A sleep quality assessment method and system based on network analysis

Abstract

The present invention relates to the technical field of sleep quality assessment, and provides a sleep quality assessment method based on network analysis, comprising: step 100, collecting sleep quality and mental health relationship data; step 200, using the EBICglasso model, using the sleep quality and mental health relationship data, to obtain a sleep quality assessment model; step 300, collecting the predicted sleep quality node data input by the user to be assessed, using the obtained sleep quality assessment model, analyzing the user data to be assessed, and obtaining the minimum prediction interval of sleep quality; step 400, outputting the sleep quality interval according to the obtained minimum prediction interval of sleep quality, and realizing sleep quality assessment. The present invention can improve the efficiency of sleep quality assessment.

Description

Technical Field

[0001] The present invention relates to the technical field of sleep quality assessment, and in particular to a sleep quality assessment method and system based on network analysis. Background Art

[0002] At present, sleep quality assessment is mainly carried out by collecting various physiological indicators through instruments and equipment. For example, by collecting heart rate, blood pressure, bioelectricity and other indicators, the user's mental health level or sleep quality can be measured or predicted.

[0003] However, the sleep quality assessment of the existing technology generally requires external additional equipment, which is inconvenient to use and has a strong physical sensation, affecting the user's sleep quality; for example, wearing a bracelet or wearing electrodes, resistance blocks, etc., makes the user feel uncomfortable when sleeping. Summary of the invention

[0004] The present invention mainly solves the technical problems that the sleep quality assessment in the prior art generally requires external additional equipment, is inconvenient to use, has a strong physical sensation, and affects the user's sleep quality. A sleep quality assessment method based on network analysis is proposed to improve the efficiency of sleep quality assessment.

[0005] The present invention provides a sleep quality assessment method based on network analysis, comprising the following process:

[0006] Step 100, collecting data on the relationship between sleep quality and mental health;

[0007] Step 200, using the EBICglasso model and the data on the relationship between sleep quality and mental health to obtain a sleep quality assessment model;

[0008] Step 300, collecting the predicted sleep quality node data input by the user to be evaluated, using the obtained sleep quality evaluation model to analyze the user data to be evaluated, and obtaining the minimum prediction interval of sleep quality;

[0009] Step 400: output a sleep quality interval according to the obtained minimum prediction interval of sleep quality, thereby achieving sleep quality assessment.

[0010] Furthermore, the types of mental health include but are not limited to: depression, anxiety and stress.

[0011] Furthermore, the step 200 includes the following process:

[0012] Step 201, according to the evolution path of the following EBICglasso model, determine the connection coefficient between each node of the model:

[0013] y=Xβ+∈

[0014] Among them, sleep quality, anxiety, depression, and stress are nodes of the model, X represents the input of the model, and the input parameters are any single or combined depression, anxiety, and stress, y represents the output of the model, and the output is sleep quality; for n observations and p parameters, y is the n×1 vector of results, β is the p×1 vector of β coefficients, ∈ is the n×1 error vector, and X is the n×p design matrix;

[0015] Step 202: Step 202: construct a sleep quality assessment model according to the determined connection coefficients between the nodes of the model.

[0016] Furthermore, in step 201, the connection coefficients between the nodes of the model are calibrated through the following process:

[0017] 1) Calculate the mental health index assessment error value using the following formula:

[0018] (y-Xβ) T (y-Xβ)

[0019] 2) Using the Lasso method, the error value of the mental health index assessment is adjusted to calibrate the connection coefficients between the nodes of the model; the residual sum of squares (RSS) formula is as follows:

[0020]

[0021] Here, λ is a penalty or tuning parameter introduced to avoid the problems faced by ordinary least squares.

[0022] Furthermore, the step 300 includes the following process:

[0023] Step 301, obtaining a predicted sleep quality node selected by a user to be evaluated, and obtaining psychological assessment scale data corresponding to the predicted sleep quality node filled in by the user;

[0024] Step 302: Assuming that the first predicted sleep quality node is X and the sleep quality to be predicted is Y, and the first predicted sleep quality node X is known, then according to the connection coefficient b1 between the first predicted sleep quality node X and the sleep quality to be predicted Y, the following estimation interval for the sleep quality to be predicted Y is generated:

[0025] [x-b1y,x+b1y], where x-b1y=m, x+b1y=n;

[0026] Step 303: According to the network relationship of the mental health data, the second predicted sleep quality node Z is another variable associated with the sleep quality Y to be predicted. According to the connection coefficient b2 between the first predicted sleep quality node X and the second predicted sleep quality node Z, the following estimated interval of the first predicted sleep quality node X to the second predicted sleep quality node Z is outputted:

[0027] [x-b2z,x+b2z], where x-b2z=a, x+b2z=b;

[0028] In step 304, a prediction can be made for the evaluation interval of the sleep quality Y to be predicted based on the second predicted sleep quality node Z, and two estimation intervals can be output using the maximum and minimum values ​​of the path estimation from the first predicted sleep quality node X to the second predicted sleep quality node Z, where the connection coefficient between the second predicted sleep quality node Z and the sleep quality Y to be predicted is b3, and the minimum estimation interval of the sleep quality Y value to be predicted is determined using the interval intersection as follows:

[0029] [a-b3y,a+b3y], where a-b3y=p, a+b3y=q

[0030] [b-b3y,b+b3y], where b-b3y=e, b+b3y=f

[0031] Among them, m, n, a, b, p, q, e, and f represent the mental health index respectively;

[0032] Step 305: According to the estimated interval of the sleep quality Y to be predicted generated in step 302, the estimated interval of the first predicted sleep quality node X to the second predicted sleep quality node Z output in step 303, and the minimum estimated interval of the sleep quality Y value to be predicted formed in step 304, the mental health indexes m, n, a, b, p, q, e, and f are taken on the number axis, and the interval between the fourth point and the fifth point is taken as the minimum prediction interval of the sleep quality Y value to be predicted. The average of the maximum value and the minimum value of the interval is the best prediction point of the sleep quality Y value to be predicted.

[0033] Correspondingly, the present invention also provides a sleep quality assessment system based on network analysis, comprising: a data pre-input subsystem A, a network model construction subsystem B, an assessment data input and analysis subsystem C, and an assessment data output feedback subsystem D;

[0034] The data is pre-input into subsystem A, which is used to collect data on the relationship between sleep quality and mental health;

[0035] The network model construction subsystem B is used to adopt the EBICglasso model and the data on the relationship between sleep quality and mental health to obtain a sleep quality assessment model;

[0036] The evaluation data input and analysis subsystem C is used to collect the predicted sleep quality node data input by the user to be evaluated, and use the obtained sleep quality evaluation model to analyze the user data to be evaluated to obtain the minimum prediction interval of sleep quality;

[0037] The evaluation data output feedback subsystem D is used to output a sleep quality interval according to the obtained minimum prediction interval of sleep quality, so as to achieve sleep quality evaluation.

[0038] The present invention provides a sleep quality assessment method based on network analysis, a sleep quality assessment model obtained based on the network analysis method, and uses the sleep quality assessment model to perform sleep quality assessment, thereby improving the efficiency of user sleep quality assessment. It produces multi-dimensional and multi-directional predictions of the user's sleep quality, and can output the sleep quality assessment results and suggested interpretations for the user. Compared with the sleep quality measurement method using specific biological instruments, watches and other additional hardware devices, the present invention can not only save the cost of the hardware itself, but also reduce the user's physical sensation, and can more directly and conveniently assess the user's sleep quality. The present invention uses a network analysis method. Compared with the correlation coefficient generated by the Pearson product moment correlation method or the prediction of a single linear regression, the network analysis method emphasizes the mutual influence between mental health data, and assesses the user's sleep quality level in a three-dimensional manner in the network structure, which is more accurate.

[0039] The present invention is a mental health assessment network with sleep quality as the core. Users only need to complete a small number of mental health test questions to obtain the predicted values ​​of other mental health levels including the completed questions, which greatly improves the efficiency of user psychological assessment and solves the problem that hundreds of questions in previous mental health tests take half an hour or even longer. The present invention can be applied to various APP / web pages for single-point rapid psychological assessment and prediction of sleep quality (central node), or reverse path prediction of user mental health level through sleep quality assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of the implementation of the sleep quality assessment method based on network analysis provided by the present invention;

[0041] Figure 2 is a schematic diagram of the network structure of the sleep quality assessment model provided by the present invention;

[0042] Figure 3 is an analysis schematic diagram of analyzing the user data to be evaluated provided by the present invention;

[0043] Figure 4 It is a connection block diagram of the sleep quality assessment system based on network analysis provided by the present invention. DETAILED DESCRIPTION

[0044] In order to make the technical problems solved by the present invention, the technical solutions adopted and the technical effects achieved clearer, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention are shown in the accompanying drawings, rather than all the contents.

[0045] Embodiment 1

[0046] like Figure 1 As shown, the sleep quality assessment method based on network analysis provided by the embodiment of the present invention includes the following process:

[0047] Step 100, collecting data on the relationship between sleep quality and mental health.

[0048] The sleep quality and mental health relationship data is data reflecting the relationship between mental health and sleep quality. The types of mental health include but are not limited to: depression, anxiety and stress.

[0049] This step collects the data on the relationship between sleep quality and mental health that have been formed and stored in the database; the data on the relationship between sleep quality and mental health are summarized by collecting a large amount of mental health data and sleep quality data. When the database is established, basic information data can also be collected while collecting mental health data. The basic information data includes but is not limited to gender, age, occupation, income, region, etc. When collecting data, try to collect as much heterogeneous data as possible within the subject group. For example, when applied to adult groups, it should cover as many demographic characteristics data types as possible, such as occupation, income, region, etc.

[0050] The relationship between sleep quality and mental health has the following characteristics:

[0051] First, networks that are densely connected and have more connections are considered risky networks. This is due to the snowball effect of symptoms: the activation of one symptom may lead to other strongly associated symptoms.

[0052] Second, more central nodes also lead to a snowball effect, again leading to the activation of connected nodes.

[0053] Therefore, a tightly connected network with more central nodes indicates possible risks for users.

[0054] Similarly, in network analysis, node activation or endorsement of a given symptom can be predicted in the network. Symptoms with highly predictive networks are more likely to be controlled by neighboring symptoms, indicating a more promising path for treatment and intervention. Network analysis can also determine disease severity and predict treatment attrition.

[0055] Step 200, using the EBICglasso model and the data on the relationship between sleep quality and mental health, a sleep quality assessment model is obtained.

[0056] Step 201, according to the evolution path of the following EBICglasso model, determine the connection coefficient between each node of the model:

[0057] y=Xβ+∈

[0058] Among them, sleep quality, anxiety, depression, and stress are the nodes of the model, X represents the input of the model, and the input parameters are any single or combined depression, anxiety, and stress. Y represents the output of the model, and the output is sleep quality. For n observations and p parameters, y is the n×1 vector of results, β is the p×1 vector of β coefficients, ∈ is the n×1 error vector, and X is the n×p design matrix (all 1s in the first column and the values ​​of the predictors in the second column).

[0059] The EBICglasso model used in the present invention is based on a linear regression model, and the assumption of the linear regression model is that the data are independent. Another key assumption is that the residual of the outcome variable follows a normal distribution.

[0060] Preferably, the connection coefficients between the nodes of the model are calibrated by the following process:

[0061] 1) Calculate the error value of mental health index assessment.

[0062] Note that this assumption will carry over to the case of multivariate models, such as the network approach we consider here. When we attempt to predict sleep quality with the following three depression, anxiety, and stress nodes, we assume that all variables present in the network are independent of each other and that the residuals of each variable conform to a normal distribution. The following formula is used to calculate the error value of the mental health index assessment:

[0063] (y-Xβ) T (y-Xβ)

[0064] 2) Using the Lasso method, the error value of the mental health index assessment is adjusted to calibrate the connection coefficients between the nodes of the model.

[0065] Lasso (Least absolute shrinkage and selection operator) is a regularization method that aims to minimize the residual sum of squares (RSS) by adding additional penalties. The residual sum of squares (RSS) formula is as follows:

[0066]

[0067] Among them, λ is a penalty or tuning parameter introduced to avoid the problems faced by ordinary least squares. When λ tends to 0, the equation tends to minimize RSS. As λ approaches infinity, it may face the problem of underfitting. Therefore, it is necessary to choose a suitable λ value. There are two ways to choose the λ value:

[0068] A common method is cross-validation, which is very suitable for the characteristics of network analysis methods when combined with Lasso. The combination produces Glasso (graphical lasso, graphical penalty maximum likelihood method / penalized likelihood Gaussian graph model), which is a regularization method for estimating the covariance matrix. The formula for estimating the covariance matrix error is as follows:

[0069] log(|Θ|)-tr(SΘ)-λ∑ i≠j (|Θ i,j |)

[0070] Where log(|Θ|) is the logarithm of the determinant of Θ, a p×p precision matrix. tr(SΘ) is the sum of the eigenvalues ​​S, in the p×p sample covariance matrix (an estimate of Σ after multiplying by Θ). λ∑ i≠j (|Θ i,j |) is a penalty factor that multiplies the tuning parameter λ by the sum of the absolute values ​​of the opposite elements of the covariance matrix. Glasso can be further modified to be a λ selection criterion. k-fold cross validation is often used for λ selection. However, k-fold cross validation overfits when selecting tuning parameters.

[0071] Another approach is to use information criteria to select tuning parameters. EBICglasso selects tuning parameters by minimizing the extended Bayesian information criterion.

[0072] Step 202: construct a sleep quality assessment model according to the determined connection coefficients between the nodes of the model.

[0073] In the sleep quality assessment model (network structure diagram and connection coefficients between nodes) constructed in this step, Figure 2 As shown, sleep quality P, anxiety Q1, depression Q2, and stress Q3 are nodes, and each node has a different connection coefficient with other nodes. The more information input around the predicted variable sleep quality P, the more accurate the prediction evaluation of the predicted variable.

[0074] Step 300 , collect the predicted sleep quality node data input by the user to be evaluated, use the obtained sleep quality evaluation model to analyze the user data to be evaluated, and obtain the minimum prediction interval of sleep quality.

[0075] Step 301 : obtaining a predicted sleep quality node selected by a user to be evaluated, and obtaining psychological assessment scale data corresponding to the predicted sleep quality node filled in by the user.

[0076] Step 302: Assuming that the first predicted sleep quality node is X and the sleep quality to be predicted is Y, and the first predicted sleep quality node X is known, then according to the connection coefficient b1 between the first predicted sleep quality node X and the sleep quality to be predicted Y, the following estimation interval for the sleep quality to be predicted Y is generated:

[0077] [x-b1y,x+b1y], where x-b1y=m, x+b1y=n

[0078] Step 303: According to the network relationship of the mental health data, the second predicted sleep quality node Z is another variable associated with the sleep quality Y to be predicted. According to the connection coefficient b2 between the first predicted sleep quality node X and the second predicted sleep quality node Z, the following estimated interval of the first predicted sleep quality node X to the second predicted sleep quality node Z is outputted:

[0079] [x-b2z,x+b2z], where x-b2z=a, x+b2z=b

[0080] Step 304: Similarly, a prediction can be made for the evaluation interval of the sleep quality Y to be predicted based on the second predicted sleep quality node Z, and two estimated intervals are output using the maximum and minimum values ​​of the path estimated from the first predicted sleep quality node X→the second predicted sleep quality node Z, where the connection coefficient between the second predicted sleep quality node Z and the sleep quality Y to be predicted is b3, and the minimum estimated interval of the sleep quality Y value to be predicted is determined using the interval intersection as follows:

[0081] [a-b3y,a+b3y], where a-b3y=p, a+b3y=q

[0082] [b-b3y,b+b3y], where b-b3y=e, b+b3y=f

[0083] Among them, m, n, a, b, p, q, e, and f represent mental health indexes respectively.

[0084] Step 305: According to the estimated interval of the sleep quality Y to be predicted generated in step 302, the estimated interval of the first predicted sleep quality node X to the second predicted sleep quality node Z output in step 303, and the minimum estimated interval of the sleep quality Y value to be predicted formed in step 304, the mental health indexes m, n, a, b, p, q, e, and f are taken on the number axis, and the interval between the fourth point and the fifth point is taken as the minimum prediction interval of the sleep quality Y value to be predicted. The average of the maximum value and the minimum value of the interval is the best prediction point of the sleep quality Y value to be predicted, such as Figure 3 shown.

[0085] Step 400: output a sleep quality interval according to the obtained minimum prediction interval of sleep quality, thereby achieving sleep quality assessment.

[0086] The estimated interval of the sleep quality Y result to be predicted is confirmed. If all of them are in a certain estimated interval, the interval is output as the sleep quality interval; if the estimated interval of the sleep quality Y result to be predicted spans two horizontal estimated intervals, the two horizontal estimated intervals are divided, and the interval with a larger proportion after the division is output as the sleep quality interval to achieve sleep quality assessment. The sleep quality intervals are, for example, low sleep quality, medium sleep quality and high sleep quality; or sleep quality 0-30 points, 30-70 points and 70-100 points.

[0087] Provide other variable prediction results and suggestions based on network analysis results to the user to be evaluated. Output result example:

[0088] Output: Your sleep quality score for Y may be Probably average sleep quality.

[0089] Interval output: There is a 60% chance that your Y sleep quality is at a medium sleep quality level.

[0090] The sleep quality assessment method of the present invention can be used in the psychological assessment module in the mobile phone APP, and the process and example are as follows: the user logs in to the account and starts to select a node mapping psychological assessment scale. Taking the above expression as an example, the user selects Q2 as depression, and the number of questions is about 7. After completing the answer, the user is provided with the results and suggestions of the depression assessment.

[0091] The sleep quality assessment method based on network analysis of the present invention utilizes the obtained sleep quality assessment model. The mental health level value of any node input by the user is obtained, and an estimated interval can be generated according to its predicted value for the two connected points. The midpoint of the smallest interval where multiple intervals overlap is the best predicted value for the variable. For example, through the depression node and its connected coefficients, the covariation of the connected nodes can be inferred, that is, after inputting the data, the estimated value is obtained according to the interval algorithm of the minimum error estimate. The EBICglasso model adopted in the present invention is a network analysis model (Network AnalysisModel), which is used to evaluate sleep quality. Network analysis can identify the risk of psychopathology. This model has the ability to identify risk factors and predict nodes.

[0092] Embodiment 2

[0093] like Figure 4As shown, the sleep quality assessment system based on network analysis provided by the embodiment of the present invention includes: a data pre-input subsystem A, a network model construction subsystem B, an assessment data input and analysis subsystem C and an assessment data output feedback subsystem D;

[0094] The data is pre-input into subsystem A, which is used to collect data on the relationship between sleep quality and mental health;

[0095] The network model construction subsystem B is used to adopt the EBICglasso model and the data on the relationship between sleep quality and mental health to obtain a sleep quality assessment model;

[0096] The evaluation data input and analysis subsystem C is used to collect the predicted sleep quality node data input by the user to be evaluated, and use the obtained sleep quality evaluation model to analyze the user data to be evaluated to obtain the minimum prediction interval of sleep quality;

[0097] The evaluation data output feedback subsystem D is used to output a sleep quality interval according to the obtained minimum prediction interval of sleep quality, so as to achieve sleep quality evaluation.

[0098] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that modifying the technical solutions described in the aforementioned embodiments, or replacing some or all of the technical features therein by equivalents, does not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A sleep quality assessment method based on network analysis, characterized in that: The process includes: Step 100, collecting data on the relationship between sleep quality and mental health; Step 200, using the EBICglasso model and the data on the relationship between sleep quality and mental health, to obtain a sleep quality assessment model; said step 200 includes the following steps 201 to 202: Step 201, according to the evolution path of the following EBICglasso model, determine the connection coefficient between each node of the model: y=xβ+∈ Among them, sleep quality, anxiety, depression, and stress are nodes of the model, x represents the input of the model, and the input parameters are any single or combined depression, anxiety, and stress, y represents the output of the model, and the output is sleep quality; for n observations and p parameters, y is the n×1 vector of results, β is the p×1 vector of β coefficients, ∈ is the n×1 error vector, and x is the n×p design matrix; Step 202, constructing a sleep quality assessment model according to the determined connection coefficients between the nodes of the model; Step 300, collect the predicted sleep quality node data input by the user to be evaluated, use the obtained sleep quality evaluation model to analyze the user data to be evaluated, and obtain the minimum prediction interval of sleep quality; step 300 includes the following steps 301 to 305: Step 301, obtaining a predicted sleep quality node selected by a user to be evaluated, and obtaining psychological assessment scale data corresponding to the predicted sleep quality node filled in by the user; Step 302: Assuming that the first predicted sleep quality node is x, the sleep quality to be predicted is y, and the first predicted sleep quality node x is known, then according to the connection coefficient b1 between the first predicted sleep quality node x and the sleep quality to be predicted y, generate the following estimation interval for the sleep quality to be predicted y: [x-b1y,x+b1y], where x-b1y=m, x+b1y=n; Step 303: According to the network relationship of the mental health data, the second predicted sleep quality node z is another variable associated with the sleep quality y to be predicted. According to the connection coefficient b2 between the first predicted sleep quality node x and the second predicted sleep quality node z, the following estimated interval of the first predicted sleep quality node x to the second predicted sleep quality node z is outputted: [x-b2z,x+b2z], where x-b2z=a, x+b2z=b; Step 304, predict the evaluation interval of the sleep quality y to be predicted according to the second predicted sleep quality node z, and output two estimated intervals using the maximum and minimum values ​​of the path estimated from the first predicted sleep quality node x→the second predicted sleep quality node z, where the connection coefficient between the second predicted sleep quality node z and the sleep quality y to be predicted is b3, and the minimum estimated interval of the sleep quality y to be predicted is determined using the interval intersection as follows: [a-b3y,a+b3y], where a-b3y=p, a+b3y=q [b-b3y,b+b3y], where b-b3y=e, b+b3y=f Among them, m, n, a, b, p, q, e, and f represent the mental health index respectively; Step 305: based on the estimated interval of the sleep quality y to be predicted generated in step 302, the estimated interval of the first predicted sleep quality node x to the second predicted sleep quality node z output in step 303, and the minimum estimated interval of the sleep quality y value to be predicted formed in step 304, the mental health indexes m, n, a, b, p, q, e, and f are taken on the number axis, and the interval between the fourth point and the fifth point is taken as the minimum predicted interval of the sleep quality y value to be predicted. The average of the maximum value and the minimum value of the interval is the best prediction point of the sleep quality y value to be predicted. Step 400: output a sleep quality interval according to the obtained minimum prediction interval of sleep quality, thereby achieving sleep quality assessment.

2. The sleep quality assessment method based on network analysis according to claim 1, characterized in that: The types of mental health include: depression, anxiety and stress.

3. The sleep quality assessment method based on network analysis according to claim 1, characterized in that: In step 201, the connection coefficients between the nodes of the model are calibrated by the following process: 1) Calculate the mental health index assessment error value using the following formula: (y-xβ) T (y-xβ) 2) Using the Lasso method, the error value of the mental health index assessment is adjusted to calibrate the connection coefficients between the nodes of the model; the residual square sum formula is as follows: Here, λ is a penalty or tuning parameter introduced to avoid the problems faced by ordinary least squares.

4. A sleep quality assessment system based on network analysis, used to execute the sleep quality assessment method based on network analysis according to any one of claims 1 to 3, characterized in that: The sleep quality assessment system based on network analysis includes: a data pre-input subsystem A, a network model construction subsystem B, an assessment data input and analysis subsystem C, and an assessment data output feedback subsystem D; The data pre-input subsystem A is used to execute the step 100; The network model construction subsystem B is used to execute the step 200; The evaluation data input and analysis subsystem C is used to execute the step 300; The evaluation data output feedback subsystem D is used to execute the step 400.

Images