the historical sample data, calculate the second correlation value between any two correction factor nodes based on the historical correction data, and calculate the third correlation value between any two correction factor nodes based on the literature data. Calculate the first intervention value between any correction measure node and any correction factor node based on the historical correction data, and calculate the second intervention value between any correction measure node and any correction factor node based on the literature data; The association value between any two correction factor nodes is determined based on the first association value, the second association value, and the third association value. The association value between any correction measure node and any correction factor node is determined based on the first intervention value and the second intervention value. The association value between any two correction measure nodes is also determined, generating a correction factor network containing multiple correction factor nodes, multiple correction measure nodes, and inter-node association edges.

2. The method for constructing the corrective factor network and recommending corrective measures according to claim 1, characterized in that, The associated edges include a first associated edge, a second associated edge, and a third associated edge; The first associated edge connects two correction factor nodes, and the associated value of the first associated edge is determined based on at least one of the first associated value, the second associated value, and the third associated value. The second associated edge connects the correction factor node and the correction measure node, and the associated value of the second associated edge is determined based on the first intervention value and the second intervention value; The third association edge connects the two corrective measure nodes, and the association value of the third association edge is determined based on the number of times the two corrective measure nodes co-occur in the historical corrective data or the literature data.

3. The method for constructing the corrective factor network and recommending corrective measures according to claim 1, characterized in that, The corrective factors include at least one of the following: personal basic attribute factors, environmental attribute factors, psychological indicator factors, and physiological indicator factors.

4. The method for constructing the corrective factor network and recommending corrective measures according to claim 1, characterized in that, The corrective measures include reasoning correction, reading appreciation correction, reading and writing discussion correction, meditation and mind cultivation correction, introspection correction, labor correction, craft activity correction, gardening activity correction, music correction, mindfulness meditation correction, yoga correction, cognitive change correction, employment skills training, and exercise correction.

5. The method for constructing the corrective factor network and recommending corrective measures according to claim 1, characterized in that, Calculate the first correlation value between any two correction factor nodes based on the historical sample data, including: Obtain the values ​​of each correction factor in the historical sample data, calculate the correlation coefficient between any two correction factor nodes, and use it as the first correlation value; Calculate the second correlation value between any two correction factor nodes based on the historical correction data, including: Obtain the pre-measured value of each correction factor in the historical correction data before the application of correction measures and the post-measured value after the application of correction measures, and calculate the correlation coefficient between the rate of change of any two correction factors as the second correlation value; Calculate the third correlation value between any two correction factor nodes based on the literature data, including: The number of literature articles in which each corrective factor node appears and the number of literature articles in which any two corrective factor nodes appear together are obtained from the literature data. The co-occurrence coefficient is calculated and used as the third association value.

6. The method for constructing the corrective factor network and recommending corrective measures according to claim 1, characterized in that, The step of calculating the first intervention value between any corrective measure node and any corrective factor node based on the historical corrective data includes: Obtain the pre-measured value of the target correction factor before the application of the target correction measure and the post-measured value of the target correction factor after the application of the target correction measure from historical correction data. Based on the rate of change of the difference between the pre-measured value and the post-measured value relative to the pre-measured value, determine the first intervention value between the target correction measure node and the target correction factor node.

7. The method for constructing the corrective factor network and recommending corrective measures according to claim 1, characterized in that, The calculation of the second intervention value between any corrective measure node and any corrective factor node based on the literature data includes: A large language model is used to perform semantic analysis on the literature data to extract the intervention direction and effect intensity of the target correction measures on the target correction factors in the literature, and the second intervention value is determined based on the intervention direction and effect intensity.

8. The method for constructing the corrective factor network and recommending corrective measures according to claim 1, characterized in that, Calculate the comprehensive impact score of each corrective measure node in the corrective factor network on the target enhancement factor set and the target reduction factor set, including: Calculate the sum of the correlation values ​​between each corrective measure node and each corrective factor node in the target improvement factor set, and the sum of the correlation values ​​between each corrective measure node and each corrective factor node in the target reduction factor set. The comprehensive impact score of each corrective measure node is determined based on the sum of the correlation values ​​of each corrective factor node in the target enhancement factor set and the sum of the correlation values ​​of each corrective factor node in the target reduction factor set.

9. The method for constructing the corrective factor network and recommending corrective measures according to claim 1, characterized in that, Selecting at least one target corrective action node from the plurality of corrective action nodes includes: The corrective measure node with the highest comprehensive impact score is determined as the target corrective measure node; Alternatively, all corrective action nodes whose comprehensive impact score exceeds a preset threshold are identified as the target corrective action nodes; Alternatively, the corrective measure nodes can be sorted according to the comprehensive impact score, and the top K corrective measure nodes can be selected from the sorting results as the target corrective measure nodes based on the total number of corrective factors in the target enhancement factor set and the target reduction factor set, where K is a positive integer.

10. An apparatus for constructing a correction factor network and recommending correction measures, characterized in that, include: The acquisition module is used to acquire a target improvement factor set and a target reduction factor set. The target improvement factor set contains at least one corrective factor for expected numerical improvement, and the target reduction factor set contains at least one corrective factor for expected numerical reduction. The corrective factors are used to characterize the user's psychological or physiological state. The calculation module is used to query a pre-constructed correction factor network, which includes multiple correction measure nodes. Based on the target enhancement factor set and the target reduction factor set, the module calculates the comprehensive influence score of each correction measure node in the correction factor network on the target enhancement factor set and the target reduction factor set. The recommendation module is used to select at least one target corrective measure node from the plurality of corrective measure nodes based on the comprehensive impact score, and to use the corrective measure corresponding to the target corrective measure node as the recommendation result; A network construction module is used to construct a correction factor network. The correction factor network includes multiple correction factor nodes and association edges connecting any two nodes. Each association edge is configured with an association value representing the influence relationship between the two nodes. The construction of the correction factor network includes: acquiring historical sample data, historical correction data, and literature data; calculating a first association value between any two correction factor nodes based on the historical sample data; calculating a second association value between any two correction factor nodes based on the historical correction data; calculating a third association value between any two correction factor nodes based on the literature data; and calculating a third association value between any two correction factor nodes based on the historical correction data. The first intervention value between any corrective measure node and any corrective factor node is calculated based on positive data. The second intervention value between any corrective measure node and any corrective factor node is calculated based on the literature data. The correlation value between any two corrective factor nodes is determined based on the first correlation value, the second correlation value, and the third correlation value. The correlation value between any corrective measure node and any corrective factor node is determined based on the first intervention value and the second intervention value. The correlation value between any two corrective measure nodes is also determined, generating a corrective factor network containing multiple corrective factor nodes, multiple corrective measure nodes, and correlation edges between nodes.

Images