Alcohol addiction recurrence prediction method and system based on deep learning

Abstract

The invention discloses an alcohol addiction recurrence prediction method and system based on deep learning. The method comprises the following steps: S1, data acquisition, data preprocessing and feature construction; s2, constructing a prediction model based on the acquired data; s3, training and optimizing the prediction model; and S4, outputting a recurrence probability based on the prediction model, and predicting a recurrence risk of the patient through an LSTM model by collecting and processing multi-dimensional time sequence data of the patient, including physiological indexes and psychological state scores. The alcohol addiction recurrence prediction method and system based on deep learning provided by the invention have the effects of improving the accuracy of recurrence risk prediction and facilitating the formulation of a personalized treatment scheme.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a method and system for predicting alcohol addiction relapse based on deep learning. Background Art

[0002] Alcohol addiction is a global public health problem with a high relapse rate, which imposes a long-term burden on patients and their families. Traditionally, the prediction of relapse risk relies mainly on doctors' experience and static data (such as patients' health records, living habits, etc.). However, these methods often lack a comprehensive analysis of the patient's dynamic state, and therefore have great limitations in predictive accuracy.

[0003] In recent years, with the development of artificial intelligence and big data technology, time series prediction methods based on deep learning models have received increasing attention. Among them, the LSTM (Long Short-Term Memory) model has shown significant advantages in processing and analyzing time series data. The LSTM model can capture the temporal dependencies in the data, which is particularly important for prediction tasks such as alcohol addiction relapse that rely on multi-dimensional time series data of patients. Summary of the invention

[0004] In response to the above problems, a method and system for predicting alcohol addiction relapse based on deep learning is provided, aiming to solve the problems existing in the prior art.

[0005] The specific technical solutions are as follows:

[0006] A method for predicting alcohol addiction relapse based on deep learning, the method comprising the following steps:

[0007] S1, data collection and data preprocessing;

[0008] S2, building a prediction model based on the acquired data;

[0009] S3, prediction model training and optimization;

[0010] S4. Output the recurrence probability based on the prediction model.

[0011] The above-mentioned alcohol addiction relapse prediction method based on deep learning also has the following characteristics: the correction S1 comprises the following steps:

[0012] S11, setting a first interval time, and collecting the patient's physical characteristic data based on the first interval time, and the data is recorded as high-frequency data;

[0013] S12, summarizing and transmitting the collected high-frequency data to a central server;

[0014] S13, resampling, setting a second interval time, downsampling the high-frequency data based on the second interval time, and calculating the mean thereof;

[0015] S14, data standardization processing, standardizing all body feature data to ensure that different features are within a similar data range;

[0016] S15, feature reconstruction, constructing the time series feature matrix for input prediction model.

[0017] The above-mentioned alcohol addiction relapse prediction method based on deep learning also has the following characteristics: step S2 comprises the following steps:

[0018] S21. Build a prediction model and use historical patient data to train the LSTM model. The model input is the preprocessed time series data and the output is the recurrence probability in the next week.

[0019] S22. Data encryption and access restriction. All patient data input into the model is encrypted, and only authorized medical personnel can access patient data.

[0020] S23. Model update and concept drift processing: Regularly retrain the model based on new data to respond to changes in patient behavior and status, monitor model performance, and detect and adjust model performance degradation caused by changes in patient behavior.

[0021] The above-mentioned alcohol addiction relapse prediction method based on deep learning also has the following characteristics: step S3 comprises the following steps:

[0022] S31, data set division, divide the data set of the input model into training set, validation set and test set to ensure the generalization ability of the model;

[0023] S32, model training, using the training set to train the prediction model, using the adaptive momentum optimization method to minimize the loss function, where the loss function is

[0024]

[0025] In the formula,

[0026] N: total number of samples in the training set;

[0027] y i : The actual label of the i-th sample;

[0028] The predicted recurrence probability of the i-th sample;

[0029] S33, model optimization, hyperparameter optimization, adjust hyperparameters such as learning rate, batch size, and number of units in the LSTM layer through grid search and Bayesian optimization methods, optimize model performance, use the Dropout method to prevent model overfitting, randomly discard some neurons in the LSTM layer to improve the generalization ability of the model, and adopt an early stopping strategy to stop training in advance when the performance of the validation set no longer improves to prevent overfitting;

[0030] S34, model evaluation, use the test set to evaluate the performance of the model, the evaluation indicators include accuracy, area under the receiver operating characteristic curve and F1 score.

[0031] The above-mentioned alcohol addiction relapse prediction method based on deep learning also has the following characteristics: step S4 includes the following steps:

[0032] S41, risk assessment, generating a recurrence risk assessment report based on the output of the prediction model;

[0033] S42. Generate personalized treatment recommendations based on the model prediction results so that doctors can adjust the treatment plan.

[0034] A deep learning-based alcohol addiction relapse prediction system, the system comprising a data acquisition module, a data preprocessing module, a prediction model module and a result output module;

[0035] The data acquisition module is used to collect the patient's physiological data, psychological state score and fixed feature data;

[0036] The data preprocessing module is used to resample, standardize and construct features for the collected physiological data, psychological state scores and fixed feature data;

[0037] The prediction model module is used to input the pre-processed time series data and predict the recurrence probability of the patient;

[0038] The result output module is used to display the prediction results and generate personalized treatment suggestions.

[0039] The above-mentioned alcohol addiction relapse prediction system based on deep learning also has the following characteristics: the physiological data collected by the data acquisition module include heart rate, body temperature, and blood alcohol concentration, the fixed feature data collected include height and weight, and the psychological state scores collected include anxiety scores and depression scores.

[0040] The above-mentioned alcohol addiction relapse prediction system based on deep learning also has the following feature: the prediction model module includes multiple LSTM units for processing input multi-dimensional time series data.

[0041] The above-mentioned alcohol addiction relapse prediction system based on deep learning also has the following feature: the data preprocessing module performs a sliding window operation on the time series data to generate a feature vector that is input into the prediction model module.

[0042] The above-mentioned deep learning-based alcohol addiction relapse prediction system also has the following characteristics: it also includes a data privacy and security module and a model update module. The data privacy and security module uses data encryption and access control technology to ensure the security and confidentiality of patient data. The model update module regularly retrains the model to adapt to changes in patient behavior and status.

[0043] In summary, the beneficial effects of this scheme are:

[0044] In the alcohol addiction relapse prediction method and system based on deep learning provided by the present invention, the patient's relapse risk is predicted by collecting and processing multi-dimensional time series data of the patient, including physiological indicators and psychological state scores, and using the LSTM model. The alcohol addiction relapse prediction method and system based on deep learning provided by the present invention has the effect of improving the accuracy of relapse risk prediction and facilitating the formulation of personalized treatment plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flowchart of the alcohol addiction relapse prediction method based on deep learning described in the present invention;

[0046] Figure 2 This is a specific flowchart of step S1 of the alcohol addiction relapse prediction method based on deep learning described in the present invention;

[0047] Figure 3 This is a specific flowchart of step S2 of the alcohol addiction relapse prediction method based on deep learning described in the present invention;

[0048] Figure 4 This is a specific flow chart of step S3 of the alcohol addiction relapse prediction method based on deep learning described in the present invention. DETAILED DESCRIPTION

[0049] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0051] The present invention will be further described below in conjunction with specific embodiments, but the present invention is not limited thereto.

[0052] In the accompanying drawings, Figure 1 This is a flowchart of the alcohol addiction relapse prediction method based on deep learning recorded in the present invention. Figure 2 This is a specific flowchart of step S1 of the alcohol addiction relapse prediction method based on deep learning described in the present invention. Figure 3 This is a specific flow chart of step S2 of the alcohol addiction relapse prediction method based on deep learning described in the present invention. Figure 4 The specific flow chart of step S3 of the alcohol addiction relapse prediction method based on deep learning recorded in the present invention is as follows: Figure 1 As shown, the alcohol addiction relapse prediction method based on deep learning provided in this embodiment includes the following steps:

[0053] S1, data collection and data preprocessing;

[0054] S2, building a prediction model based on the acquired data;

[0055] S3, prediction model training and optimization;

[0056] S4. Output the recurrence probability based on the prediction model.

[0057] like Figure 2 As shown, in the above embodiment, step S1 includes the following steps:

[0058] S11, setting a first interval time, and collecting the patient's physical characteristic data based on the first interval time, and the data is recorded as high-frequency data;

[0059] S12, summarizing and transmitting the collected high-frequency data to a central server;

[0060] S13, resampling, setting a second interval time, downsampling the high-frequency data based on the second interval time, and calculating the mean thereof;

[0061] S14, data standardization processing, standardizing all body feature data to ensure that different features are within a similar data range;

[0062] S15, feature reconstruction, constructing the time series feature matrix for input prediction model.

[0063] It should be noted that feature reconstruction uses the sliding window method to construct the time series feature matrix:

[0064] X t =[HR t-24:t ,BT t-24:t ,BAC t-24:t ,W t,AS t ,DS t ]

[0065] In the formula,

[0066] HR t-24:t : Heart rate data series for the past 24 hours;

[0067] BT t-24:t : The temperature data series for the past 24 hours;

[0068] BAC t-24:t : Blood alcohol concentration data series for the past 24 hours;

[0069] W t : Weight of the week;

[0070] AS t : Anxiety score of the week;

[0071] DS t : Depression score for the week.

[0072] like Figure 3 As shown, in the above embodiment, step S2 includes the following steps:

[0073] S21. Build a prediction model and use historical patient data to train the LSTM model. The model input is the preprocessed time series data and the output is the recurrence probability in the next week.

[0074] S22. Data encryption and access restriction. All patient data input into the model is encrypted, and only authorized medical personnel can access patient data.

[0075] S23. Model update and concept drift processing: Regularly retrain the model based on new data to respond to changes in patient behavior and status, monitor model performance, and detect and adjust model performance degradation caused by changes in patient behavior.

[0076] It should be noted that the LSTM model formula and parameter calculation,

[0077] Forget gate calculation:

[0078] f t =σ(W f *[h t-1 ,x t ]+b f )

[0079] In the formula,

[0080] f t : The output vector of the forget gate, indicating the proportion of memory that needs to be forgotten;

[0081] σ: Sigmoid activation function, which maps the output to between 0 and 1;

[0082] W f : The weight matrix of the forget gate, used to linearly transform the current input x t and the hidden state h at the previous time step t-1 ;

[0083] h t-1 : The hidden state vector of the previous time step;

[0084] x t : Input feature vector of the current time step;

[0085] b f : Bias vector of forget gate.

[0086] Input gate calculation:

[0087] i t =σ(W i *[h t-1 ,x t ]+bx)

[0088]

[0089] In the formula,

[0090] i t : The output vector of the input gate, indicating the new information ratio;

[0091] W i : weight matrix of input gate;

[0092] b i : bias vector of input gate;

[0093] Candidate memory cell states, representing new potential memories;

[0094] W c : The weight matrix of the candidate memory unit;

[0095] b c : Bias vector of candidate memory unit;

[0096] tanh: Hyperbolic tangent function, which maps candidate memory state values ​​to between -1 and 1.

[0097] Memory unit status update:

[0098]

[0099] In the formula,

[0100] Ct : The memory cell state vector at the current time step;

[0101] C t-1 : The memory cell state vector at the previous time step;

[0102] f t *C t-1 : The forget gate determines the memory of the previous time step to be retained;

[0103] The output gate determines the new information that is added.

[0104] Output gate calculation:

[0105] O t =σ(W o *[h t-1 ,x t ]+b o )

[0106] h t =O t *tanh(C t )

[0107] In the formula,

[0108] O t : The output vector of the output gate, which controls the output of the hidden state;

[0109] W o : The weight matrix of the output gate;

[0110] b o : bias vector of output gate;

[0111] h t : The hidden state vector at the current time step, used to output the prediction.

[0112] Output prediction:

[0113]

[0114] In the formula,

[0115] Predicted probability of recurrence;

[0116] W y : The weight matrix of the output layer;

[0117] b y : Bias vector of the output layer.

[0118] like Figure 4 As shown, in the above embodiment, step S3 includes the following steps:

[0119] S31, data set division, divide the data set of the input model into training set, validation set and test set to ensure the generalization ability of the model;

[0120] It should be noted that the data is usually divided into 70% of the data for model training, 15% of the data for model validation, and the remaining 15% of the data for testing model performance.

[0121] S32, model training, using the training set to train the prediction model, using the adaptive momentum optimization method to minimize the loss function, where the loss function is

[0122]

[0123] In the formula,

[0124] N: total number of samples in the training set;

[0125] y i : The actual label of the i-th sample;

[0126] The predicted recurrence probability of the i-th sample;

[0127] S33, model optimization, hyperparameter optimization, adjust hyperparameters such as learning rate, batch size, and number of units in the LSTM layer through grid search and Bayesian optimization methods, optimize model performance, use the Dropout method to prevent model overfitting, randomly discard some neurons in the LSTM layer to improve the generalization ability of the model, and adopt an early stopping strategy to stop training in advance when the performance of the validation set no longer improves to prevent overfitting;

[0128] S34, model evaluation, use the test set to evaluate the performance of the model, the evaluation indicators include accuracy, area under the receiver operating characteristic curve and F1 score.

[0129] It should be noted that the accuracy rate indicates the proportion of samples predicted correctly; the area under the receiver operating characteristic curve is used to evaluate the performance of the classification model; the F1 score represents a comprehensive indicator that combines precision and recall, and is used for performance evaluation of unbalanced data sets.

[0130] In the above embodiment, step S4 includes the following steps:

[0131] S41, risk assessment, generating a recurrence risk assessment report based on the output of the prediction model;

[0132] S42. Generate personalized treatment recommendations based on the model prediction results so that doctors can adjust the treatment plan.

[0133] In another embodiment, a deep learning-based alcohol addiction relapse prediction system is provided, the system includes a data acquisition module, a data preprocessing module, a prediction model module, and a result output module;

[0134] A data collection module is used to collect the patient's physiological data, psychological state scores, and fixed characteristic data;

[0135] Data preprocessing module, used to resample, standardize and construct features of collected physiological data, psychological state scores and fixed feature data;

[0136] The prediction model module is used to input the preprocessed time series data and predict the recurrence probability of the patient;

[0137] The result output module is used to display the prediction results and generate personalized treatment recommendations.

[0138] It should be noted that wearable devices and other monitoring tools can be used to collect real-time physiological data from patients.

[0139] In the above embodiment, the physiological data collected by the data collection module include heart rate, body temperature, and blood alcohol concentration, the fixed feature data collected by the data collection module include height and weight, and the psychological state scores collected by the data collection module include anxiety scores and depression scores.

[0140] In the above embodiment, the prediction model module includes multiple LSTM units for processing input multi-dimensional time series data.

[0141] In the above embodiment, the data preprocessing module performs a sliding window operation on the time series data to generate a feature vector that is input into the prediction model module.

[0142] In the above embodiment, a data privacy and security module and a model update module are also included. The data privacy and security module uses data encryption and access control technology to ensure the security and confidentiality of patient data. The model update module regularly retrains the model to adapt to changes in patient behavior and status.

[0143] Working principle: This application collects multi-dimensional time series data of patients, and integrates the collected data into a data set after pre-processing. Part of the data set is used to train the prediction model, part of the data set is used to verify the prediction model, and the remaining part of the data is used to test the prediction model. Recurrence prediction of patients is performed based on the obtained model, which improves the accuracy of the prediction and helps to customize personalized treatment plans.

[0144] The above are only preferred embodiments of the present invention, and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the contents of the specification of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting alcohol addiction relapse based on deep learning, characterized in that: The method comprises the following steps: S1, data collection and data preprocessing; S2, building a prediction model based on the acquired data; S3, prediction model training and optimization; S4. Output the recurrence probability based on the prediction model.

2. The method for predicting alcohol addiction relapse based on deep learning according to claim 1, characterized in that: The correction S1 comprises the following steps: S11, setting a first interval time, and collecting the patient's physical characteristic data based on the first interval time, and the data is recorded as high-frequency data; S12, summarizing and transmitting the collected high-frequency data to a central server; S13, resampling, setting a second interval time, downsampling the high-frequency data based on the second interval time, and calculating the mean thereof; S14, data standardization processing, standardizing all body feature data to ensure that different features are within a similar data range; S15, feature reconstruction, constructing the time series feature matrix for input prediction model.

3. The method for predicting alcohol addiction relapse based on deep learning according to claim 1, characterized in that: The step S2 comprises the following steps: S21. Build a prediction model and use historical patient data to train the LSTM model. The model input is the preprocessed time series data and the output is the recurrence probability in the next week. S22. Data encryption and access restriction. All patient data input into the model is encrypted, and only authorized medical personnel can access patient data. S23. Model update and concept drift processing: Regularly retrain the model based on new data to respond to changes in patient behavior and status, monitor model performance, and detect and adjust model performance degradation caused by changes in patient behavior.

4. The method for predicting alcohol addiction relapse based on deep learning according to claim 1, characterized in that: The step S3 comprises the following steps: S31, data set division, divide the data set of the input model into training set, validation set and test set to ensure the generalization ability of the model; S32, model training, using the training set to train the prediction model, using the adaptive momentum optimization method to minimize the loss function, where the loss function is In the formula, N: total number of samples in the training set; y i : The actual label of the i-th sample; The predicted recurrence probability of the i-th sample; S33, model optimization, hyperparameter optimization, adjust hyperparameters such as learning rate, batch size, and number of units in the LSTM layer through grid search and Bayesian optimization methods, optimize model performance, use the Dropout method to prevent model overfitting, randomly discard some neurons in the LSTM layer to improve the generalization ability of the model, and adopt an early stopping strategy to stop training in advance when the performance of the validation set no longer improves to prevent overfitting; S34, model evaluation, use the test set to evaluate the performance of the model, the evaluation indicators include accuracy, area under the receiver operating characteristic curve and F1 score.

5. The method for predicting alcohol addiction relapse based on deep learning according to claim 1, characterized in that: The step S4 includes the following steps: S41, risk assessment, generating a recurrence risk assessment report based on the output of the prediction model; S42. Generate personalized treatment recommendations based on the model prediction results so that doctors can adjust the treatment plan.

6. A deep learning-based alcohol addiction relapse prediction system, characterized by: The system includes a data acquisition module, a data preprocessing module, a prediction model module and a result output module; The data acquisition module is used to collect the patient's physiological data, psychological state score and fixed feature data; The data preprocessing module is used to resample, standardize and construct features for the collected physiological data, psychological state scores and fixed feature data; The prediction model module is used to input the pre-processed time series data and predict the recurrence probability of the patient; The result output module is used to display the prediction results and generate personalized treatment suggestions.

7. The alcohol addiction relapse prediction system based on deep learning according to claim 6, characterized in that: The data acquisition module collects physiological data including heart rate, body temperature, and blood alcohol concentration, collects fixed feature data including height and weight, and collects psychological state scores including anxiety scores and depression scores.

8. The alcohol addiction relapse prediction system based on deep learning according to claim 6, characterized in that: The prediction model module includes multiple LSTM units for processing input multi-dimensional time series data.

9. The alcohol addiction relapse prediction system based on deep learning according to claim 7 or 8, characterized in that: The data preprocessing module performs a sliding window operation on the time series data to generate a feature vector that is input into the prediction model module.

10. The alcohol addiction relapse prediction system based on deep learning according to claim 9, characterized in that: It also includes a data privacy and security module and a model update module. The data privacy and security module uses data encryption and access control technology to ensure the security and confidentiality of patient data. The model update module regularly retrains the model to adapt to changes in patient behavior and status.

Images