Internet self-assessment intervention system for AIDS infection of male and male behavior persons

Abstract

The invention relates to the technical field of disease infection early warning, and discloses a male and male behavior AIDS infection internet self-assessment intervention system, which comprises a feature sequence generation module, which is used for collecting feature data of known AIDS virus infected persons and contact persons thereof to generate feature sequences; the social network graph construction module is used for constructing a social network graph; the social network graph splitting module is used for dividing the social network graph into a plurality of sub-graphs through a community detection algorithm based on modularity optimization; the risk level early warning module is used for inputting each sub-graph into an AIDS early warning model and outputting the risk level of a known AIDS virus infected person or a person in contact with the known AIDS virus infected person; the early warning measure generation module is used for generating early warning measures according to the risk levels; according to the AIDS early warning method, the social network graph is subjected to space-time analysis through the AIDS early warning model, the risk level of each person in the social network graph is automatically predicted, and the early warning measures are accurately generated according to the risk levels, so that the AIDS early warning efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of AIDS infection early warning, and more specifically, it relates to an Internet self-assessment intervention system for AIDS infection among men who have sex with men. Background Art

[0002] Acquired immunodeficiency syndrome (AIDS) is a serious infectious disease caused by the human immunodeficiency virus (HIV), which poses a major challenge to global public health. Traditional AIDS assessment and early warning mainly rely on government and medical institution personnel for offline visits and publicity. Offline visits directly contact high-risk groups for health examinations, consultation and guidance, and HIV antibody tests. Offline publicity popularizes AIDS prevention and control knowledge to the public through forms such as holding lectures and trainings to enhance people's self-protection awareness. However, traditional AIDS assessment and early warning require a large amount of manpower and material resources, have problems of low efficiency and limited coverage, and the information transmission channels are single, making it difficult to achieve precise early warning of the development of AIDS.

[0003] The transmission routes of AIDS mainly include: sexual transmission, blood transmission and mother-to-child transmission. Given that the transmission of AIDS is closely related to the social network, it is particularly important to construct a social network model based on AIDS virus-infected individuals and their contacts. Therefore, there is an urgent need for an Internet self-assessment intervention system for AIDS infection among men who have sex with men to solve the above problems. Summary of the Invention

[0004] The present invention provides an Internet self-assessment intervention system for AIDS infection among men who have sex with men to solve the technical problems in the above background art.

[0005] The present invention provides an Internet self-assessment intervention system for AIDS infection among men who have sex with men, including the following steps:

[0006] Step S101, within the first preset time period T1, collect the characteristic data of known AIDS virus-infected individuals and their contacts at a preset time interval t, and perform preprocessing to generate a characteristic sequence;

[0007] The characteristic sequence includes N sequence units, and the nth sequence unit represents the characteristic data at the nth time point after preprocessing, where 1 ≤ n ≤ N, and N = T1 / t;

[0008] The characteristic data includes: age, gender, HIV infection status, viral load, CD4+ T cell count, whether there is a history of drug use, whether there is a medication history of antiretroviral therapy, relationship with the contact, contact duration with the contact, and contact frequency with the contact;

[0009] Step S102, construct a social network graph based on the characteristic sequences of known HIV-infected individuals and their contacts;

[0010] The social network graph includes: nodes, the characteristics of nodes, and the edges between nodes;

[0011] A mapping relationship is established between the nodes and the known HIV-infected individuals or their contacts;

[0012] The characteristics of the nodes are represented by the characteristic sequences of the known HIV-infected individuals or their contacts with which the nodes establish a mapping relationship;

[0013] Edges are constructed between the nodes to represent the existence of an interaction relationship between the known HIV-infected individuals or their contacts with which the nodes establish a mapping relationship;

[0014] Step S103, divide the social network graph into multiple subgraphs through a community detection algorithm based on modularity optimization; the subgraphs are represented in the same way as the social network graph, and the number of nodes in the subgraphs is less than the number of nodes in the social network graph;

[0015] Step S104, input each subgraph into the AIDS early warning model, and the output value represents the risk level of the known HIV-infected individuals or their contacts with which the nodes in each subgraph establish a mapping relationship within the second preset time period T2;

[0016] The risk levels include: low risk level, medium risk level, and high risk level;

[0017] Step S105, generate early warning measures according to the risk levels of the known HIV-infected individuals or their contacts within the second preset time period T2.

[0018] Furthermore, the first preset time period T1, the preset time interval t, and the second preset time period T2 are all user-defined parameters.

[0019] Furthermore, preprocess the feature data, including the following steps:

[0020] Step S201, convert the gender, HIV infection status, whether there is a history of drug use, whether there is a history of taking antiretroviral therapy drugs, and the relationship with contacts in the feature data into numerical representations;

[0021] The gender is represented by the real number 0 or 1, where 0 represents female gender and 1 represents male gender;

[0022] The HIV infection status, whether there is a history of drug use, and whether there is a history of taking antiretroviral therapy drugs are all represented by the real numbers 0, 1, or 2, where 0 represents no, 1 represents unknown, and 2 represents yes;

[0023] The relationship with the contact is represented by real numbers 0, 1, or 2. 0 represents a social friend, 1 represents a family member, and 2 represents a sexual partner relationship;

[0024] Step S202: Grade the age, viral load, and CD4+ T cell count in the feature data and convert them into numerical representations;

[0025] An age less than 18 indicates being underage, represented by the real number 0. An age greater than or equal to 18 and less than 60 indicates being an adult, represented by the real number 1. An age greater than or equal to 60 is represented by the real number 2;

[0026] A viral load less than 50 indicates normal, represented by the real number 0. A viral load greater than or equal to 50 and less than 1000 indicates a low viral load, represented by the real number 1. A viral load greater than or equal to 1000 indicates a high viral load, represented by the real number 2;

[0027] A CD4+ T cell count greater than or equal to 500 indicates normal immunity, represented by the real number 0. A CD4+ T cell count less than 500 and greater than or equal to 200 indicates moderate immunosuppression, represented by the real number 1. A CD4+ T cell count less than 200 indicates severe immunosuppression, represented by the real number 2;

[0028] Step S203: Normalize each value in the feature data by the Min-Max method.

[0029] Furthermore, the social network graph is divided into multiple subgraphs by a community detection algorithm based on modularity optimization, including the following steps:

[0030] Step S301: Take each node in the social network graph as a subgraph, initialize the modularity of each subgraph to 0, and initialize the current iteration count to 1;

[0031] Step S302: Merge the subgraphs connected by edges and calculate the gain value of the modularity of the merged subgraph;

[0032] Step S303: Determine that if the gain value of the modularity of the merged subgraph is less than or equal to 0 or the number of nodes in the merged subgraph reaches the preset maximum number of nodes, no merging is performed. Otherwise, retain the merged subgraph;

[0033] Where the preset maximum number of nodes is a custom parameter;

[0034] Step S304: Increment the current iteration count by 1, and repeat Steps S302 to S303 until the iteration termination condition is met, and output multiple subgraphs;

[0035] The iteration termination conditions include: the current iteration number reaches the preset maximum iteration number; within three consecutive iteration numbers, the difference in the number of subgraphs is less than the preset difference; where the preset maximum iteration number and the preset difference are both user-defined parameters.

[0036] Further, the calculation formula for the gain value ΔQ of modularity includes:

[0037] ΔQ = Q after - Q before ;

[0038]

[0039] where Q before and Q after respectively represent the modularity of the subgraphs before and after merging, R1 and R3 respectively represent the number of edges between nodes in the subgraphs before and after merging, R2 and R4 respectively represent the number of edges between nodes in the subgraph and nodes in other subgraphs before and after merging, represents the weight value of the r1-th edge in the subgraph before merging, represents the weight value of the r2-th edge between nodes in the subgraph and nodes in other subgraphs before merging, represents the weight value of the r3-th edge in the subgraph after merging, represents the weight value of the r4-th edge between nodes in the subgraph and nodes in other subgraphs after merging, the contact duration with the contact and the number of contacts with the contact, and the weight value is obtained by weighted summing the contact duration and the number of contacts after Min-Max normalization of the known HIV-infected individuals or their contacts mapped to the nodes, and the weighting coefficients are all user-defined parameters, and the sum value of the weighting coefficients is 1.

[0040] Further, the AIDS early warning model includes: a first hidden layer, a second hidden layer, an extraction layer, and a first classifier;

[0041] The first hidden layer includes N hidden units. The i-th hidden unit inputs the i-th sequence unit of the feature sequence of a node in the subgraph and outputs a hidden vector as the feature of the node, where 1 ≤ i ≤ N;

[0042] Each hidden unit in the first hidden layer is constructed based on a gated recurrent unit;

[0043] The second hidden layer inputs the subgraph and outputs an update matrix, and each row vector of the update matrix corresponds to an update vector of a node in the subgraph;

[0044] The extraction layer is used to extract each row vector of the update matrix and input it into the first classifier. The classification space representation of the first classifier represents the risk level of known HIV-infected individuals or their contacts who have established a mapping relationship with the node corresponding to the row vector within the second preset time period T2.

[0045] Furthermore, the calculation formula of the second hidden layer is as follows:

[0046]

[0047] Rel u，v = MLP(Pile(H u , H v ));

[0048] Where Matrix represents the update matrix input to the second hidden layer, represents the update vector of the u-th node, Col u represents the set of nodes that have an edge connection with the u-th node of the subgraph input to the second hidden layer, H u and H v represent the hidden vectors of the u-th node and the v-th node respectively, Rel u，v represents the correlation coefficient between the u-th node and the v-th node, and the value range of the correlation coefficient is between 0 and 1, W 1 and W 2 represent the first weight parameter and the second weight parameter respectively, b represents the bias parameter, Pile represents the stacking operation, MLP represents the multi-layer perceptron, and Swish represents the Swish activation function.

[0049] Furthermore, the sample labels of the training samples used to train the AIDS early warning model are obtained through manual annotation, and it is pre-trained before training the AIDS early warning model. The PageRank value of each node in the subgraph is obtained by performing random walks on the subgraph. The extraction layer is used to extract each row vector of the update matrix and input it into the second classifier. The classification space representation of the second classifier represents the PageRank value of the node corresponding to the row vector;

[0050] The PageRank value of each node in the subgraph is obtained by performing random walks on the subgraph, including the following steps:

[0051] Step S401, initialize the PageRank value of each node in the subgraph to 1 / M, where M represents the number of nodes in the subgraph, and initialize the current iteration count q to 1;

[0052] Step S402, calculate the PageRank values of all nodes in the subgraph for the next iteration count;

[0053] The PageRank value of the u-th node at the current iteration number q + 1 is calculated as follows:

[0054]

[0055] where d represents the damping factor, assigned a value of 0.85, represents the PageRank value of the u-th node at the current iteration number q, col u represents the set of nodes with edge connections to the u-th node, col v represents the set of nodes with edge connections to the v-th node, L v represents the number of edges connected to the v-th node, F u，v represents the weight value of the edge between the u-th node and the v-th node, F v，k represents the weight value of the edge between the v-th node and the k-th node. The weight value is obtained by weighted summation of the contact duration and contact times after Min-Max normalization of the known HIV-infected individuals or their contacts mapped to the nodes. The weighting coefficients are all user-defined parameters, and the sum of the weighting coefficients is 1;

[0056] Step S403, increment the current iteration number q by 1, and repeat Step S402 until the iteration termination condition is met;

[0057] The iteration termination conditions include: the current iteration number q reaches the maximum iteration number Q; within two consecutive iteration numbers, the PageRank values of all nodes in the subgraph are less than or equal to the PageRank value threshold; where the maximum iteration number Q and the PageRank value threshold are both user-defined parameters.

[0058] Furthermore, the warning measures corresponding to the low-risk level include: continuous monitoring, health education, and psychological support. The warning measures corresponding to the medium-risk level include: enhanced monitoring, intensified intervention, and behavioral intervention. The warning measure corresponding to the high-risk level is emergency intervention.

[0059] The beneficial effects of the present invention are as follows: The present invention performs spatio-temporal analysis on the social network graph through the AIDS warning model, realizes automatic prediction of the risk levels of each person in the social network graph, and accurately generates warning measures according to the risk levels, thereby improving the AIDS warning efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a schematic diagram of an Internet self-assessment and intervention system for AIDS infection among men who have sex with men according to the present invention;

[0061] Figure 2It is a flowchart for preprocessing feature data in the present invention;

[0062] Figure 3 It is a flowchart for dividing a social network graph into multiple subgraphs by a community detection algorithm based on modularity optimization in the present invention;

[0063] Figure 4 It is a flowchart for obtaining the PageRank value of each node in a subgraph by performing random walks on the subgraph in the present invention.

[0064] In the figure: feature sequence generation module 101, social network graph construction module 102, social network graph splitting module 103, risk level warning module 104, warning measure generation module 105. Detailed implementation manners

[0065] Now, the subject matter described herein will be discussed with reference to exemplary embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the scope of protection of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in relation to some examples can also be combined in other examples.

[0066] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should have the ordinary meaning understood by those of ordinary skill in the art to which the present invention pertains. The "first", "second", and similar terms used in one or more embodiments of the present invention do not denote any order, quantity, or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or items appearing before this word cover the elements or items listed after this word and their equivalents, without excluding other elements or items. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0067] As Figures 1 to 4 shown, an Internet self-assessment intervention system for AIDS infection among men who have sex with men includes the following steps:

[0068] Step S101, within a first preset time period T1, collect the feature data of known AIDS virus infected persons and their contacts at a preset time interval t, and perform preprocessing to generate a feature sequence;

[0069] The feature sequence includes N sequence units, and the nth sequence unit represents the feature data at the nth time point after preprocessing, where 1 ≤ n ≤ N and N = T1 / t;

[0070] The feature data includes: age, gender, HIV infection status, viral load, CD4+ T cell count, whether there is a history of drug use, whether there is a medication history of antiretroviral therapy (ART), relationship with the contact, contact duration with the contact, and number of contacts with the contact;

[0071] Step S102, construct a social network graph based on the feature sequences of known HIV-infected individuals and their contacts;

[0072] The social network graph includes: nodes, features of the nodes, and edges between the nodes;

[0073] A mapping relationship is established between the nodes and known HIV-infected individuals or their contacts;

[0074] The features of the nodes are represented by the feature sequences of known HIV-infected individuals or their contacts with whom the nodes have established a mapping relationship;

[0075] Edges are constructed between the nodes to represent the existence of an interaction relationship between known HIV-infected individuals or their contacts with whom the nodes have established a mapping relationship;

[0076] Step S103, divide the social network graph into multiple subgraphs through a community detection algorithm based on modularity optimization;

[0077] The subgraphs are represented in the same way as the social network graph, and the number of nodes in the subgraphs is less than the number of nodes in the social network graph;

[0078] Step S104, input each subgraph into the AIDS early warning model, and the output value represents the risk level of known HIV-infected individuals or their contacts with whom the nodes have established a mapping relationship within the second preset time period T2;

[0079] The risk levels include: low risk level, medium risk level, and high risk level;

[0080] Step S105, generate early warning measures based on the risk levels of known HIV-infected individuals or their contacts within the second preset time period T2.

[0081] It should be noted that since the pathological mechanism of AIDS has not been fully clarified and it is difficult to completely cure AIDS, for known HIV-infected individuals, risk level prediction may not be necessary, and the risk level of known HIV-infected individuals can be directly set to the highest; however, the viral load of known HIV-infected individuals can also change, and if they actively follow the treatment plan recommended by the doctor, then the risk level of known HIV-infected individuals still has evaluation value.

[0082] In an embodiment of the present invention, the first preset time period T1, the preset time interval t, and the second preset time period T2 are all user-defined parameters. For short-term early warning, preferably, the first preset time period T1 is set to 6 months, the preset time interval t is set to 1 month, and the second preset time period T2 is set to 3 months. For medium- and long-term early warning, preferably, the first preset time period T1 is set to 1 year, the preset time interval t is set to 2 months, and the second preset time period T2 is set to 6 months.

[0083] In an embodiment of the present invention, as Figure 2 shown, preprocessing the feature data includes the following steps:

[0084] Step S201, convert the gender, HIV infection status, whether there is a history of drug use, whether there is a history of taking antiretroviral therapy (ART) medications, and the relationship with the contact in the feature data into numerical representations;

[0085] Gender is represented by the real number 0 or 1, where 0 represents female gender and 1 represents male gender;

[0086] The HIV infection status, whether there is a history of drug use, and whether there is a history of taking antiretroviral therapy (ART) medications are all represented by the real numbers 0, 1, or 2, where 0 represents no, 1 represents unknown, and 2 represents yes;

[0087] The relationship with the contact is represented by the real numbers 0, 1, or 2, where 0 represents social friend, 1 represents family member, and 2 represents sexual partner relationship;

[0088] The relationship with the contact can also be refined. For example, social friends can include close friends and ordinary friends, family members can include parents, children, siblings, and other relatives, and can also include drug abuse-related relationships and medical staff relationships, etc.;

[0089] Step S202, divide the age, viral load, and CD4+ T cell count in the feature data into grades and convert them into numerical representations;

[0090] An age less than 18 indicates being underage and is represented by the real number 0. An age greater than or equal to 18 and less than 60 indicates being an adult and is represented by the real number 1. An age greater than or equal to 60 is represented by the real number 2;

[0091] A viral load less than 50 indicates normal, represented by the real number 0. A viral load greater than or equal to 50 and less than 1000 indicates a low viral load, represented by the real number 1. A viral load greater than or equal to 1000 indicates a high viral load, represented by the real number 2;

[0092] A CD4+ T cell count greater than or equal to 500 indicates normal immunity, represented by the real number 0. A CD4+ T cell count less than 500 and greater than or equal to 200 indicates moderate immunosuppression, represented by the real number 1. A CD4+ T cell count less than 200 indicates severe immunosuppression, represented by the real number 2;

[0093] Step S203, normalize each value in the feature data by the Min-Max method.

[0094] In one embodiment of the present invention, as Figure 3 shown, divide the social network graph into multiple subgraphs by a community detection algorithm based on modularity optimization, including the following steps:

[0095] Step S301, take each node in the social network graph as a subgraph, initialize the modularity of each subgraph to 0, and initialize the current iteration number to 1;

[0096] Step S302, merge the subgraphs connected by edges, and calculate the gain value of the modularity of the merged subgraph;

[0097] Step S303, determine that if the gain value of the modularity of the merged subgraph is less than or equal to 0 or the number of nodes in the merged subgraph reaches the preset maximum number of nodes, then no merging is performed, otherwise retain the merged subgraph;

[0098] Wherein the preset maximum number of nodes is a custom parameter. Preferably, the preset maximum number of nodes is set to 1 / 6 of the number of nodes in the social network graph;

[0099] Step S304, increment the current iteration number by 1, and repeat steps S302 to S303 until the iteration termination condition is met, and output multiple subgraphs;

[0100] The iteration termination conditions include: the current iteration number reaches the preset maximum iteration number; within 3 consecutive iteration numbers, the difference in the number of subgraphs is less than the preset difference; wherein the preset maximum iteration number and the preset difference are both custom parameters. Preferably, the preset maximum iteration number is set to 50, and the preset difference is set to 1.

[0101] For example, within 3 consecutive iteration numbers, if the number of subgraphs is 11, 10, and 10 respectively, it means that the structure of dividing the social network graph into multiple subgraphs no longer changes, and the iteration is terminated.

[0102] In an embodiment of the present invention, the calculation formula for the gain value ΔQ of modularity includes:

[0103] ΔQ = Q after - Q before ;

[0104]

[0105] where Q before and Q after respectively represent the modularity of the subgraphs before and after merging. R1 and R3 respectively represent the number of edges between nodes in the subgraphs before and after merging. R2 and R4 respectively represent the number of edges between nodes in the subgraph and nodes in other subgraphs before and after merging. represents the weight value of the r1-th edge in the subgraph before merging, represents the weight value of the r2-th edge between nodes in the subgraph and nodes in other subgraphs before merging, represents the weight value of the r3-th edge in the subgraph after merging, represents the weight value of the r4-th edge between nodes in the subgraph and nodes in other subgraphs after merging, the contact duration with the contact and the number of contacts with the contact. The weight value is obtained by weighted summation of the contact duration and the number of contacts after Min - Max normalization of the known HIV-infected persons or their contacts that establish a mapping relationship with the node. The weighting coefficients are all user-defined parameters, and the sum value of the weighting coefficients is 1.

[0106] It should be noted that since the social network graph is usually relatively large and the number of nodes in the social network graph is relatively large, dividing the social network graph into multiple subgraphs can accelerate the computational complexity of the subsequent AIDS early warning model, thereby improving the computational speed of the AIDS early warning model.

[0107] In an embodiment of the present invention, the AIDS early warning model includes: a first hidden layer, a second hidden layer, an extraction layer, and a first classifier;

[0108] The first hidden layer includes N hidden units. The i-th hidden unit inputs the i-th sequence unit of the feature sequence of a node in the subgraph and outputs a hidden vector as the feature of the node, where 1 ≤ i ≤ N;

[0109] That is, through the first hidden layer, the features of all nodes in the subgraph are changed from being represented by the feature sequence to being represented by the hidden vector;

[0110] The second hidden layer inputs the subgraph and outputs an update matrix. Each row vector of the update matrix corresponds to an update vector of a node in the subgraph;

[0111] The extraction layer is used to extract each row vector of the update matrix and input it into the first classifier. The classification space representation of the first classifier represents the risk level of known HIV-infected individuals or their contacts who have established a mapping relationship with the corresponding node of the row vector within the second preset time period T2.

[0112] In one embodiment of the present invention, each hidden unit of the first hidden layer is constructed based on GRU (Gated Recurrent Unit), or can also be constructed based on LSTM (Long Short-Term Memory), which will not be elaborated here.

[0113] In one embodiment of the present invention, the calculation formula of the second hidden layer is as follows:

[0114]

[0115] Rel u，v = MLP(Pile(H u ,H v ));

[0116] Where Matrix represents the update matrix input to the second hidden layer, represents the update vector of the u-th node, Col u represents the set of nodes that have an edge connection with the u-th node of the subgraph input to the second hidden layer, H u and H v respectively represent the hidden vectors of the u-th node and the v-th node, Rel u，v represents the correlation coefficient between the u-th node and the v-th node, and the value range of the correlation coefficient is between 0 and 1. W 1 and W 2 respectively represent the first weight parameter and the second weight parameter, b represents the bias parameter, Pile represents the stacking operation, MLP represents the multi-layer perceptron, and Swish represents the Swish activation function.

[0117] It should be noted that the weight parameters and bias parameters in the AIDS early warning model are all learnable hyperparameters. For example, if the hidden vector of the node is designed to be 1×8 in size, then the first weight parameter and the second weight parameter can be designed as matrices of 8×16 in size, and the size of the update vector is 1×16. MLP is used to map the hidden vectors of the two stacked nodes into the correlation coefficient, and the value range of the correlation coefficient is controlled between 0 and 1 through the sigmoid activation function, which is used to represent the degree of association of one node with another node.

[0118] In one embodiment of the present invention, the sample labels of the training samples for training the AIDS early warning model are obtained through manual annotation, and pre-training is performed on it before training the AIDS early warning model. The PageRank value of each node in the subgraph is obtained by performing random walks on the subgraph. The extraction layer is used to extract each row vector of the update matrix and input it into the second classifier. The classification space representation of the second classifier represents the PageRank value of the node corresponding to the row vector;

[0119] As Figure 4 shown, obtaining the PageRank value of each node in the subgraph by performing random walks on the subgraph includes the following steps:

[0120] Step S401, initialize the PageRank value of each node in the subgraph to 1 / M, where M represents the number of nodes in the subgraph, and initialize the current iteration number q to 1;

[0121] Step S402, calculate the PageRank values of all nodes in the subgraph for the next iteration number.

[0122] The PageRank value of the u-th node with the current iteration number q + 1 is calculated as follows:

[0123]

[0124] where d represents the damping factor, assigned a value of 0.85, represents the PageRank value of the u-th node with the current iteration number q, col u represents the set of nodes connected to the u-th node by an edge, col v represents the set of nodes connected to the v-th node by an edge, L v represents the number of edges connected to the v-th node, F u，v represents the weight value of the edge between the u-th node and the v-th node, F v，k represents the weight value of the edge between the v-th node and the k-th node. The weight value is obtained by weighted summation of the normalized contact duration and contact times of known AIDS virus infected persons or their contacts who have established a mapping relationship with the node. The weighting coefficients are all user-defined parameters, and the sum value of the weighting coefficients is 1;

[0125] Step S403, increment the current iteration number q by 1, and repeat Step S402 until the iteration termination condition is met;

[0126] The iteration termination conditions include: the current iteration number q reaches the maximum iteration number Q; within two consecutive iteration numbers, the PageRank values of all nodes in the sub-graph are less than or equal to the PageRank value threshold; where the maximum iteration number Q and the PageRank value threshold are both user-defined parameters. Preferably, the maximum iteration number Q is set to 10, and the PageRank value threshold is set to 1 / 1000.

[0127] It should be noted that obtaining the PageRank value of each node in the sub-graph through random walk is completely based on the structure of the sub-graph, without manual intervention, which can greatly reduce the number of training samples for training the AIDS early warning model. Moreover, the PageRank value is highly correlated with the risk level. Pre-training and fitting the PageRank value helps to capture the complex interaction patterns between nodes, speeds up the training speed of the AIDS early warning model, and improves the robustness of the model.

[0128] In an embodiment of the present invention, the early warning measures corresponding to the low-risk level include: continuous monitoring (regular HIV testing), health education (providing personalized health education, emphasizing the importance of adhering to ART, and how to correctly take protective measures to prevent virus transmission), and psychological support (providing psychological counseling and support services for patients and their contacts to help them cope with the stress and anxiety brought by the disease and maintain mental health). The early warning measures corresponding to the medium-risk level include: enhanced monitoring (increasing the frequency of HIV testing), intensive intervention (providing medication adherence training to ensure that patients take their medications on time), and behavioral intervention (providing one-on-one counseling to help them identify and change high-risk behaviors). The early warning measure corresponding to the high-risk level is emergency intervention (immediately arranging relevant personnel to provide treatment and psychological counseling for them).

[0129] In addition, for the medium-risk level and high-risk level, they can also be reminded of their high-risk behaviors by sending text messages and asked to go to the local AIDS prevention and control center for testing.

[0130] The above has described the embodiments of this embodiment, but this embodiment is not limited to the above specific implementation manners. The above specific implementation manners are only illustrative and not restrictive. Under the inspiration of this embodiment, those of ordinary skill in the art can also make many forms, all of which fall within the protection scope of this embodiment.

Claims

1. An Internet self-assessment intervention system for HIV infection among men who have sex with men, characterized by: include: The feature sequence generation module 101 is used to collect feature data of known HIV-infected persons and their contacts at preset time intervals t within a first preset time period T1, and perform preprocessing to generate feature sequences; The feature sequence includes N sequence units, and the nth sequence unit represents the feature data of the nth time point after preprocessing, where 1≤n≤N, N=T1 / t; Characteristic data included: age, sex, HIV infection status, viral load, CD4+ T-cell count, history of drug abuse, history of antiretroviral therapy, relationship with the contact, duration of contact with the contact, and number of contacts with the contact; A social network graph construction module 102, which is used to construct a social network graph based on feature sequences of known HIV-infected persons and their contacts; A social network graph includes: nodes, features of nodes, and edges between nodes; The nodes are mapped to known HIV-infected persons or their contacts; The characteristics of the node are represented by the characteristic sequences of known HIV-infected persons or their contacts that establish a mapping relationship with the node; The edges between nodes indicate that there is a relationship between known HIV-infected persons or their contacts who have established a mapping relationship with the nodes; A social network graph splitting module 103 is used to divide the social network graph into multiple subgraphs by using a community detection algorithm based on modularity optimization; the subgraphs have the same representation as the social network graph, and the number of nodes in the subgraphs is less than the number of nodes in the social network graph; The risk level warning module 104 is used to input each subgraph into the AIDS warning model, and the output value represents the risk level of the known HIV-infected persons or persons in contact with the known HIV-infected persons who have established a mapping relationship with the node in each subgraph within the second preset time period T2; Risk levels include: low risk level, medium risk level and high risk level; The early warning measure generating module 105 is used to generate early warning measures according to the risk level of the known HIV-infected persons or persons in contact with them within the second preset time period T2.

2. According to claim 1, the Internet self-assessment intervention system for HIV infection among men who have sex with men is characterized by: The first preset time period T1, the preset time interval t, and the second preset time period T2 are all custom parameters.

3. According to claim 1, the Internet self-assessment intervention system for HIV infection among men who have sex with men is characterized by: Preprocessing the feature data includes the following steps: Step S201, converting the gender, HIV infection status, history of drug abuse, history of antiretroviral treatment and relationship with the contact in the feature data into numerical representation; Gender is represented by a real number 0 or 1, 0 represents female and 1 represents male; HIV infection status, history of drug abuse, and history of antiretroviral therapy were all represented by real numbers 0, 1, or 2, with 0 representing no, 1 representing unknown, and 2 representing yes; The relationship with the contact is represented by a real number 0, 1, or 2, where 0 represents social friends, 1 represents family members, and 2 represents sexual partners; Step S202, classifying age, viral load and CD4+T cell count in the characteristic data into grades and converting them into numerical representations; Age less than 18 indicates a minor, represented by the real number 0; age greater than or equal to 18 and less than 60 indicates an adult, represented by the real number 1; age greater than or equal to 60 indicates a real number 2; A viral load less than 50 indicates normal, represented by the real number 0; a viral load greater than or equal to 50 and less than 1000 indicates a low viral load, represented by the real number 1; a viral load greater than or equal to 1000 indicates a high viral load, represented by the real number 2; A CD4+T cell count greater than or equal to 500 indicates normal immunity, represented by the real number 0, a CD4+T cell count less than 500 and greater than or equal to 200 indicates moderate immunosuppression, represented by the real number 1, and a CD4+T cell count less than 0 indicates severe immunosuppression, represented by the real number 2; Step S203: normalize each value in the feature data using the Min-Max method.

4. According to claim 1, the Internet self-assessment intervention system for HIV infection among men who have sex with men is characterized by: The social network graph is divided into multiple subgraphs through a community detection algorithm based on modularity optimization, including the following steps: Step S301, taking each node in the social network graph as a subgraph, and initializing the modularity of each subgraph to 0, and initializing the current iteration number to 1; Step S302, merging the subgraphs with edge connections, and calculating the gain value of the modularity of the merged subgraphs; Step S303, when it is determined that the gain value of the modularity of the merged subgraph is less than or equal to 0 or the number of nodes of the merged subgraph reaches a preset maximum number of nodes, the merge is not performed, otherwise the merged subgraph is retained; The preset maximum number of nodes is a custom parameter; Step S304, the current iteration number is incremented by 1, and steps S302 to S303 are repeatedly executed until the iteration termination condition is met, and multiple subgraphs are output; The iteration termination conditions include: the current number of iterations reaches the preset maximum number of iterations; within three consecutive iterations, the difference in the number of sub-graphs is less than the preset difference; the preset maximum number of iterations and the preset difference are both custom parameters.

5. According to claim 4, the Internet self-assessment intervention system for HIV infection among men who have sex with men is characterized by: The calculation formula of the gain value ΔQ of modularity includes: ΔQ=Q after -Q before ; Where Q before and Q after Respectively represent the modularity of the subgraph before and after the merger, R1 and R3 represent the number of edges between nodes in the subgraph before and after the merger, R2 and R4 represent the number of edges between nodes in the subgraph and other subgraphs before and after the merger, represents the weight value of the r1th edge in the subgraph before merging, Indicates the weight value of the r2th edge between the node in the subgraph before merging and the nodes in other subgraphs. Represents the weight value of the r3th edge in the merged subgraph, It represents the weight value of the r4th edge between the node in the merged subgraph and the nodes in other subgraphs, the contact time and the number of contacts with the contacts. The weight value is obtained by weighted summing up the Min-Max normalized contact time and number of contacts of known HIV-infected persons or contacts with whom a mapping relationship is established. The weight coefficients are all custom parameters, and the sum of the weight coefficients is 1.

6. The Internet self-assessment intervention system for HIV infection among men who have sex with men according to claim 1 is characterized in that: The AIDS early warning model includes: a first hidden layer, a second hidden layer, an extraction layer and a first classifier; The first hidden layer includes N hidden units, the i-th hidden unit inputs the i-th sequence unit of the feature sequence of a node of the subgraph, and outputs a hidden vector as the feature of the node, where 1≤i≤N; Each hidden unit of the first hidden layer is built based on a gated recurrent network unit; The second hidden layer inputs the subgraph and outputs an update matrix, where each row vector of the update matrix corresponds to an update vector for a node in the subgraph; The extraction layer is used to extract each row vector of the update matrix and input it into the first classifier. The classification space of the first classifier represents the risk level of known HIV-infected persons or their contacts who establish a mapping relationship with the node corresponding to the row vector within the second preset time period T2.

7. The Internet self-assessment intervention system for HIV infection among men who have sex with men according to claim 6 is characterized in that: The calculation formula for the second hidden layer is as follows: Rel u，v =MLP(Pile(H u ,H v )); Where Matrix represents the update matrix of the second hidden layer input, represents the update vector of the u-th node, Col u represents the set of nodes that have edge connections with the u-th node of the second hidden layer input subgraph, H u and H v Represent the hidden vectors of the u-th node and the v-th node respectively, Rel u，v represents the association coefficient between the u-th node and the v-th node, and the value range of the association coefficient is between 0 and 1, W1 and W2 represent the first weight parameter and the second weight parameter respectively, b represents the bias parameter, Pile represents the stacking operation, MLP represents the multi-layer perceptron, and Swish represents the Swish activation function.

8. According to claim 1, the Internet self-assessment intervention system for HIV infection among men who have sex with men is characterized by: The sample labels of the training samples used to train the AIDS early warning model are obtained through manual annotation, and the AIDS early warning model is pre-trained before training. The PageRank value of each node in the subgraph is obtained by random walk on the subgraph. The extraction layer is used to extract each row vector of the update matrix and input it into the second classifier. The classification space of the second classifier represents the PageRank value of the node corresponding to the row vector. The PageRank value of each node in the subgraph is obtained by performing a random walk on the subgraph, including the following steps: Step S401, initializing the PageRank value of each node in the subgraph to 1 / M, where M represents the number of nodes in the subgraph, and initializing the current iteration number q to 1; Step S402, calculating the PageRank values ​​of all nodes in the subgraph of the next iteration number; The PageRank value of the u-th node whose current iteration number is q+1 The calculation formula is as follows: Where d represents the damping factor, which is assigned a value of 0.

85. Indicates the PageRank value of the u-th node with the current iteration number q, col u represents the set of nodes that have edges connected to the u-th node, col v represents the set of nodes that have edges connected to the vth node, L v represents the number of edges connected to the vth node, F u，v represents the weight value of the edge between the uth node and the vth node, F v，k Represents the weight value of the edge between the vth node and the kth node. The weight value is obtained by weighted summing the normalized Min-Max contact time and contact times of known HIV-infected persons or their contacts who have established a mapping relationship with the node. The weight coefficients are all custom parameters, and the sum of the weight coefficients is 1; Step S403, the current iteration number q is incremented by 1, and step S402 is repeated until the iteration termination condition is met; The iteration termination conditions include: the current iteration number q reaches the maximum iteration number Q; within 2 consecutive iterations, the PageRank values ​​of all nodes in the subgraph are less than or equal to the PageRank value threshold; the maximum iteration number Q and the PageRank value threshold are both custom parameters.

9. The Internet self-assessment intervention system for HIV infection among men who have sex with men according to claim 1 is characterized in that: The warning measures corresponding to the low risk level include: continuous monitoring, health education and psychological support; the warning measures corresponding to the medium risk level include: enhanced monitoring, intensified intervention and behavioral intervention; the warning measures corresponding to the high risk level include emergency intervention.

Images