PTSD risk prediction system based on multi-dimensional data analysis

Abstract

The invention discloses a PTSD risk prediction system based on multi-dimensional data analysis, and relates to the technical field of data processing, and the system comprises the steps: obtaining a trauma event multi-dimensional data stream of a target user; performing standardization processing and risk feature screening on the wound event multi-dimensional data stream to obtain a wound risk key feature set; building a PTSD risk prediction model, performing risk prediction on the trauma risk key feature set, and outputting target PTSD risk prediction parameters; constructing an intervention strategy parameter space, performing strategy analysis on the target PTSD risk prediction parameters based on the intervention strategy parameter space, and determining target intervention strategy parameters; and performing intervention interaction control on the target user based on the target intervention strategy parameter. The technical problem that intervention measures are not accurate enough due to insufficient accuracy of PTSD risk prediction in the prior art is solved, and the technical effect of accurate PTSD risk prediction and intervention is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a PTSD risk prediction system based on multidimensional data analysis. Background Art

[0002] Existing technologies often rely on traditional assessment methods when predicting PTSD risk, usually based on single or limited factors, such as psychological assessment questionnaires and clinical experience. These methods not only lack sufficient personalization and accuracy, but also fail to comprehensively consider multi-dimensional data, such as the type of traumatic event, severity, duration, exposure, gene expression, psychological characteristics and other multiple factors. Therefore, existing technologies have a high error rate and limitations when dealing with complex PTSD risk predictions, and the predictions are not comprehensive and scientific, resulting in inaccurate intervention measures. Summary of the invention

[0003] The present application provides a PTSD risk prediction system based on multidimensional data analysis, which is used to solve the technical problem that the existing technology is not accurate enough in PTSD risk prediction, resulting in inaccurate intervention measures.

[0004] In view of the above problems, the present application provides a PTSD risk prediction system based on multidimensional data analysis.

[0005] The present application provides a PTSD risk prediction system based on multidimensional data analysis, the system comprising:

[0006] A data input module is used to collect and acquire a multidimensional data stream of traumatic events of a target user, wherein the multidimensional data stream of traumatic events includes event type, severity, duration, degree of exposure, gene expression data, psychological characteristics, and relevant medical history; a data processing module is used to perform standardization processing and risk feature screening on the multidimensional data stream of traumatic events to obtain a key feature set of trauma risk; a risk prediction module is used to build a PTSD risk prediction model, perform risk prediction on the key feature set of trauma risk based on the PTSD risk prediction model, and output target PTSD risk prediction parameters; an intervention strategy recommendation module is used to construct an intervention strategy parameter space, perform strategy analysis on the target PTSD risk prediction parameters based on the intervention strategy parameter space, and determine target intervention strategy parameters; an intervention interaction module is used to perform intervention interaction management and control on the target user based on the target intervention strategy parameters.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The present application collects and obtains a multidimensional data stream of traumatic events of a target user, wherein the multidimensional data stream of traumatic events includes event type, severity, duration, degree of exposure, gene expression data, psychological characteristics, and relevant medical history; the multidimensional data stream of traumatic events is subjected to standardization processing and risk feature screening to obtain a key feature set of traumatic risk; a PTSD risk prediction model is constructed, and risk prediction is performed on the key feature set of traumatic risk based on the PTSD risk prediction model, and target PTSD risk prediction parameters are output; an intervention strategy parameter space is constructed, and strategy analysis is performed on the target PTSD risk prediction parameters based on the intervention strategy parameter space to determine target intervention strategy parameters; and intervention interaction management and control are performed on the target user based on the target intervention strategy parameters. The present invention solves the technical problem that the prior art is not accurate enough in predicting PTSD risk, resulting in inaccurate intervention measures. The present invention collects a multidimensional data stream of traumatic events of a target user, and performs standardization processing and feature screening to extract key features of traumatic risk. A PTSD risk prediction model is constructed and predicted using a risk prediction module, and target risk parameters are output. Strategy analysis is performed based on the prediction results, and intervention strategy parameters are optimized. Finally, personalized intervention management and control are performed on the user, so as to achieve the technical effect of accurate PTSD risk prediction and intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A schematic diagram of the structure of a PTSD risk prediction system based on multidimensional data analysis provided in an embodiment of the present application;

[0011] Figure 2 A schematic diagram of the process of obtaining a key feature set of trauma risk in a PTSD risk prediction system based on multidimensional data analysis provided in an embodiment of the present application.

[0012] Explanation of the reference numerals: data input module 11 , data processing module 12 , risk prediction module 13 , intervention strategy recommendation module 14 , intervention interaction module 15 . DETAILED DESCRIPTION

[0013] This application provides a PTSD risk prediction system based on multidimensional data analysis to solve the technical problem that the existing technology is not accurate enough in PTSD risk prediction, resulting in inaccurate intervention measures. By collecting multidimensional data streams of traumatic events of target users, and performing standardized processing and feature screening, key features of trauma risk are extracted, and a PTSD risk prediction model is constructed and predicted using a risk prediction module. The target risk parameters are output, and strategy analysis is performed based on the prediction results. The intervention strategy parameters are optimized, and ultimately personalized intervention management and control are performed on users, thereby achieving the technical effect of accurate PTSD risk prediction and intervention.

[0014] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0015] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0016] Examples, such as Figure 1 As shown, the embodiment of the present application provides a PTSD risk prediction system based on multidimensional data analysis, the system comprising:

[0017] The data input module 11 is used to collect and acquire the multi-dimensional data stream of traumatic events of the target user, wherein the multi-dimensional data stream of traumatic events includes event type, severity, duration, exposure level, gene expression data, psychological characteristics and relevant medical history.

[0018] In an embodiment of the present application, first, the data input module interacts with the target user through questionnaires or electronic health records to collect personal experiences and relevant background information. In the process of interaction with the user, specific data of the traumatic event is obtained, including event type, severity, duration, and exposure. Event types include traffic accidents, natural disasters, etc. Severity refers to the serious impact that the user believes the traumatic event has on his or her life, which is divided into three levels: high, medium and low, and is determined by the target user. Duration is the length of time the user experiences the traumatic event. Exposure refers to the user's role in the event, such as a witness, direct victim, or indirect victim.

[0019] Furthermore, high-throughput gene sequencing technology (such as RNA sequencing) is used to obtain the target user's PTSD-related gene expression information. This step is done with the user's consent and provision of samples (such as blood samples), and biomarkers that may be affected after trauma are extracted through genetic analysis technology. When collecting psychological characteristics, standardized psychological assessment tools are used to interact with users to obtain relevant information about their psychological state. During the user interaction process, their previous medical history data is also collected, including any known mental illnesses.

[0020] Through the above process, the data input module collects and obtains the multidimensional data stream of traumatic events of the target user, where the multidimensional data stream of traumatic events includes event type, severity, duration, exposure level, gene expression data, psychological characteristics and relevant medical history.

[0021] The data processing module 12 is used to perform standardization processing and risk feature screening on the multi-dimensional data stream of the trauma event to obtain a key feature set of trauma risk.

[0022] In the embodiment of the present application, when the data processing module performs standardization processing and risk feature screening on the multidimensional data stream of trauma events, it first ensures the validity of the data through outlier identification and data cleaning. Then, it obtains and constructs a data attribute coding system to complete the attribute classification coding of the data. Then, the coded data is standardized to eliminate dimensional differences. Finally, through risk feature extraction and screening, a key feature set of trauma risk is obtained.

[0023] Further, such as Figure 2 As shown, in the system provided by the application embodiment, the data processing module 12 is also used for:

[0024] Perform outlier identification and data cleaning on the multidimensional data stream of trauma events to obtain a usable multidimensional data stream of trauma events; obtain data attribute factor information, the data attribute factor information includes dimension type, data format and timestamp, and construct a data attribute coding system according to the data attribute factor information; perform attribute classification coding on the available multidimensional data stream of trauma events according to the data attribute coding system to obtain multidimensional data attribute coding information; perform standardization processing on the available multidimensional data stream of trauma events based on the multidimensional data attribute coding information to obtain a standard multidimensional data stream of trauma events; perform risk feature extraction and feature screening on the standard multidimensional data stream of trauma events to obtain the trauma risk key feature set.

[0025] In the embodiment of the present application, firstly, the multidimensional data stream of trauma events is subjected to outlier identification and data cleaning. In this process, the Z-score outlier detection method is used to identify and remove outliers that exceed the mean ± 3 times the standard deviation by calculating the standard deviation of each data point. For missing data, the mean filling method is used to repair it to ensure the integrity and accuracy of the data, thereby obtaining a usable multidimensional data stream of trauma events.

[0026] To obtain data attribute factor information, we rely on preset rules for data structure. Data dimensions are identified by data type annotations, such as metadata definitions in CSV files or databases. Through this process, we obtain data attribute factor information, which includes dimension type (such as numeric, text, date, etc.), data format (such as integer, floating point number or string), and timestamp (recording the specific time of data collection).

[0027] Next, based on the extracted data attribute factor information, a data attribute encoding system is constructed. Among them, for categorical data (such as event type), the one-hot encoding method is used. One-hot encoding converts each categorical variable into a new binary feature. For example, the trauma event type field "accident" is encoded as [1,0,0], indicating that it belongs to the category of "accident"; while "natural disaster" is encoded as [0,1,0], indicating that it belongs to "natural disaster". For ordered data, the label encoding method is used. Label encoding maps each category to an integer.

[0028] After constructing the data attribute coding system, attribute classification coding is performed on each data field according to the system. When performing classification coding, for text-type data (such as "event type" and "coping method", etc.), one-hot encoding is used to convert each category into a binary vector. When label encoding, for category data with a sequential relationship, label encoding is used to map different levels to integers. When facing numerical data, such as the "severity" or "duration" of traumatic events, directly keep its numerical format. After this step, the multidimensional data stream of traumatic events is converted into unified numerical data, and multidimensional data attribute coding information is obtained.

[0029] After coding, all numerical data were standardized using the Z-score standardization method. Each data feature was subtracted from its mean and then divided by its standard deviation, so that the mean of each feature was 0 and the standard deviation was 1. This process ensured that the data of all features were within the same scale range, avoiding deviations caused by different feature dimensions, and obtaining a standard trauma event multidimensional data stream.

[0030] Finally, the standardized data was used for risk feature extraction and feature screening. The contribution of each feature was calculated using the random forest algorithm to identify the most influential features for PTSD risk prediction. By screening out the most relevant features, the key feature set of trauma risk was obtained.

[0031] Furthermore, in the system provided in the embodiment of the application, the data processing module 12 is also used for:

[0032] Risk features are extracted from the multidimensional data stream of standard traumatic events to obtain a multidimensional risk feature set of associated traumatic events; random forest training and optimization are performed based on the multidimensional risk feature set of associated traumatic events to construct a random forest model of traumatic event features; contribution calculations are performed on the multidimensional risk feature set of associated traumatic events based on the random forest model of traumatic event features to obtain a set of associated multidimensional feature contribution degrees; the multidimensional risk feature set of associated traumatic events is sorted and selected according to the set of associated multidimensional feature contribution degrees to obtain the set of key features of trauma risk.

[0033] In an embodiment of the present application, risk features are first extracted from a multidimensional data stream of a standard traumatic event, and principal component analysis is used to reduce the dimensionality of the data, and the most representative principal components are extracted from it. PCA maps the original data to these principal components by calculating the covariance matrix of the data and finding the direction that maximizes the data variance. PCA is used to extract key risk features of traumatic events, such as event type, severity, exposure, and psychological characteristics. These extracted principal components constitute a multidimensional risk feature set of associated traumatic events.

[0034] Next, based on the extracted multidimensional risk feature set of associated traumatic events, the random forest algorithm was used for training and tuning. Random forest improves the accuracy and stability of the model by building multiple decision trees and voting on their outputs. In the training phase, the model was trained through cross-validation to ensure that the model has good generalization ability, and grid search was used to tune the hyperparameters. Through this process, a random forest model of traumatic event features was obtained.

[0035] The random forest model of traumatic event features was then used to calculate the contribution of the multidimensional risk feature set associated with traumatic events. Specifically, the Gini index of each feature at the node split was first calculated. The Gini index measures the "impurity" of the data set, that is, the degree of confusion in the classification. If a feature can effectively divide the data set into relatively pure subsets, the Gini index of the feature is low, indicating that it has a good effect on the classification of the data. The initial Gini index of the entire data set was calculated, and then the data set was split according to different values ​​of the feature. The Gini index of each subset was calculated after each split. The smaller the Gini index, the higher the purity of the data set, indicating that the feature contributes more to the classification result. Then the improvement of the Gini index after the feature split was calculated. This means that by calculating the change in the Gini index before and after the feature split in the decision tree, the effect of reducing impurity when the feature splits the data is evaluated. If the feature split can significantly reduce impurity, then the feature contributes more to the data set and has a higher contribution value. Each decision tree calculates the improvement of the Gini index of the feature during the training process, and summarizes the contribution in each tree. By summing up the contributions in all trees, we can get the contribution value of each feature in the entire random forest. This contribution value reflects the importance of the feature in the model. Features with higher contributions have a greater impact on the final prediction results. Finally, the contribution values ​​of each feature are standardized to ensure that the contributions of all features are on the same scale. The standardized contribution value represents the relative importance of the feature to the model prediction results. The larger the value, the greater the contribution of the feature to the model prediction. After standardization, the sum of the contributions of all features is 1. Through this process, we get a set of associated multidimensional feature contributions containing the contribution of each feature.

[0036] Finally, the features are sorted and selected according to the associated multidimensional feature contribution set. All features are sorted from high to low according to the contribution value, and the features with the top 30% contribution are selected to proceed to the next step, while the features with smaller contribution are eliminated. In this way, the key feature set of trauma risk is obtained.

[0037] The risk prediction module 13 is used to build a PTSD risk prediction model, perform risk prediction on the trauma risk key feature set based on the PTSD risk prediction model, and output target PTSD risk prediction parameters.

[0038] In an embodiment of the present application, the risk prediction module builds and optimizes the PTSD risk prediction model based on the multidimensional feature data of the traumatic event, and then predicts the key feature set of the traumatic risk, and outputs the PTSD risk prediction parameters. Specifically, the traumatic event data is first collected and desensitized and feature extracted to obtain a multidimensional feature database of traumatic events suitable for modeling. Next, the PTSD risk prediction target is determined, and the risk prediction type, level and scope of influence are clarified by disassembling the prediction target, and then a suitable model framework set is selected. Based on these frameworks, the multidimensional feature database of traumatic events is trained, verified and optimized to construct a PTSD risk prediction model. Finally, the PTSD risk prediction model is used to predict the key feature set of trauma risk, and the target PTSD risk prediction parameters are output.

[0039] Furthermore, in the system provided in the application embodiment, the risk prediction module 13 is also used for:

[0040] Acquire a multidimensional database of traumatic events, perform desensitization processing and feature extraction on the multidimensional database of traumatic events, and obtain a multidimensional feature database of traumatic events; obtain a PTSD risk prediction target, decompose the PTSD risk prediction target into indicators, and determine the risk prediction type, risk prediction level, and risk prediction impact range; perform prediction characteristic analysis on the risk prediction type, risk prediction level, and risk prediction impact range, and select a risk prediction model framework set; train, verify, and optimize the multidimensional feature database of traumatic events based on the risk prediction model framework set, and build the PTSD risk prediction model.

[0041] In the embodiments of the present application, a multidimensional database of traumatic events is first collected, and information related to the traumatic events is collected through various data collection methods (such as questionnaires, medical records, etc.). These data include the type of traumatic event, severity, degree of exposure, psychological characteristics, etc., and also include basic information of individuals (such as age, gender, etc.). In order to ensure data security and protect user privacy, the collected multidimensional database of traumatic events is desensitized, and sensitive information is removed or hidden through technical means such as data anonymization and encryption. Afterwards, through feature extraction technology, features that are helpful in predicting PTSD risks are extracted from the original data, such as converting psychological assessment results into quantitative indicators, encoding the type and degree of exposure of traumatic events, and finally obtaining a multidimensional feature database of traumatic events suitable for modeling.

[0042] Next, we enter the stage of defining the PTSD risk prediction target, breaking down the target indicators and clarifying the target dimensions of the prediction. The goal of this stage is to determine the type, level, and scope of impact of risk prediction. First, the risk prediction type refers to the form of the prediction task, such as a binary classification model (for example, predicting whether an individual will develop PTSD) or regression prediction (predicting the probability of PTSD). Secondly, clarify the risk prediction level, such as dividing the risk into high risk, medium risk, and low risk, so that different intervention measures can be implemented according to the prediction results. Finally, determine the scope of impact of the risk prediction, that is, whether the assessment object is the risk of a single individual, or an overall risk assessment for a specific group (such as a specific population or region).

[0043] After defining the prediction target, the prediction characteristics of each target are analyzed. The purpose of this step is to deeply analyze the characteristics of each risk prediction target and provide a basis for the selection of the model framework. For example, through correlation analysis, statistical modeling and other methods, evaluate which features best reflect the risk of PTSD. For example, use the Pearson correlation coefficient or chi-square test to analyze the relationship between the exposure and severity of traumatic events and the occurrence of PTSD. Next, select a suitable risk prediction model framework set, which includes random forests, support vector machines, gradient boosting machines and neural networks. Among them, random forests are suitable for dealing with complex nonlinear relationships and common models of high-dimensional data, and can effectively capture the relationship between trauma features and PTSD risk. Support vector machines are suitable for dealing with binary classification problems, especially when the data dimension is high. Gradient boosting machines are an integrated learning method suitable for regression and classification tasks, and are particularly good at capturing complex patterns from feature data. Neural networks can deeply explore the potential complexity of data and explore deeper feature relationships under large data sets.

[0044] Finally, the selected risk prediction model framework set was trained and optimized using the multidimensional feature database of traumatic events to build a PTSD risk prediction model. Specifically, the multidimensional feature database of traumatic events was first trained for prediction targets to obtain the initial risk type prediction model, risk level prediction model, and risk impact range prediction model. Then, these initial models were verified by cross-validation to obtain the performance parameters of each model. Subsequently, based on these performance parameters, the initial model was grid search optimized and the hyperparameters were adjusted to improve the prediction effect of the model, and finally the target risk type prediction model, target risk level prediction model, and target risk impact range prediction model were generated. Finally, these optimized models were fused in series to build the final PTSD risk prediction model.

[0045] Furthermore, in the system provided in the application embodiment, the risk prediction module 13 is also used for:

[0046] The risk prediction model framework set is used to perform prediction target training on the multidimensional feature database of trauma events to obtain an initial risk type prediction model, an initial risk level prediction model and an initial risk impact range prediction model; the initial risk type prediction model, the initial risk level prediction model and the initial risk impact range prediction model are cross-validated to obtain risk type prediction model performance parameters, risk level prediction model performance parameters and risk impact range prediction model performance parameters; based on the risk type prediction model performance parameters, risk level prediction model performance parameters and risk impact range prediction model performance parameters, the initial risk type prediction model, the initial risk level prediction model and the initial risk impact range prediction model are optimized by parameter grid search to generate a target risk type prediction model, a target risk level prediction model and a target risk impact range prediction model; the target risk type prediction model, the target risk level prediction model and the target risk impact range prediction model are fused in series to build the PTSD risk prediction model.

[0047] In an embodiment of the present application, a risk prediction model framework set (such as random forest, support vector machine, gradient boosting machine, etc.) is first used to train a multidimensional feature database of traumatic events to obtain a risk type prediction model (used to predict whether PTSD will occur), a risk level prediction model (used to predict the risk level of PTSD, such as high, medium, and low risk), and a risk impact range prediction model (used to assess the impact range of the risk, such as the impact on individuals or groups).

[0048] Next, the initial risk type prediction model, the initial risk level prediction model, and the initial risk impact range prediction model were cross-validated. Specifically, the multidimensional feature database of traumatic events was divided into multiple subsets as a data set, and the K-fold cross-validation method was used. Each time, K-1 subsets were used as training sets, and the remaining subset was used as a validation set. Multiple rounds of training and validation were performed alternately to evaluate the performance of the model on different data subsets. The goal of cross-validation is to ensure that the model can make effective predictions on unseen data and avoid overfitting of the model on the training data. Through this method, the performance parameters of each model are obtained, and the performance parameters of the risk type prediction model, the risk level prediction model, and the risk impact range prediction model are obtained. These performance parameters include the accuracy of the model, etc.

[0049] Subsequently, based on the performance parameters of the risk type prediction model, the performance parameters of the risk level prediction model, and the performance parameters of the risk impact range prediction model, the initial risk type prediction model, the initial risk level prediction model, and the initial risk impact range prediction model were optimized by parameter grid search. Grid search is a technique that uses exhaustive methods to find the optimal parameter combination. In this stage, multiple hyperparameters, such as the number of trees, maximum depth, and learning rate, are adjusted to ensure that each model can maximize its performance. Grid search traverses possible hyperparameter combinations and evaluates the performance of each combination, and finally generates a target risk type prediction model, a target risk level prediction model, and a target risk impact range prediction model, ensuring that each model has strong accuracy and stability in the trauma risk prediction task.

[0050] Finally, the target risk type prediction model, target risk level prediction model and target risk impact range prediction model are fused in series to build a PTSD risk prediction model. Specifically, serial fusion is to use the output results of multiple models as the output of the PTSD risk prediction model. In this process, each model (target risk type prediction model, target risk level prediction model and target risk impact range prediction model) independently predicts its own output, and then the prediction results of these three models are combined to generate the final output of the PTSD risk prediction model.

[0051] The intervention strategy recommendation module 14 is used to construct an intervention strategy parameter space, perform a strategy analysis on the target PTSD risk prediction parameters based on the intervention strategy parameter space, and determine target intervention strategy parameters.

[0052] In the present application embodiment, the intervention strategy recommendation module first constructs the intervention strategy parameter space, which contains all possible intervention measures and their related parameters, such as treatment type, intensity, frequency, etc. The intervention strategy parameter space refers to the set of all possible intervention measures and strategy parameters. This space includes different combinations of intervention methods such as psychotherapy, drug therapy, and social support. When constructing the intervention strategy parameter space, all available intervention measures and their related parameters (such as treatment duration, treatment intensity, drug type, medication frequency, psychological support intensity, etc.) are listed by technical experts according to existing literature, and a multidimensional strategy space is established to cover different intervention schemes.

[0053] Next, the target PTSD risk prediction parameters are used as constraint parameters to match and divide the intervention strategy parameter space and determine the intervention strategy parameter thresholds that match these prediction parameters.

[0054] After obtaining the matching strategy parameter threshold, multiple intervention strategy parameters are randomly selected within the threshold, and the intervention effect of these parameters is evaluated. The fitness of each strategy, that is, the quality of the intervention effect, is obtained through evaluation. Based on these evaluation results, multiple intervention strategy parameters are clustered and allocated to form multiple intervention strategy parameter clusters. Finally, these clusters are globally compared and optimized to determine the target intervention strategy parameters.

[0055] Furthermore, in the system provided in the application embodiment, the intervention strategy recommendation module 14 is also used for:

[0056] The target PTSD risk prediction parameter is used as a constraint parameter, the intervention strategy parameter space is matched and divided, and a matching intervention strategy parameter threshold is obtained; multiple intervention strategy parameters are randomly selected within the matching intervention strategy parameter threshold, and intervention effect evaluations are performed on the multiple intervention strategy parameters to obtain multiple intervention effect fitnesses; the multiple intervention strategy parameters are clustered and allocated based on the multiple intervention effect fitnesses to obtain multiple intervention strategy parameter clusters; a global comparison and optimization is performed on the multiple intervention strategy parameter clusters to determine the target intervention strategy parameters.

[0057] In the embodiment of the present application, the target PTSD risk prediction parameter is first used as a constraint parameter to match and divide the intervention strategy parameter space. Specifically, according to the target prediction parameters (such as individual high, medium, and low risk levels, the probability of PTSD, etc.), the conditional constraint optimization method is used to screen out the appropriate range of intervention strategies. Based on these prediction parameters, the choice of intervention strategy is limited to obtain matching intervention strategy parameter thresholds, which represent the limits of treatment intensity, treatment frequency, and other intervention parameters that can be selected according to individual risk levels.

[0058] Then, based on the matching intervention strategy parameter threshold, multiple intervention strategy parameters are randomly selected and the intervention effect is evaluated. The effect of each strategy combination is predicted by the method of historical data fitting. Using regression analysis, such as random forest regression and linear regression, each group of intervention strategies is fitted based on historical patient data (such as treatment records, efficacy feedback, etc.) to evaluate its therapeutic effect on individual PTSD. Each intervention strategy will get a specific intervention effect fitness value, which indicates the performance of the intervention program in relieving symptoms or therapeutic effects. For example, for a group of randomly selected intervention strategies, assuming that cognitive behavioral therapy (CBT) combined with drug treatment is used for a high-risk individual, the fitness value of the efficacy evaluation is 0.85. This means that under this strategy, the individual's PTSD symptoms have an 85% chance of relief, and the treatment effect is good. On the other hand, the intervention strategy for another group of low-risk individuals may include mild psychological counseling and short-term drug treatment, with a fitness value of 0.65, indicating that the efficacy of the intervention program is relatively mild and suitable for low-risk individuals. Through this process, multiple intervention effect fitness values ​​are obtained.

[0059] After evaluating the intervention effect, the K-means clustering algorithm is used to cluster multiple intervention strategies and group intervention strategies with similar fitness. For example, all treatment plans with high intensity and high frequency may be clustered into one group, while low intensity and low frequency treatment plans may be clustered into another group. Through this clustering, multiple intervention strategy parameter clusters are obtained, and each cluster contains intervention plans with similar effects.

[0060] Finally, a global comparison and optimization is performed on multiple intervention strategy parameter clusters. First, the center of each cluster is determined based on the multiple intervention strategy parameter clusters. The cluster center represents the strategy parameter with the highest intervention effect fitness in the cluster. Then, a learning factor is set, and other strategy parameters in the cluster are optimized toward the cluster center based on the factor, and each strategy parameter is updated to make it closer to the optimal solution, thereby forming multiple intervention strategy optimization parameter clusters. Then, the sum of the fitness of multiple intervention strategy optimization parameter clusters is calculated and compared, the effect of each optimization cluster is evaluated, and finally the target intervention strategy optimization parameter cluster with the best fitness is selected. Finally, based on the selected target intervention strategy optimization parameter cluster, the target intervention strategy optimization parameter with the largest fitness in the target intervention strategy optimization parameter cluster is selected as the target intervention strategy parameter.

[0061] Furthermore, in the system provided in the application embodiment, the intervention strategy recommendation module 14 is also used for:

[0062] According to the multiple intervention strategy parameter clusters, multiple strategy parameter cluster centers are determined, and the multiple strategy parameter cluster centers are respectively the strategy parameters with the largest intervention effect fitness in the multiple intervention strategy parameter clusters; a learning factor is set, and the remaining strategy parameters in the multiple intervention strategy parameter clusters are learned, optimized and updated toward the multiple strategy parameter cluster centers according to the learning factor to obtain multiple intervention strategy optimization parameter clusters; the sum of the fitness of the multiple intervention strategy optimization parameter clusters is calculated, compared and optimized to obtain a target intervention strategy optimization parameter cluster, and the target intervention strategy parameters are preferably determined according to the target intervention strategy optimization parameter cluster.

[0063] In the embodiment of the present application, firstly, multiple strategy parameter cluster centers are determined based on multiple intervention strategy parameter clusters, wherein the multiple strategy parameter cluster centers are respectively the strategy parameters with the largest intervention effect fitness in the multiple intervention strategy parameter clusters. Specifically, the intervention effect fitness corresponding to the multiple intervention strategy parameters in each intervention strategy parameter cluster is compared, and the intervention strategy parameter with the highest fitness is selected as the strategy parameter cluster center. Through this process, multiple strategy parameter cluster centers are obtained.

[0064] Technical experts then set a learning factor for each cluster, and based on this factor, learn and optimize other policy parameters in the cluster toward the cluster center. The role of the learning factor is to control the step size of each optimization to ensure that the parameter update is not too fast or too slow. Using optimization algorithms such as gradient descent or particle swarm optimization, the position of each policy parameter in the cluster is gradually adjusted to ensure that it is constantly approaching the cluster center. Specifically, the difference between the cluster center and the policy parameter is calculated by the gradient descent method, and the parameters are adjusted based on this difference to gradually improve the effectiveness of the strategy. Particle swarm optimization simulates the movement of multiple particles in the search space, and the particles continuously adjust their positions to achieve the optimal solution. Through this optimization process, multiple intervention strategy optimization parameter clusters are obtained.

[0065] Next, the fitness of each intervention strategy optimization parameter in multiple intervention strategy optimization parameter clusters is calculated by the same fitness calculation method as above, and the fitness sum of the intervention strategy optimization parameter clusters is calculated, and the intervention strategy optimization parameter cluster with the largest fitness sum is taken as the target intervention strategy optimization parameter cluster. Finally, the intervention strategy optimization parameter with the largest fitness in the target intervention strategy optimization parameter cluster is taken as the target intervention strategy optimization parameter.

[0066] The intervention interaction module 15 is used to perform intervention interaction management and control on the target user based on the target intervention strategy parameters.

[0067] In the embodiment of the present application, the intervention interaction module performs intervention interaction management and control on the target user by receiving the target intervention strategy optimization parameters. Specifically, according to the target intervention strategy parameters, including treatment type, intensity, frequency and cycle, a personalized intervention plan is executed to ensure the implementation of treatment measures.

[0068] Furthermore, in the system provided in the embodiment of the application, the intervention interaction module 15 is also used for:

[0069] Based on the target intervention strategy parameters, an intervention interaction simulation evaluation is performed on the target user to obtain an intervention simulation evaluation effect; if the intervention simulation evaluation effect does not reach the preset intervention effect, the remaining strategy parameters of the target intervention strategy optimization parameter cluster are compared and optimized to obtain actual intervention strategy parameters, and intervention optimization management and control of the target user is performed through the actual intervention strategy parameters.

[0070] In the embodiment of the present application, the target user is firstly evaluated for intervention interaction simulation according to the target intervention strategy parameters. Specifically, the intervention strategy parameters of historical users and the corresponding intervention evaluation effects are firstly collected, and the effects are annotated by technical experts. Then, the BP neural network is trained by using the obtained data to obtain a virtual patient model. Next, the target intervention strategy parameters are input into the virtual patient model for prediction to obtain the intervention simulation evaluation effect.

[0071] After that, the intervention simulation evaluation effect is compared with the preset intervention effect. If some pre-simulation evaluation effects do not reach the preset intervention effect, the remaining strategy parameters of the target intervention strategy optimization parameter cluster are compared and optimized. Among them, the preset intervention effect is set by technical experts. Specifically, when comparing and optimizing the remaining strategy parameters of the target intervention strategy optimization parameter cluster, the strategy parameters with the largest fitness among the remaining strategy parameters of the target intervention strategy optimization parameter cluster are selected as the actual intervention strategy parameters.

[0072] Finally, intervention optimization and control of target users are carried out through actual intervention strategy parameters.

[0073] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0074] The present application collects and obtains a multidimensional data stream of traumatic events of a target user, wherein the multidimensional data stream of traumatic events includes event type, severity, duration, degree of exposure, gene expression data, psychological characteristics, and relevant medical history; the multidimensional data stream of traumatic events is subjected to standardization processing and risk feature screening to obtain a key feature set of traumatic risk; a PTSD risk prediction model is constructed, and risk prediction is performed on the key feature set of traumatic risk based on the PTSD risk prediction model, and target PTSD risk prediction parameters are output; an intervention strategy parameter space is constructed, and strategy analysis is performed on the target PTSD risk prediction parameters based on the intervention strategy parameter space to determine target intervention strategy parameters; and intervention interaction management and control are performed on the target user based on the target intervention strategy parameters. The present invention solves the technical problem that the prior art is not accurate enough in predicting PTSD risk, resulting in inaccurate intervention measures. The present invention collects a multidimensional data stream of traumatic events of a target user, and performs standardization processing and feature screening to extract key features of traumatic risk. A PTSD risk prediction model is constructed and predicted using a risk prediction module, and target risk parameters are output. Strategy analysis is performed based on the prediction results, and intervention strategy parameters are optimized. Finally, personalized intervention management and control are performed on the user, so as to achieve the technical effect of accurate PTSD risk prediction and intervention.

[0075] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0076] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0077] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A PTSD risk prediction system based on multidimensional data analysis, characterized in that: The system comprises: A data input module, used to collect and obtain a multi-dimensional data stream of traumatic events of a target user, wherein the multi-dimensional data stream of traumatic events includes event type, severity, duration, degree of exposure, gene expression data, psychological characteristics, and relevant medical history; A data processing module, used for performing standardization processing and risk feature screening on the multi-dimensional data stream of the trauma event to obtain a key feature set of trauma risk; A risk prediction module, used to build a PTSD risk prediction model, perform risk prediction on the trauma risk key feature set based on the PTSD risk prediction model, and output target PTSD risk prediction parameters; An intervention strategy recommendation module is used to construct an intervention strategy parameter space, perform a strategy analysis on the target PTSD risk prediction parameters based on the intervention strategy parameter space, and determine the target intervention strategy parameters; An intervention interaction module is used to perform intervention interaction management and control on the target user based on the target intervention strategy parameters.

2. The PTSD risk prediction system based on multidimensional data analysis according to claim 1, characterized in that: The key feature set of trauma risk is obtained, including: Performing outlier identification and data cleaning on the multidimensional data stream of trauma events to obtain a usable multidimensional data stream of trauma events; Acquire data attribute factor information, wherein the data attribute factor information includes dimension type, data format, and timestamp, and construct a data attribute coding system according to the data attribute factor information; Performing attribute classification encoding on the available trauma event multidimensional data stream according to the data attribute encoding system to obtain multidimensional data attribute encoding information; Standardizing the available trauma event multidimensional data stream based on the multidimensional data attribute encoding information to obtain a standard trauma event multidimensional data stream; Risk feature extraction and feature screening are performed on the multidimensional data stream of the standard trauma event to obtain the trauma risk key feature set.

3. The PTSD risk prediction system based on multidimensional data analysis according to claim 2, characterized in that: The step of obtaining the trauma risk key feature set comprises: Extracting risk features from the multidimensional data stream of the standard traumatic event to obtain a multidimensional risk feature set of associated traumatic events; Performing random forest training and tuning based on the multi-dimensional risk feature set of associated traumatic events to construct a random forest model of traumatic event features; Calculating the contribution of the multidimensional risk feature set of the associated traumatic events based on the traumatic event feature random forest model to obtain an associated multidimensional feature contribution set; The multidimensional risk feature set of the associated trauma events is sorted and selected according to the associated multidimensional feature contribution set to obtain the trauma risk key feature set.

4. The PTSD risk prediction system based on multidimensional data analysis according to claim 1, characterized in that: The PTSD risk prediction model is constructed, including: Acquiring a multidimensional database of traumatic events, performing desensitization processing and feature extraction on the multidimensional database of traumatic events, and obtaining a multidimensional feature database of traumatic events; Obtaining a PTSD risk prediction target, breaking down the PTSD risk prediction target into indicators, and determining a risk prediction type, a risk prediction level, and a risk prediction impact range; Perform prediction characteristic analysis on the risk prediction type, risk prediction level and risk prediction impact range respectively, and select a risk prediction model framework set; Based on the risk prediction model framework set, the multidimensional feature database of traumatic events is trained, verified and optimized to build the PTSD risk prediction model.

5. The PTSD risk prediction system based on multidimensional data analysis according to claim 4, characterized in that: The step of building the PTSD risk prediction model comprises: The risk prediction model framework set is used to respectively perform prediction target training on the multidimensional feature database of trauma events to obtain an initial risk type prediction model, an initial risk level prediction model, and an initial risk impact range prediction model; Cross-validating the initial risk type prediction model, the initial risk level prediction model, and the initial risk impact range prediction model to obtain performance parameters of the risk type prediction model, the risk level prediction model, and the risk impact range prediction model; Based on the risk type prediction model performance parameters, risk level prediction model performance parameters and risk impact range prediction model performance parameters, the initial risk type prediction model, the initial risk level prediction model and the initial risk impact range prediction model are optimized by parameter grid search to generate a target risk type prediction model, a target risk level prediction model and a target risk impact range prediction model; The target risk type prediction model, the target risk level prediction model and the target risk impact range prediction model are connected in series and integrated to build the PTSD risk prediction model.

6. The PTSD risk prediction system based on multidimensional data analysis according to claim 1, characterized in that: Determining target intervention strategy parameters includes: Using the target PTSD risk prediction parameter as a constraint parameter, matching and dividing the intervention strategy parameter space to obtain a matching intervention strategy parameter threshold; Randomly selecting multiple intervention strategy parameters within the matching intervention strategy parameter threshold, performing intervention effect evaluation on the multiple intervention strategy parameters, and obtaining multiple intervention effect fitnesses; Clustering and allocating the multiple intervention strategy parameters based on the multiple intervention effect fitnesses to obtain multiple intervention strategy parameter clusters; A global comparison and optimization is performed on the multiple intervention strategy parameter clusters to determine the target intervention strategy parameters.

7. The PTSD risk prediction system based on multidimensional data analysis according to claim 6, characterized in that: Determining the target intervention strategy parameters includes: Determine a plurality of strategy parameter cluster centers according to the plurality of intervention strategy parameter clusters, wherein the plurality of strategy parameter cluster centers are respectively strategy parameters with the largest intervention effect fitness among the plurality of intervention strategy parameter clusters; Setting a learning factor, and performing learning optimization and updating on the remaining strategy parameters in the multiple intervention strategy parameter clusters toward the center of the multiple strategy parameter clusters according to the learning factor, to obtain multiple intervention strategy optimization parameter clusters; The fitness sums of the multiple intervention strategy optimization parameter clusters are calculated, compared and optimized to obtain a target intervention strategy optimization parameter cluster, and the target intervention strategy parameters are preferably determined based on the target intervention strategy optimization parameter cluster.

8. The PTSD risk prediction system based on multidimensional data analysis according to claim 7, characterized in that: The intervening and interactively controlling the target user based on the target intervention strategy parameters includes: Performing an intervention interaction simulation evaluation on the target user based on the target intervention strategy parameters to obtain an intervention simulation evaluation effect; If the intervention simulation evaluation effect does not reach the preset intervention effect, the remaining policy parameters of the target intervention strategy optimization parameter cluster are compared and optimized to obtain actual intervention policy parameters, and the target user is intervened and optimized and controlled through the actual intervention policy parameters.

Images