A Smart Data Analysis Method for Hospital Infection Control

Abstract

This invention relates to the field of intelligent medical data analysis, specifically to an intelligent data analysis method for hospital infection control. The method includes the following steps: multi-source infection control behavior data collection, infection control behavior feature modeling, infection control behavior correlation modeling, graph-based behavior pattern learning, multi-modal fusion infection risk assessment calculation, and dynamic evolution analysis of infection risk. This invention, by constructing an infection control behavior relationship graph and introducing a joint embedding representation mechanism of behavior nodes and related features, achieves modeling of complex correlations between multi-source infection control behaviors, overcoming the limitations of existing technologies that rely solely on single rules or independent data indicators for analysis. Furthermore, this invention, by constructing an anomaly identification mechanism based on the deviation between predicted and actual behaviors, and combining cross-modal attention fusion and feature contribution analysis methods, achieves dynamic assessment and interpretable analysis of infection risk.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical data analysis, specifically to an intelligent data analysis method for hospital infection control. Background Technology

[0002] With the continuous development of hospital informatization and smart healthcare technologies, hospital infection control is gradually shifting from traditional manual inspection methods to intelligent data analysis methods. Currently, most hospital infection control systems primarily use rule-based thresholds or single data indicators for analysis, independently judging hand hygiene practices, medical operation standardization, or environmental monitoring indicators. They lack the overall modeling capability to understand the relationships between multiple infection control behaviors, making it difficult to effectively characterize the potential transmission links and dynamic interactions between different infection control behaviors. This results in low accuracy in infection risk identification and a tendency for missed or misjudged abnormal behaviors. Furthermore, existing technologies typically use static statistical methods to assess infection risk, lacking analytical mechanisms for the temporal evolution of infection control behaviors and the deviation between predicted and actual behaviors. This makes it difficult to dynamically identify and provide early warnings of potential infection risks. Existing risk assessment models lack effective integration and interpretation capabilities regarding the importance of characteristics from different sources, leading to poor interpretability of infection risk assessment results and hindering the provision of effective data support and decision-making basis for hospital infection control. Summary of the Invention

[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an intelligent data analysis method for hospital infection control. This invention constructs an infection control behavior relationship graph and introduces a joint embedding representation mechanism of behavior nodes and associated features. This enables the modeling of complex relationships between multi-source infection control behaviors, overcoming the limitations of existing technologies that rely solely on single rules or independent data indicators. This effectively characterizes the dynamic relationships and potential infection transmission relationships between different infection control behaviors, improving the accuracy of infection risk identification. Furthermore, this invention constructs an anomaly identification mechanism based on the deviation between predicted and actual behaviors, combined with cross-modal attention fusion and feature contribution analysis methods. This enables dynamic assessment and interpretable analysis of infection risks, effectively identifying abnormal behaviors and key risk factors in the infection control process, thereby improving infection risk early warning capabilities and providing a reliable basis for hospital infection control decisions.

[0004] The technical solution adopted by this invention is as follows: This invention provides an intelligent data analysis method for hospital infection control, which includes the following steps:

[0005] Step S1: Multi-source infection control behavior data collection. Collect multi-source data during the hospital infection control process, including medical staff operation behavior data, medical operation log data, environmental monitoring data, equipment usage records and patient status data. Standardize and time-align the multi-source data to obtain pre-processed data.

[0006] Step S2: Sensory control behavior feature modeling, extract features from preprocessed data, construct sensory control behavior feature vectors, map different types of sensory control behavior feature vectors to behavior nodes, and form a unified behavior representation space;

[0007] Step S3: Model the relationship between sensory control behaviors. Based on the behavior nodes, construct the relationship between the feature vectors of sensory control behaviors to form a sensory control behavior relationship graph.

[0008] Step S4: Based on graph model behavior pattern learning, model and learn the relationship graph of infection control behavior, construct a behavior pattern prediction model based on the time series information of multi-source data in step S1, obtain the predicted behavior features of the corresponding behavior nodes, compare the predicted behavior features with the infection control behavior feature vector, calculate the behavior deviation value, and extract infection control abnormal features based on the behavior deviation value.

[0009] Step S5: Multimodal fusion infection risk assessment calculation, which integrates infection prevention and control abnormal features with infection control behavior feature vectors to construct an infection risk assessment model and calculate the infection risk score for the corresponding user;

[0010] Step S6: Dynamic evolution analysis of infection risk. Based on the user's infection risk score, analyze the dynamic trend of infection risk and generate an infection risk level.

[0011] Furthermore, step S3 specifically includes the following steps:

[0012] Step S31: Constructing multidimensional infection control behavior relationships. Based on behavior nodes and corresponding infection control behavior feature vectors, extract multidimensional association features between behavior nodes, including medical operation time interval features, operation intensity change features, and infection control measure execution response features. Based on the multidimensional association features, construct the initial association relationship between behavior nodes.

[0013] Step S32: Joint Embedding Representation of Behavior Nodes and Relationship Features. A joint mapping process is performed on the behavior nodes and multidimensional association features to obtain the embedded representation of the behavior nodes, expressed in the following form:

[0014] ;

[0015] in, Indicates the first Feature vectors of each behavioral node Represents behavior nodes With behavioral nodes The correlation features between them Indicates the embedding of mapping functions;

[0016] Step S33: Adaptive relation filtering based on association strength. Based on the embedded representation of behavior nodes, calculate the association strength between any two behavior nodes. The association strength is defined as follows:

[0017] ;

[0018] in, Indicates the first Embedded representation of each behavior node Represents behavior nodes With behavioral nodes The strength of the correlation between them;

[0019] Based on the strength of association, the association relationships between behavioral nodes are filtered to obtain the final behavioral association relationships;

[0020] Step S34: Multi-type behavior relationship hierarchical modeling. Based on the final behavior association, the behavior nodes are used as graph nodes and the final behavior association is used as graph edges to construct the sensory control behavior relationship graph.

[0021] Furthermore, step S4 specifically includes the following steps:

[0022] Step S41: Construction of temporal features of behavior relationship diagram. Based on the temporal series information of sensory control behavior relationship diagram and multi-source data, extract the temporal features of each behavior node within a continuous time window and construct the temporal feature sequence of behavior nodes.

[0023] Step S42: Behavioral pattern prediction modeling based on the behavioral relationship graph. Based on the structure of the behavioral relationship graph and the temporal feature sequence of the behavioral nodes, a behavioral pattern prediction model is constructed to predict the sensory control behavioral feature vector of each behavioral node at the current moment, and obtain the corresponding predicted behavioral features, which are expressed as follows:

[0024] ;

[0025] in, Represents a node At any moment Predictive behavioral characteristics This represents the embedding representation of the node at the previous time step. Represents a node The set of adjacent nodes, Represents the prediction function;

[0026] Step S43: Calculate the behavior deviation value. Based on the predicted behavior features and the sensory control behavior feature vector, calculate the behavior deviation value for each behavior node. The behavior deviation value is defined as follows:

[0027] ;

[0028] in: Represents the feature vector of sensory control behavior. Indicates predictive behavioral characteristics, Indicates the behavioral deviation value;

[0029] Step S44: Abnormal behavior identification based on deviation distribution, statistical analysis of the behavior deviation value of each behavior node, determination of the abnormal judgment threshold based on the deviation distribution characteristics, marking behavior nodes whose behavior deviation value exceeds the threshold, and obtaining a set of abnormal behavior nodes for infection control.

[0030] Step S45: Extract abnormal features of infection control. Based on the set of abnormal behavior nodes in infection control, extract abnormal features of infection control.

[0031] Furthermore, step S5 specifically includes the following steps:

[0032] Step S51: Multimodal feature construction. Based on the infection control behavior feature vector and infection prevention and control anomaly features, features from different sources and with different semantics are classified and processed to construct a multimodal feature set, as shown below:

[0033] ;

[0034] in, Indicates the first A subset of modal features Represents a set of multimodal features;

[0035] Step S52: Cross-modal attention fusion calculation, weighted fusion of multimodal feature sets to obtain a unified fusion feature vector, which is defined as follows:

[0036] ;

[0037] ;

[0038] in, Indicates time The fused feature vector, Indicates the first The modality at time... Feature representation, Indicates the first Attention weights for each modality Indicates the first The importance score function value of each modal feature Indicates the number of modal feature subsets. Index variables representing subsets of modal features. Indicates the first The importance score function value of each modality feature;

[0039] Step S53: Construction of an infection risk assessment model based on behavioral bias enhancement. Based on the fusion of feature vectors and behavioral bias features, an infection risk assessment model is constructed to calculate the user's risk at time [time value missing]. The infection risk score is expressed as follows:

[0040] ;

[0041] in, Indicates the user at a certain time. Infection risk score, Represents the fused feature vector. This represents a behavioral deviation feature vector consisting of a set of behavioral deviation values. Represents the model weight parameters. Indicates the bias term. Indicates the activation function;

[0042] Step S54: Individualized adaptive weight adjustment. Based on the stability of the object's historical infection control behavior, the parameters of the infection risk assessment model are adaptively adjusted. The degree of fluctuation in user behavior is calculated based on multi-source data and infection risk scores, as shown below:

[0043] ;

[0044] in, Indicates the degree of fluctuation in user behavior. This represents a sequence of infection risk scores within a time window. This is the variance calculation function;

[0045] The weights of the infection risk assessment model are adjusted based on the degree of fluctuation in user behavior, using the following adjustment formula:

[0046] ;

[0047] in, This represents the adjusted weight parameters. Indicates the adjustment coefficient;

[0048] Step S55: Infection risk level classification. Based on the infection risk score, interval mapping is performed to obtain the infection risk level. The mapping function is defined as follows:

[0049] ;

[0050] in, Indicates the user at a certain time. The level of infection risk, Represents a piecewise mapping function;

[0051] Step S56: Infection risk feature contribution analysis. Based on the fused feature vector, calculate the contribution of each feature to the infection risk score to obtain the infection risk interpretation results. The feature contribution is defined as:

[0052] ;

[0053] in, Indicates the first Each feature at time Contribution to infection risk score Represents the first in the fused feature vector Each feature component.

[0054] The beneficial effects achieved by the present invention using the above solution are as follows:

[0055] (1) By constructing a relationship diagram of infection control behavior and introducing a joint embedding representation mechanism of behavior nodes and associated features, this invention realizes the modeling of complex relationships between multi-source infection control behaviors, overcomes the limitations of existing technologies that only analyze based on a single rule or independent data indicators, and can effectively characterize the dynamic relationship between different infection control behaviors and the potential infection transmission relationship, thereby improving the accuracy of infection risk identification.

[0056] (2) By constructing an anomaly identification mechanism based on the deviation between predicted behavior and actual behavior, and combining cross-modal attention fusion and feature contribution analysis methods, this invention realizes dynamic assessment and interpretable analysis of infection risk, which can effectively identify abnormal behavior and key risk factors in the infection prevention and control process, thereby improving the early warning capability of infection risk and providing a reliable basis for hospital infection prevention and control decisions. Attached Figure Description

[0057] Figure 1 This is a flowchart illustrating an intelligent data analysis method for hospital infection control provided by the present invention.

[0058] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0059] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0060] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0061] Example 1, see Figure 1 This invention provides an intelligent data analysis method for hospital infection control, which includes the following steps:

[0062] Step S1: Multi-source infection control behavior data collection. Collect multi-source data during the hospital infection control process, including medical staff operation behavior data, medical operation log data, environmental monitoring data, equipment usage records and patient status data. Standardize and time-align the multi-source data to obtain pre-processed data.

[0063] Step S2: Sensory control behavior feature modeling, extract features from preprocessed data, construct sensory control behavior feature vectors, map different types of sensory control behavior feature vectors to behavior nodes, and form a unified behavior representation space;

[0064] Step S3: Model the relationship between sensory control behaviors. Based on the behavior nodes, construct the relationship between the feature vectors of sensory control behaviors to form a sensory control behavior relationship graph.

[0065] Step S4: Based on graph model behavior pattern learning, model and learn the relationship graph of infection control behavior, construct a behavior pattern prediction model based on the time series information of multi-source data in step S1, obtain the predicted behavior features of the corresponding behavior nodes, compare the predicted behavior features with the infection control behavior feature vector, calculate the behavior deviation value, and extract infection control abnormal features based on the behavior deviation value.

[0066] Step S5: Multimodal fusion infection risk assessment calculation, which integrates infection prevention and control abnormal features with infection control behavior feature vectors to construct an infection risk assessment model and calculate the infection risk score for the corresponding user;

[0067] Step S6: Dynamic evolution analysis of infection risk. Based on the user's infection risk score, analyze the dynamic trend of infection risk and generate an infection risk level.

[0068] Example 2, this example is based on the above example, step S3 specifically includes the following steps:

[0069] Step S31: Constructing multidimensional infection control behavior relationships. Based on behavior nodes and corresponding infection control behavior feature vectors, extract multidimensional association features between behavior nodes, including medical operation time interval features, operation intensity change features, and infection control measure execution response features. Based on the multidimensional association features, construct the initial association relationship between behavior nodes.

[0070] Step S32: Joint Embedding Representation of Behavior Nodes and Relationship Features. A joint mapping process is performed on the behavior nodes and multidimensional association features to obtain the embedded representation of the behavior nodes, expressed in the following form:

[0071] ;

[0072] in, Indicates the first Embedded representation of each behavior node Indicates the first Feature vectors of each behavioral node Represents behavior nodes With behavioral nodes The correlation features between them Indicates the embedding of mapping functions;

[0073] Step S33: Adaptive relation filtering based on association strength. Based on the embedded representation of behavior nodes, calculate the association strength between any two behavior nodes. The association strength is defined as follows:

[0074] ;

[0075] in, Represents behavior nodes With behavioral nodes The strength of the correlation between them;

[0076] Based on the strength of association, the association relationships between behavioral nodes are filtered to obtain the final behavioral association relationships;

[0077] Step S34: Multi-type behavior relationship hierarchical modeling. Based on the final behavior association, the behavior nodes are used as graph nodes and the final behavior association is used as graph edges to construct the sensory control behavior relationship graph.

[0078] In this embodiment, the code used is as follows:

[0079] import numpy as np

[0080] # ============================================

[0081] # Example 2: Implementation Code for Infection Control Behavioral Relationship Modeling

[0082] # ============================================

[0083] print("========== Hospital Infection Control Behavioral Relationship Modeling==========\n")

[0084] # ============================================

[0085] # Step S31: Building a Multidimensional Relationship Between Infection Control and Behavior

[0086] # ============================================

[0087] # Nurse A's hand hygiene behavior node feature vector

[0088] x1 = np.array([0.82, 0.76, 0.71])

[0089] # Feature vector of disinfection behavior nodes for ventilator equipment

[0090] x2 = np.array([0.63, 0.58, 0.66])

[0091] print("Nurse A's hand hygiene behavior feature vector x1:")

[0092] print(x1)

[0093] print("\nVentilator equipment disinfection behavior node feature vector x2:")

[0094] print(x2)

[0095] # Multidimensional association features

[0096] # [Time interval characteristics, operational intensity variation characteristics, infection control measure execution response characteristics]

[0097] r12 = np.array([0.42, 0.68, 0.74])

[0098] print("\nBehavior node associated feature r12:")

[0099] print(r12)

[0100] # ============================================

[0101] # Step S32: Joint Embedding Representation of Behavior Nodes

[0102] # h_i = f(x_i, r_ij)

[0103] # ============================================

[0104] def embedding_mapping(x, r):

[0105] """

[0106] Joint embedding mapping function

[0107] Employing a weighted fusion approach to simulate embedded representation

[0108] """

[0109] return 0.7 * x + 0.3 * r

[0110] # Node Embedding Representation

[0111] h1 = embedding_mapping(x1, r12)

[0112] h2 = embedding_mapping(x2, r12)

[0113] print("\nNurse A's behavior node embedding representation h1:")

[0114] print(np.round(h1, 3))

[0115] print("\nEquipment disinfection node embedded representation h2:")

[0116] print(np.round(h2, 3))

[0117] # ============================================

[0118] # Step S33: Correlation Strength Calculation

[0119] # w_ij = (h_i · h_j) / (||h_i|| * ||h_j||)

[0120] # ============================================

[0121] def cosine_similarity(a, b):

[0122] numerator = np.dot(a, b)

[0123] denominator = np.linalg.norm(a) * np.linalg.norm(b)

[0124] return numerator / denominator

[0125] # Calculate the association strength

[0126] w12 = cosine_similarity(h1, h2)

[0127] print("\nBehavior node association strength w12:")

[0128] print(round(w12, 3))

[0129] # ============================================

[0130] # Adaptive Relationship Filtering

[0131] # ============================================

[0132] threshold = 0.75

[0133] print("\nAssociation strength screening threshold:")

[0134] print(threshold)

[0135] if w12 > threshold:

[0136] relation_result = "Preserve the association relationship"

[0137] else:

[0138] relation_result = "Remove associations"

[0139] print("\nRelationship filtering results:")

[0140] print(relation_result)

[0141] # ============================================

[0142] # Step S34: Construction of Infection Control Behavior Relationship Diagram

[0143] # ============================================

[0144] # Construct a behavior relationship graph (adjacency list format)

[0145] behavior_graph = {}

[0146] if w12 > threshold:

[0147] behavior_graph["Nurse A's Hand Hygiene Node"] = ["Ventilator Equipment Disinfection Node"]

[0148] behavior_graph["ventilator equipment disinfection node"] = ["nurse A's hand hygiene node"]

[0149] print("\n========== Infection Control Behavioral Relationship Diagram==========")

[0150] for node, neighbors in behavior_graph.items():

[0151] print(f"{node} --> {neighbors}")

[0152] # ============================================

[0153] # Potential Infection Transmission Risk Analysis

[0154] # ============================================

[0155] print("\n========= Risk Analysis==========")

[0156] if w12 > 0.9:

[0157] print("A highly correlated propagation path has been detected")

[0158] print("Nurse A frequently touches the breathing equipment")

[0159] print("Equipment disinfection behavior is highly coupled with personnel contact behavior")

[0160] print("There is a potential risk of cross-infection")

[0161] else:

[0162] print("The current risk of transmission is within a normal range")

[0163] # ============================================

[0164] # Output the final analysis results

[0165] # ============================================

[0166] print("\n========== Final Result==========")

[0167] print(f"Node 1 Feature Vector x1: {x1}")

[0168] print(f"Node 2 feature vector x2: {x2}")

[0169] print(f"Associated feature r12: {r12}")

[0170] print(f"\nNode embedding representation h1: {np.round(h1, 3)}")

[0171] print(f"Node embedding representation h2: {np.round(h2, 3)}")

[0172] print(f"\nAssociation strength w12: {round(w12, 3)}")

[0173] print(f"Relation filtering results: {relation_result}")

[0174] print("\nInfection control behavior relationship diagram construction complete").

[0175] Example 3, this example is based on the above example, step S4 specifically includes the following steps:

[0176] Step S41: Construction of temporal features of behavior relationship diagram. Based on the temporal series information of sensory control behavior relationship diagram and multi-source data, extract the temporal features of each behavior node within a continuous time window and construct the temporal feature sequence of behavior nodes.

[0177] Step S42: Behavioral pattern prediction modeling based on the behavioral relationship graph. Based on the structure of the behavioral relationship graph and the temporal feature sequence of the behavioral nodes, a behavioral pattern prediction model is constructed to predict the sensory control behavioral feature vector of each behavioral node at the current moment, and obtain the corresponding predicted behavioral features, which are expressed as follows:

[0178] ;

[0179] in, Represents a node At any moment Predictive behavioral characteristics This represents the embedding representation of the node at the previous time step. Represents a node The set of adjacent nodes, Represents the prediction function;

[0180] Step S43: Calculate the behavior deviation value. Based on the predicted behavior features and the sensory control behavior feature vector, calculate the behavior deviation value for each behavior node. The behavior deviation value is defined as follows:

[0181] ;

[0182] in: Represents the feature vector of sensory control behavior. Indicates predictive behavioral characteristics, Indicates the behavioral deviation value;

[0183] Step S44: Abnormal behavior identification based on deviation distribution, statistical analysis of the behavior deviation value of each behavior node, determination of the abnormal judgment threshold based on the deviation distribution characteristics, marking behavior nodes whose behavior deviation value exceeds the threshold, and obtaining a set of abnormal behavior nodes for infection control.

[0184] Step S45: Extract abnormal features of infection control. Based on the set of abnormal behavior nodes in infection control, extract abnormal features of infection control.

[0185] In this city's embodiment, a hospital infection control intelligent analysis system was deployed in the intensive care unit of a tertiary hospital. The system continuously collects hand hygiene records of medical staff, patient contact behavior records, medical equipment disinfection records, and ward environmental monitoring data, and constructs a corresponding infection control behavior relationship diagram. The behavior nodes include medical staff operation nodes, patient contact nodes, equipment disinfection nodes, and environmental status nodes. The nodes are connected through time correlation, spatial correlation, and contact transmission relationships.

[0186] Using a 30-minute continuous time window, temporal features are extracted from the historical behavioral data of each behavioral node to form a temporal feature sequence for the behavioral node. For the hand hygiene behavioral node corresponding to Nurse A, the following temporal features are extracted within the continuous time window:

[0187] Number of times hand hygiene is performed per unit of time;

[0188] Hand hygiene response time after contact with a patient;

[0189] Duration of hand hygiene practice;

[0190] The implementation status of disinfection after handling high-frequency contact equipment;

[0191] The system obtains the node embedding representation of nurse A at the previous moment as [0.72, 0.65, 0.81];

[0192] And obtain the set of adjacent nodes that are associated with nurse A, including:

[0193] Patient P1 contact behavior node;

[0194] Disinfection procedures for ventilator equipment;

[0195] ward environment status nodes;

[0196] Based on the behavioral relationship graph structure and the temporal feature sequence of behavioral nodes, a behavioral pattern prediction model is constructed to predict the infection control behavioral characteristics of nurse A at the current moment, according to the following formula:

[0197] ;

[0198] The predicted behavioral characteristics of nurse A are obtained as follows:

[0199] =[0.83,0.79,0.88];

[0200] The features in each dimension correspond to:

[0201] Degree of adherence to hand hygiene standards;

[0202] Timeliness of disinfection response;

[0203] The degree of risk control over contact behavior;

[0204] Obtain the feature vector of nurse A's actual infection control behavior at the current moment:

[0205] =[0.51,0.46,0.52];

[0206] Calculate the behavioral deviation value based on predicted behavioral characteristics and actual sensory control behavioral characteristics:

[0207] ;

[0208] The calculation yielded the following results:

[0209] =0.59;

[0210] Statistical analysis was performed on the behavioral deviation values ​​corresponding to all behavioral nodes in the ICU ward to obtain the distribution of behavioral deviation values ​​within the current time window. Based on historical deviation distribution data, the system dynamically determined the anomaly judgment threshold to be 0.45.

[0211] Due to the behavioral deviation value corresponding to Nurse A:

[0212] 0.59 > anomaly detection threshold;

[0213] Therefore, the behavior node corresponding to Nurse A is marked as an abnormal behavior node for infection control and added to the set of abnormal behavior nodes for infection control.

[0214] Abnormal features were extracted from nodes exhibiting abnormal behaviors related to infection control, resulting in the following abnormal features:

[0215] The frequency of hand hygiene practices has decreased significantly;

[0216] Delayed disinfection response after patient contact;

[0217] Disinfection procedures were not performed promptly after high-frequency contact equipment was used.

[0218] Increased contact with high-risk patients;

[0219] Analysis of the ward environment revealed that Nurse A frequently came into contact with patient P1, who was at risk of respiratory infection, within a continuous time window. At the same time, the disinfection frequency of the corresponding ventilator equipment was lower than the normal threshold. Therefore, the system generated an infection risk warning and sent a high-risk behavior alert to the hospital infection control management platform.

[0220] Example 4: This example is based on the above examples. Step S5 specifically includes the following steps:

[0221] Step S51: Multimodal feature construction. Based on the infection control behavior feature vector and infection prevention and control anomaly features, features from different sources and with different semantics are classified and processed to construct a multimodal feature set, as shown below:

[0222] ;

[0223] in, Indicates the first A subset of modal features Represents a set of multimodal features;

[0224] Step S52: Cross-modal attention fusion calculation, weighted fusion of multimodal feature sets to obtain a unified fusion feature vector, which is defined as follows:

[0225] ;

[0226] ;

[0227] in, Indicates time The fused feature vector, Indicates the first The modality at time... Feature representation, Indicates the first Attention weights for each modality Indicates the first The importance score function value of each modal feature Indicates the number of modal feature subsets. Index variables representing subsets of modal features. Indicates the first The importance score function value of each modality feature;

[0228] Step S53: Construction of an infection risk assessment model based on behavioral bias enhancement. Based on the fusion of feature vectors and behavioral bias features, an infection risk assessment model is constructed to calculate the user's risk at time [time value missing]. The infection risk score is expressed as follows:

[0229] ;

[0230] in, Indicates the user at a certain time. Infection risk score, Represents the fused feature vector. This represents a behavioral deviation feature vector consisting of a set of behavioral deviation values. Represents the model weight parameters. Indicates the bias term. Indicates the activation function;

[0231] Step S54: Individualized adaptive weight adjustment. Based on the stability of the object's historical infection control behavior, the parameters of the infection risk assessment model are adaptively adjusted. The degree of fluctuation in user behavior is calculated based on multi-source data and infection risk scores, as shown below:

[0232] ;

[0233] in, Indicates the degree of fluctuation in user behavior. This represents a sequence of infection risk scores within a time window. This is the variance calculation function;

[0234] The weights of the infection risk assessment model are adjusted based on the degree of fluctuation in user behavior, using the following adjustment formula:

[0235] ;

[0236] in, This represents the adjusted weight parameters. Indicates the adjustment coefficient;

[0237] Step S55: Infection risk level classification. Based on the infection risk score, interval mapping is performed to obtain the infection risk level. The mapping function is defined as follows:

[0238] ;

[0239] in, Indicates the user at a certain time. The level of infection risk, Represents a piecewise mapping function;

[0240] Step S56: Infection risk feature contribution analysis. Based on the fused feature vector, calculate the contribution of each feature to the infection risk score to obtain the infection risk interpretation results. The feature contribution is defined as:

[0241] ;

[0242] in, Indicates the first Each feature at time Contribution to infection risk score Represents the first in the fused feature vector Each feature component.

[0243] In this embodiment, the code used is as follows:

[0244] import numpy as np

[0245] # =========================

[0246] # Step S51: Multimodal Feature Construction

[0247] # =========================

[0248] # Hand hygiene behavioral characteristics

[0249] F1_t = np.array([0.72, 0.65, 0.81])

[0250] # Characteristics of Equipment Disinfection Behavior

[0251] F2_t = np.array([0.53, 0.61, 0.48])

[0252] # Patient contact behavior characteristics

[0253] F3_t = np.array([0.79, 0.84, 0.76])

[0254] # Environmental Monitoring Characteristics

[0255] F4_t = np.array([0.58, 0.63, 0.55])

[0256] # Multimodal feature set

[0257] F = [F1_t, F2_t, F3_t, F4_t]

[0258] # =========================

[0259] # Step S52: Cross-modal attention fusion computation

[0260] # =========================

[0261] # Importance score of each modality

[0262] s_t = np.array([1.26, 0.94, 1.51, 1.02])

[0263] # Softmax function

[0264] def softmax(x):

[0265] exp_x = np.exp(x)

[0266] return exp_x / np.sum(exp_x)

[0267] # Attention weights α_k^(t)

[0268] alpha_t = softmax(s_t)

[0269] print("Attention weights for each modality:")

[0270] for i, alpha in enumerate(alpha_t):

[0271] print(f"α_{i+1}^(t) = {alpha:.3f}")

[0272] # Calculate the fused feature vector Z^(t)

[0273] Z_t = np.zeros_like(F1_t)

[0274] for alpha, feature in zip(alpha_t, F):

[0275] Z_t += alpha * feature

[0276] print("\nFused feature vector Z^(t):")

[0277] print(Z_t)

[0278] # =========================

[0279] # Step S53: Construction of Infection Risk Assessment Model

[0280] # =========================

[0281] # Behavioral bias eigenvector D^(t)

[0282] D_t = np.array([0.41, 0.36, 0.45])

[0283] # Model parameters

[0284] W1 = np.array([0.63, 0.58, 0.61])

[0285] W2 = np.array([0.49, 0.44, 0.52])

[0286] b = 0.15

[0287] # Sigmoid activation function

[0288] def sigmoid(x):

[0289] return 1 / (1 + np.exp(-x))

[0290] # Infection risk score S^(t)

[0291] risk_input = np.dot(W1, Z_t) + np.dot(W2, D_t) + b

[0292] S_t = sigmoid(risk_input)

[0293] print("\nInfection risk score S^(t):")

[0294] print(round(S_t, 3))

[0295] # =========================

[0296] # Step S54: Individualized Adaptive Weight Adjustment

[0297] # =========================

[0298] # Historical Infection Risk Scoring Sequence

[0299] S_history = np.array([0.61, 0.66, 0.71, 0.75, S_t])

[0300] # User behavior fluctuation level Δ_u

[0301] Delta_u = np.var(S_history)

[0302] print("\nUser behavior fluctuation level Δ_u:")

[0303] print(round(Delta_u, 6))

[0304] # Adjustment coefficient λ

[0305] lambda_value = 1.5

[0306] # Adjusted weight parameters

[0307] W1_prime = W1 * (1 + lambda_value * Delta_u)

[0308] print("\nAdjusted weight parameter W1':")

[0309] print(W1_prime)

[0310] # =========================

[0311] # Step S55: Infection Risk Level Classification

[0312] # =========================

[0313] def risk_level(score):

[0314] if score < 0.4:

[0315] Return "Low Risk"

[0316] ELIF score < 0.7:

[0317] return "medium risk"

[0318] else:

[0319] return "high risk"

[0320] L_t = risk_level(S_t)

[0321] print("\nInfection risk level L^(t):")

[0322] print(L_t)

[0323] # =========================

[0324] # Step S56: Analysis of the Contribution of Infection Risk Characteristics

[0325] # =========================

[0326] # Sigmoid derivative

[0327] sigmoid_derivative = S_t * (1 - S_t)

[0328] # Feature contribution

[0329] C_t = sigmoid_derivative * W1

[0330] print("\nInfection risk characteristic contribution:")

[0331] for i, contribution in enumerate(C_t):

[0332] print(f"C_{i+1}^(t) = {contribution:.3f}")

[0333] # =========================

[0334] # Output the final analysis results

[0335] # =========================

[0336] print("\n=========================")

[0337] print("Results of infection risk analysis")

[0338] print("=========================")

[0339] print(f"Fused feature vector Z^(t): {Z_t}")

[0340] print(f"Infection risk score S^(t): {round(S_t, 3)}")

[0341] print(f"Infection risk level: {L_t}")

[0342] print(f"Degree of behavioral fluctuation Δ_u: {round(Delta_u, 6)}")

[0343] max_index = np.argmax(C_t)

[0344] feature_names = [

[0345] "Hand hygiene behavioral characteristics",

[0346] "Characteristics of Equipment Disinfection Behavior"

[0347] "Patient contact behavior characteristics" ]

[0349] print(f"Main risk factors: {feature_names[max_index]}").

[0350] It should be noted that the terms "first" and "second" used in this document are only used to distinguish different objects, modules or operations, and are not used to limit their importance, order or relationship. At the same time, "including", "containing" and similar expressions should be understood as open inclusion, that is, the relevant process, method, apparatus or system not only includes the technical features explicitly listed, but may also include other technical features that are not explicitly listed but are conventional or inherent to the process, method, apparatus or system.

[0351] Although the present invention has been shown and described in conjunction with specific embodiments, those skilled in the art should understand that various modifications, substitutions, adjustments or equivalent transformations can be made to the embodiments without departing from the core ideas and technical principles of the present invention, and all such changes should fall within the protection scope of the present invention. The protection scope of the present invention should be determined by the claims and their equivalent technical solutions.

[0352] The above description only illustrates the technical solution and implementation of the present invention. This description should not be construed as a limitation of the present invention. The embodiments shown in the accompanying drawings are only one of the embodiments of the present invention and do not constitute a limitation on the actual application structure. For those skilled in the art, any non-creative changes and improvements made to the relevant structural forms, implementation methods or embodiments under the guidance of the present invention without departing from the design concept and technical essence of the present invention should be considered as falling within the protection scope of the present invention.

Claims

1. An intelligent data analysis method for hospital infection control, characterized in that: The method includes the following steps: Step S1: Multi-source infection control behavior data collection. Collect multi-source data during the hospital infection control process, and standardize and time-align the multi-source data to obtain pre-processed data; Step S2: Sensory control behavior feature modeling, extract features from preprocessed data, construct sensory control behavior feature vectors, map different types of sensory control behavior feature vectors to behavior nodes, and form a unified behavior representation space; Step S3: Model the relationship between sensory control behaviors. Based on the behavior nodes, construct the relationship between the feature vectors of sensory control behaviors to form a sensory control behavior relationship graph. Step S4: Based on graph model behavior pattern learning, model and learn the relationship graph of infection control behavior, construct a behavior pattern prediction model based on the time series information of multi-source data in step S1, obtain the predicted behavior features of the corresponding behavior nodes, compare the predicted behavior features with the infection control behavior feature vector, calculate the behavior deviation value, and extract infection control abnormal features based on the behavior deviation value. Step S5: Multimodal fusion infection risk assessment calculation, which integrates infection prevention and control abnormal features with infection control behavior feature vectors to construct an infection risk assessment model and calculate the infection risk score for the corresponding user; Step S6: Dynamic evolution analysis of infection risk. Based on the user's infection risk score, analyze the dynamic trend of infection risk and generate an infection risk level.

2. The intelligent data analysis method for hospital infection control according to claim 1, characterized in that: Step S1 involves multi-source data during hospital infection control, specifically including data on medical staff's operational behavior, medical operation logs, environmental monitoring data, equipment usage records, and patient status data.

3. The intelligent data analysis method for hospital infection control according to claim 1, characterized in that: Step S3 specifically includes the following steps: Step S31: Constructing multidimensional infection control behavior relationships. Based on behavior nodes and corresponding infection control behavior feature vectors, extract multidimensional association features between behavior nodes, including medical operation time interval features, operation intensity change features, and infection control measure execution response features. Based on the multidimensional association features, construct the initial association relationship between behavior nodes. Step S32: Joint Embedding Representation of Behavior Nodes and Relationship Features. A joint mapping process is performed on the behavior nodes and multidimensional association features to obtain the embedded representation of the behavior nodes, expressed in the following form: ； in, Indicates the first Embedded representation of each behavior node Indicates the first Feature vectors of each behavioral node Represents behavior nodes With behavioral nodes The correlation features between them Indicates the embedding of mapping functions; Step S33: Adaptive relation filtering based on association strength. Based on the embedded representation of behavior nodes, calculate the association strength between any two behavior nodes. The association strength is defined as follows: ； in, Represents behavior nodes With behavioral nodes The strength of the correlation between them; Based on the strength of association, the association relationships between behavioral nodes are filtered to obtain the final behavioral association relationships; Step S34: Multi-type behavior relationship hierarchical modeling. Based on the final behavior association, the behavior nodes are used as graph nodes and the final behavior association is used as graph edges to construct the sensory control behavior relationship graph.

4. The intelligent data analysis method for hospital infection control according to claim 1, characterized in that: Step S4 specifically includes the following steps: Step S41: Construction of temporal features of behavior relationship diagram. Based on the temporal series information of sensory control behavior relationship diagram and multi-source data, extract the temporal features of each behavior node within a continuous time window and construct the temporal feature sequence of behavior nodes. Step S42: Behavioral pattern prediction modeling based on the behavioral relationship graph. Based on the structure of the behavioral relationship graph and the temporal feature sequence of the behavioral nodes, a behavioral pattern prediction model is constructed to predict the sensory control behavioral feature vector of each behavioral node at the current moment, and obtain the corresponding predicted behavioral features, which are expressed as follows: ； in, Represents a node At any moment Predictive behavioral characteristics This represents the embedding representation of the node at the previous time step. Represents a node The set of adjacent nodes, Represents the prediction function; Step S43: Calculate the behavior deviation value. Based on the predicted behavior features and the sensory control behavior feature vector, calculate the behavior deviation value for each behavior node. The behavior deviation value is defined as follows: ； in: Represents the feature vector of sensory control behavior. Indicates predictive behavioral characteristics, Indicates the behavioral deviation value; Step S44: Abnormal behavior identification based on deviation distribution, statistical analysis of the behavior deviation value of each behavior node, determination of the abnormal judgment threshold based on the deviation distribution characteristics, marking behavior nodes whose behavior deviation value exceeds the threshold, and obtaining a set of abnormal behavior nodes for infection control. Step S45: Extract abnormal features of infection control. Based on the set of abnormal behavior nodes in infection control, extract abnormal features of infection control.

5. The intelligent data analysis method for hospital infection control according to claim 1, characterized in that: Step S5 specifically includes the following steps: Step S51: Multimodal feature construction. Based on the infection control behavior feature vector and infection prevention and control anomaly features, features from different sources and with different semantics are classified and processed to construct a multimodal feature set, as shown below: ； in, Indicates the first A subset of modal features Represents a set of multimodal features; Step S52: Cross-modal attention fusion calculation, weighted fusion of multimodal feature sets to obtain a unified fusion feature vector, which is defined as follows: ； ； in, Indicates time The fused feature vector, Indicates the first The modality at time... Feature representation, Indicates the first Attention weights for each modality Indicates the first The importance score function value of each modal feature Indicates the number of modal feature subsets. Index variables representing subsets of modal features. Indicates the first The importance score function value of each modality feature; Step S53: Construction of an infection risk assessment model based on behavioral bias enhancement. Based on the fusion of feature vectors and behavioral bias features, an infection risk assessment model is constructed to calculate the user's risk at time [time value missing]. The infection risk score is expressed as follows: ； in, Indicates the user at a certain time. Infection risk score, Represents the fused feature vector. This represents a behavioral deviation feature vector consisting of a set of behavioral deviation values. Represents the model weight parameters. Indicates the bias term. Indicates the activation function; Step S54: Individualized adaptive weight adjustment. Based on the stability of the object's historical infection control behavior, the parameters of the infection risk assessment model are adaptively adjusted. The degree of fluctuation in user behavior is calculated based on multi-source data and infection risk scores, as shown below: ； in, Indicates the degree of fluctuation in user behavior. This represents a sequence of infection risk scores within a time window. This is the variance calculation function; The weights of the infection risk assessment model are adjusted based on the degree of fluctuation in user behavior, using the following adjustment formula: ； in, This represents the adjusted weight parameters. Indicates the adjustment coefficient; Step S55: Infection risk level classification. Based on the infection risk score, interval mapping is performed to obtain the infection risk level. The mapping function is defined as follows: ； in, Indicates the user at a certain time. The level of infection risk, Represents a piecewise mapping function; Step S56: Infection risk feature contribution analysis. Based on the fused feature vector, calculate the contribution of each feature to the infection risk score to obtain the infection risk interpretation results. The feature contribution is defined as: ； in, Indicates the first Each feature at time Contribution to infection risk score Represents the first in the fused feature vector Each feature component.

Images