Depression development trend prediction method and device

Abstract

The invention provides a depression development trend prediction method and device, and relates to the technical field of disease prediction, and the method comprises the steps: obtaining a time sequence of depression attack data of multiple years based on a GBD data set; processing the time sequence through a phase-space reconstruction method to obtain a state vector; extracting a maximum Lyapunov index, a bifurcation point feature and a fractal dimension based on the state vector; combining the time sequence, the maximum Lyapunov index, the bifurcation point feature and the fractal dimension into a feature vector, and obtaining a feature vector sequence in a preset time window; and normalizing and standardizing the feature vector sequence, inputting the feature vector sequence into a preset bidirectional long-short-term memory network model, and predicting the depression development trend. According to the method, the nonlinear dynamic characteristics in the depression disease development process can be revealed, and the prediction accuracy and reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of disease prediction, and in particular to a method and device for predicting the development trend of depression. Background Art

[0002] In the field of disease prediction research, accurately predicting the development trend of diseases has always been an important and challenging topic. With the acceleration of globalization and the increase in population mobility, the complexity and uncertainty of disease transmission have increased significantly. Especially for mental illnesses such as depression, their development process is often affected by multiple factors such as social environment, economic conditions, and cultural differences.

[0003] Traditional disease prediction methods mainly rely on epidemiological models and statistical methods. These methods are usually based on simple linear assumptions and assume that the development of the disease follows a certain deterministic pattern. For example, the SIR (Susceptible-Infected-Recovered) model and its variants are widely used in infectious disease prediction. However, such models show obvious limitations when dealing with complex nonlinear systems, and it is difficult to accurately capture the mutation characteristics and long-term evolution laws in the development of the disease.

[0004] With the development of data science, machine learning methods, especially deep learning technology, have been introduced into the field of disease prediction. These methods can improve the accuracy of predictions to a certain extent by learning from a large amount of historical data. However, most of these methods use a "black box" modeling approach and lack a deep understanding of the internal mechanism of disease development, resulting in a lack of interpretability of the prediction results and often poor performance when faced with system mutations.

[0005] In terms of theoretical research, chaos theory provides a new perspective for understanding the dynamic behavior of complex systems. Studies have shown that many natural and social systems have chaotic characteristics, that is, although the system is governed by deterministic laws, it is highly sensitive to initial conditions. This theory has achieved remarkable results in the fields of weather forecasting and financial market analysis. However, in the field of disease prediction, especially in the prediction of mental illness, the application of chaos theory is relatively limited, and its potential has not yet been fully explored. Summary of the invention

[0006] The purpose of the present invention is to provide a method and device for predicting the development trend of depression, which is used to solve the problems that the application of chaos theory in the prediction of mental illness in the prior art is relatively limited and the potential of chaos theory has not been fully explored. By extracting chaotic features in time series data, the nonlinear dynamic characteristics in the development process of depression are revealed, thereby improving the accuracy and reliability of the prediction.

[0007] In order to achieve the above object, in a first aspect, the present invention provides a method for predicting the development trend of depression, comprising:

[0008] Obtain time series of depression incidence data for multiple years based on the GBD dataset;

[0009] The time series is processed by phase space reconstruction method to obtain the state vector;

[0010] Extract the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension based on the state vector;

[0011] The time series, the maximum Lyapunov exponent, the bifurcation point characteristics and the fractal dimension array are synthesized into a feature vector, and then the feature vector sequence within the preset time window is obtained;

[0012] The feature vector sequence is normalized and standardized, and then input into the preset bidirectional long short-term memory network model to predict the development trend of depression.

[0013] According to a method for predicting the development trend of depression provided by the present invention, the expression of the time series is:

[0014]

[0015] In the formula, x(t) is the time series, μ is the mean, and σ is the standard deviation;

[0016]

[0017] Where Y t represents the global average incidence of depression in year t, y i,t represents the incidence of depression in the ith country or region in year t, and N is the total number of countries and regions.

[0018] According to a method for predicting the development trend of depression provided by the present invention, the expression of the state vector is:

[0019] X(t)={x(t),x(t+τ),…,x(t+(n-1)τ)}

[0020] Where X(t) is the state vector, τ is the time delay, and m is the embedding dimension.

[0021] According to a method for predicting the development trend of depression provided by the present invention, the value of time delay is 1, and the value of embedding dimension is 3.

[0022] According to a method for predicting the development trend of depression provided by the present invention, the expression of the maximum Lyapunov index is:

[0023]

[0024] Among them, λ is the maximum Lyapunov exponent; δX(0) represents the deviation vector between two adjacent orbits at the initial moment in the phase space; δX(t) represents the deviation vector between the two adjacent orbits after t time; ∥∥ represents the Euclidean norm.

[0025] According to a method for predicting the development trend of depression provided by the present invention, the expression of the bifurcation point feature is:

[0026]

[0027] Where B(t) is the bifurcation point characteristic at time t, is the absolute value of the change rate of depression disease development, θ is the change rate threshold, σ t is the local volatility index, and σ0 is the index threshold.

[0028] According to a method for predicting the development trend of depression provided by the present invention, the expression of the local volatility index is:

[0029]

[0030] Here, k is half the length of the time window.

[0031] According to a method for predicting the development trend of depression provided by the present invention, the time window is 12 months and k is 6.

[0032] According to a method for predicting the development trend of depression provided by the present invention, the expression of fractal dimension is:

[0033]

[0034] In the formula, D is the fractal dimension of the development of depression, r is the scale parameter, and C(r) is the correlation integral;

[0035]

[0036] Where i and j are time points in the phase space, and each time point includes three components: the average incidence of depression in the world at the current time t, Y t , the average incidence of depression worldwide after one year of delay Y t+1 , the average incidence of depression worldwide after two years of delay Y t+2 ; M is the number of time points, and H represents the unit step function.

[0037] In a second aspect, the present invention provides a depression development trend prediction device, comprising:

[0038] An acquisition unit is used to obtain the time series of depression incidence data for multiple years based on the GBD dataset;

[0039] A reconstruction unit, used to process the time series by a phase space reconstruction method to obtain a state vector;

[0040] An extraction unit, for extracting the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension based on the state vector;

[0041] A combination unit is used to synthesize a feature vector by combining the time series, the maximum Lyapunov exponent, the bifurcation point feature and the fractal dimension array, and then obtain a feature vector sequence within a preset time window;

[0042] The prediction unit is used to normalize and standardize the feature vector sequence, and then input it into the preset bidirectional long short-term memory network model to predict the development trend of depression.

[0043] The technical solution of the present invention has at least the following technical effects:

[0044] The present invention provides a method and device for predicting the development trend of depression, the method comprising: obtaining a time series of depression incidence data for multiple years based on the GBD data set; processing the time series by a phase space reconstruction method to obtain a state vector; extracting the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension based on the state vector; synthesizing a feature vector by combining the time series, the maximum Lyapunov exponent, the bifurcation point characteristics and the fractal dimension array, and then obtaining a feature vector sequence within a preset time window; normalizing and standardizing the feature vector sequence, and then inputting a preset bidirectional long short-term memory network model to predict the development trend of depression. The present invention organically combines chaos theory with data analysis technology, provides a method that can reflect the essential laws of disease development and has strong predictive ability, can reveal the nonlinear dynamic characteristics of depression in the development process, and improve the accuracy and reliability of prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] In the attached picture:

[0047] Figure 1 The present invention is a flowchart of the method for predicting the development trend of depression. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0049] Some embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0050] See also Figure 1 The embodiment of the present invention provides a method for predicting the development trend of depression, comprising:

[0051] Step 1: Obtain the time series of depression incidence data for multiple years based on the GBD dataset;

[0052] Step 2: Process the time series through the phase space reconstruction method to obtain the state vector;

[0053] Step 3, extracting the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension based on the state vector;

[0054] Step 4, synthesize the time series, the maximum Lyapunov exponent, the bifurcation point feature and the fractal dimension array into a feature vector, and then obtain a feature vector sequence within a preset time window;

[0055] Step 5: Normalize and standardize the feature vector sequence, and then input it into the preset bidirectional long short-term memory network model to predict the development trend of depression.

[0056] It should be noted that chaotic systems are highly sensitive to initial conditions, and their long-term behaviors are unpredictable but have inherent regularity. The core idea of ​​the present invention is based on this theory, and it is believed that the development trend of depression in the population also has chaotic characteristics. This chaotic characteristic is mainly reflected in: nonlinear interactions in the process of disease transmission, the complex influence of social environmental factors, and the coupling effect between multiple factors.

[0057] The present invention proposes three key chaos characteristic quantitative indicators:

[0058] The first is the maximum Lyapunov exponent, which is used to describe the sensitivity of the system to the initial conditions. In the development of depression, the degree of uncertainty in the development of the disease can be determined by calculating the separation rate of adjacent orbits in the phase space. A positive maximum Lyapunov exponent indicates that the system has chaotic characteristics, and its numerical value directly reflects the difficulty of prediction.

[0059] The second is the bifurcation feature, which is used to identify sudden changes in the system state. In the development of the disease, bifurcation points often correspond to important turning points in the transmission pattern, such as seasonal outbreaks, changes in the degree of herd immunity, and other key moments. By monitoring the local stability and rate of change of the system, it is possible to warn of the occurrence of these important turning points.

[0060] The third is the fractal dimension, which is used to characterize the complexity of the system. The time series of disease development often show self-similarity characteristics. By calculating its fractal dimension, this complexity can be quantified, providing an important reference for the prediction model.

[0061] The present invention selects depression data provided by the GBD data set as the research object. These data show obvious nonlinear characteristics and are very suitable for analysis using chaos theory. The specific implementation process is as follows: First, the phase space is reconstructed to map the one-dimensional time series to the high-dimensional phase space. By setting appropriate time delays and embedding dimensions, the dynamic characteristics of the system can be fully demonstrated. This step is the basis for chaos analysis. Then calculate each chaos characteristic index. By analyzing the time evolution characteristics of these indicators, the key evolution laws of the system can be identified. For example, the change in the maximum Lyapunov exponent can be used to determine whether the predictability of the system has changed; the bifurcation point characteristics can be used to warn of possible mutations; and the fractal dimension can be used to grasp the changing trend of the complexity of the system.

[0062] A detailed description is given below.

[0063] Specifically, in step 1, the present invention uses the Global Burden of Disease (GBD) dataset as the research object. The GBD dataset has the following characteristics:

[0064] Time span: Provides long-term disease development data, such as 1990 to 2019.

[0065] Geographical scope: Covering 195 countries and regions around the world.

[0066] Therefore, the number of depressive disorders can be extracted from it.

[0067] Since the original GBD dataset contains time series data from multiple geographical regions, appropriate preprocessing is required. This paper uses the time dimension aggregation method to calculate the average incidence of depression worldwide:

[0068]

[0069] Among them, Y t represents the global average incidence of depression in year t, yi,t represents the number of depression cases in the i-th country or region in year t, and N is the total number of countries and regions. t can be from 1990 to 2019, and N = 195. This preprocessing method can eliminate random fluctuations between countries and regions, highlight the overall trend of depression disease development, and provide a stable data basis for subsequent chaotic feature extraction.

[0070] The average incidence of depression worldwide t After standardization, the time series of depression incidence data for multiple years are obtained as follows:

[0071]

[0072] Where x(t) is the time series, μ is the mean, and σ is the standard deviation.

[0073] Specifically, in step 2, the expression of the state vector is:

[0074] X(t)={x(t),x(t+τ),…,x(t+(m-1)τ)}

[0075] Among them, X(t) is the state vector, τ is the time delay (the value can be 1), and m is the embedding dimension (the value can be 3).

[0076] On this basis, in order to deeply analyze the nonlinear characteristics in the development trend of depression, the present invention describes the nonlinear characteristics of the development trend of depression by analyzing the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension.

[0077] Specifically, in step 3, the expression of the maximum Lyapunov exponent is:

[0078]

[0079] Among them, λ is the maximum Lyapunov exponent, which can effectively describe the degree of chaos in the development of depression. If λ>0, the orbit deviation increases exponentially with time, indicating that the development of depression is chaotic; if λ≤0, it means that the development of depression is not chaotic; δX(0) represents the deviation vector (initial difference) between two adjacent orbits at the initial moment in the phase space; δX(t) represents the deviation vector between the two adjacent orbits after t time (the change of the difference over time); ∥∥ represents the Euclidean norm, which is used to calculate the size of the vector.

[0080] By analyzing the local stability and change rate of the time series, a bifurcation point indicator function is established:

[0081]

[0082] Where B(t) is the bifurcation point characteristic at time t, is the absolute value of the change rate of depression disease development, θ is the change rate threshold, σ t is the local volatility index, which is used to measure local volatility, and σ0 is the index threshold.

[0083]

[0084] Wherein, k is half of the length of the time window. When the time window w=12, k=6.

[0085] The bifurcation point characteristics can effectively identify the key turning points in the development of depression.

[0086] The present invention adopts the fractal dimension method to describe the fractal characteristics of the development trend of depression. The fractal dimension can reveal the self-similarity and complexity characteristics of depression in the development process. The calculation of the fractal dimension is based on the following formula:

[0087]

[0088] Where D is the fractal dimension of depression disease development, which is used to quantify the complexity of depression disease development; C(r) is the correlation integral; r is the scale parameter, which is used to measure the distance between depression incidence data at different time points in the phase space;

[0089]

[0090] Where i and j are time points in the phase space, and each time point includes three components: the average incidence of depression in the world at the current time t, Y t , the average incidence of depression worldwide after one year of delay Y t+1 , the average incidence of depression worldwide after two years of delay Y t+2 M is the number of time points. For example, if it is depression disease data from 1990 to 2019, then M = 30. H represents the Heaviside step function (unit step function), and H(x) is defined as:

[0091]

[0092] Specifically, in step 4, the constructed feature vector sequence can be expressed as:

[0093] {F t ,F t-1 ,…,F t-w+1}

[0094] Among them, F tis the feature vector at time t, including the time series of depression incidence data for multiple years, the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension, and w is the time window (the value is 12). This fusion strategy can fully capture the dynamic characteristics of the development of depression.

[0095] It should be noted that in step 5, the feature vector sequence is normalized and standardized, which is a common data preprocessing method in machine learning. The purpose is to unify the value ranges of different features for better model training and prediction. The normalization method can be maximum and minimum value normalization, which is to subtract the minimum value from each element in the feature vector and then divide it by the difference between the maximum and minimum values, so as to scale the feature vector to the range of [0,1], that is:

[0096]

[0097] In the formula, Z norm represents the normalized value, Z represents the value of each element, Z max , Z min are the maximum and minimum values ​​respectively.

[0098] Specifically, in step 5, a bidirectional long short-term memory network (Bidirectional LSTM) model is used to process the fused feature vector sequence. The model structure has the following characteristics:

[0099] 1. A two-layer bidirectional LSTM model is used to enhance the ability to capture bidirectional information of time series data.

[0100] 2. Introduce the Dropout mechanism between LSTM layers (the Dropout rate is set to 0.4) to effectively prevent overfitting.

[0101] 3. Use L2 regularization (regularization coefficient is 0.005) to further improve the generalization ability of the model.

[0102] 4. Introduce an adaptive learning rate adjustment mechanism to optimize the model training process.

[0103] Based on the same inventive concept, another embodiment of the present invention provides a depression development trend prediction device, which corresponds to the method of the above embodiment, and includes:

[0104] An acquisition unit is used to obtain the time series of depression incidence data for multiple years based on the GBD dataset;

[0105] A reconstruction unit, used to process the time series by a phase space reconstruction method to obtain a state vector;

[0106] An extraction unit, for extracting the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension based on the state vector;

[0107] A combination unit is used to synthesize a feature vector by combining the time series, the maximum Lyapunov exponent, the bifurcation point feature and the fractal dimension array, and then obtain a feature vector sequence within a preset time window;

[0108] The prediction unit is used to normalize and standardize the feature vector sequence, and then input it into the preset bidirectional long short-term memory network model to predict the development trend of depression.

[0109] The present invention shows through experimental comparison that the prediction method of the present invention, which combines chaotic characteristics and deep learning, has a significant performance improvement compared with the prior art. According to comparative experimental data, the MAPE (mean absolute percentage error) predicted by the traditional linear time series model (such as the ARIMA model (Autoregressive Integrated Moving Average model)) is 5.0%, the MAPE of the unidirectional LSTM is 2.5%, and the MAPE of the ordinary bidirectional LSTM is 1.2%, while the MAPE of the present invention is only 0.11%, and the prediction accuracy is significantly improved. At the same time, the RMSE (root mean square error) predicted by the present invention is 2500, which is significantly lower than the RMSE of 15000 predicted by ARIMA, the RMSE of 8000 predicted by unidirectional LSTM, and the RMSE of 4000 predicted by ordinary bidirectional LSTM.

[0110] The present invention can effectively capture the nonlinear characteristics of depression disease development by introducing chaos feature analysis, especially the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension analysis. Experiments show that the common bidirectional LSTM model is prone to the problem of reduced accuracy when processing long-term predictions, while the present invention significantly improves the prediction stability of the model through feature fusion strategy.

[0111] This paper applies chaos theory to the analysis of GBD data sets for the first time, innovatively combines the three chaos features of maximum Lyapunov exponent, bifurcation point feature and fractal dimension, and constructs a complete nonlinear feature description system. This paper provides a new research paradigm for the analysis of the development trend of depression and has important theoretical value.

[0112] In summary, the depression development trend prediction method and device provided by the present invention combines the core features of chaos theory with deep learning models for the first time to analyze and predict the development trend of depression in the GBD data set. By extracting the chaotic features in the time series data, the nonlinear dynamic characteristics in the development process of depression are revealed, and the accuracy and reliability of the prediction are improved. It has at least the following advantages:

[0113] 1. For the first time, chaos theory was applied to the prediction of the global disease burden of non-communicable mental illnesses. Through the combination of collaborative analysis of multi-dimensional chaos characteristics and deep learning methods, a high-precision prediction of the disease burden of depression was achieved.

[0114] 2. A new feature fusion strategy was designed to achieve an effective combination of chaotic features and time series features. The dynamic characteristics of the development of depression were effectively extracted through feature vector standardization and feature vector sequence construction.

[0115] 3. A dual judgment mechanism based on local stability and change rate was introduced in bifurcation point detection, which improved the accuracy of identifying turning points in the development of depression.

[0116] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the embodiments disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for predicting the development trend of depression, characterized in that: include: Obtain time series of depression incidence data for multiple years based on the GBD dataset; Processing the time series by a phase space reconstruction method to obtain a state vector; Extracting the maximum Lyapunov exponent, bifurcation point characteristics and fractal dimension based on the state vector; The time series, the maximum Lyapunov exponent, the bifurcation point feature and the fractal dimension array are synthesized into a feature vector, and then a feature vector sequence within a preset time window is obtained; The feature vector sequence is normalized and standardized, and then input into a preset bidirectional long short-term memory network model to predict the development trend of depression.

2. The method for predicting the development trend of depression according to claim 1, characterized in that: The expression of the time series is: In the formula, x(t) is the time series, μ is the mean, and σ is the standard deviation; Where Y t represents the global average incidence of depression in year t, y i,t represents the incidence of depression in the ith country or region in year t, and N is the total number of countries and regions.

3. The method for predicting the development trend of depression according to claim 2, characterized in that: The expression of the state vector is: X(t)={x(t),x(t+τ),…,x(t+(m-1)τ)} Where X(t) is the state vector, τ is the time delay, and m is the embedding dimension.

4. The method for predicting the development trend of depression according to claim 3, characterized in that: The value of the time delay is 1, and the value of the embedding dimension is 3.

5. The method for predicting the development trend of depression according to claim 3, characterized in that: The expression of the maximum Lyapunov exponent is: Among them, λ is the maximum Lyapunov exponent; δX(0) represents the deviation vector between two adjacent orbits at the initial moment in the phase space; δX(t) represents the deviation vector between the two adjacent orbits after t time; ∥∥ represents the Euclidean norm.

6. The method for predicting the development trend of depression according to claim 3, characterized in that: The expression of the bifurcation point characteristic is: Where B(t) is the bifurcation point characteristic at time t, is the absolute value of the change rate of depression disease development, θ is the change rate threshold, σ t is the local volatility index, and σ0 is the index threshold.

7. The method for predicting the development trend of depression according to claim 6, characterized in that: The expression of the local volatility index is: Here, k is half the length of the time window.

8. The method for predicting the development trend of depression according to claim 7, characterized in that: The time window is 12 months and k is 6.

9. The method for predicting the development trend of depression according to claim 3, characterized in that: The expression of the fractal dimension is: In the formula, D is the fractal dimension of the development of depression, r is the scale parameter, and C(r) is the correlation integral; Where i and j are time points in the phase space, and each time point includes three components: the average incidence of depression in the world at the current time t, Y t , the average incidence of depression worldwide after one year of delay Y t+1 , the average incidence of depression worldwide after two years of delay Y t+2 ; M is the number of time points, and H represents the unit step function.

10. A depression development trend prediction device, characterized in that: include: An acquisition unit is used to obtain the time series of depression incidence data for multiple years based on the GBD dataset; A reconstruction unit, used for processing the time series by a phase space reconstruction method to obtain a state vector; An extraction unit, used for extracting a maximum Lyapunov exponent, bifurcation point features and fractal dimension based on the state vector; A combining unit, used for combining the time series, the maximum Lyapunov exponent, the bifurcation point feature and the fractal dimension array into a feature vector, and then obtaining a feature vector sequence within a preset time window; The prediction unit is used to normalize and standardize the feature vector sequence, and then input it into a preset bidirectional long short-term memory network model to predict the development trend of depression.

Images