Pre-hospital emergency resource scheduling optimization method and system

Abstract

The invention discloses a pre-hospital emergency resource scheduling optimization method and system, and particularly relates to the technical field of resource scheduling. The method comprises the following steps: receiving a multi-source emergency call request, extracting key basic information, and realizing disease condition scoring of each request by using an emergency evaluation model fusing historical cases, emergency rules and a machine learning algorithm; then, a multi-objective optimization model including response time, scheduling cost, resource balance and illness state matching degree is constructed, the weight is dynamically adjusted, and medical oriented optimization of a scheduling decision is realized; and then an optimal dispatching scheme is generated by means of a dynamic weighted intelligent optimization algorithm, rolling updating is carried out in combination with real-time traffic and vehicle states, it is ensured that the dispatching scheme is continuous and effective, and the response priority of a dispatching system to high-risk patients and the overall operation efficiency are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of resource scheduling technology, specifically to a method and system for optimizing pre-hospital emergency medical resource scheduling. Background Technology

[0002] Pre-hospital emergency resource scheduling optimization refers to the scientific and efficient allocation and scheduling of emergency resources (such as ambulances, emergency personnel, and equipment) during the emergency medical phase before patients arrive at the hospital. This aims to shorten response time, improve treatment efficiency, and maximize the use of limited resources. Such optimization typically combines geographic information systems, real-time data analysis, and intelligent algorithms to achieve rapid, accurate, and optimal rescue decisions in emergency situations, ensuring patient safety and improving the overall operational efficiency of the emergency medical system.

[0003] The existing technology has the following shortcomings: In multi-objective optimization scheduling systems, static weight settings struggle to dynamically perceive the urgency of a patient's condition. This can lead to scheduling schemes that, while mathematically optimal, may deviate from actual medical needs. For instance, when faced with simultaneous calls from patients with minor injuries and those with severe illnesses, the system might prioritize response time and mistakenly dispatch an ambulance to the less seriously ill, delaying crucial treatment. Such problems are highly insidious and difficult to detect through routine testing. In severe cases, they directly impact patient survival rates, exposing a critical weakness in current systems: a lack of clinically intelligent judgment capabilities. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing the scheduling of pre-hospital emergency medical resources, in order to address the shortcomings in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing pre-hospital emergency resource scheduling, comprising: It receives multiple pre-hospital emergency call requests and extracts basic information for each call request, including the call location, call time, and preliminary description of the patient's condition. An integrated emergency assessment model is used to score the urgency of each call request. The model is based on historical cases, emergency rules, natural language processing, and machine learning algorithms, and outputs a score that reflects the clinical severity of the patient. A multi-objective optimization scheduling model is constructed, which includes four optimization objectives: response time, scheduling cost, resource balance, and matching degree of disease urgency. The weight parameters of each optimization objective are dynamically adjusted according to the disease urgency score. Based on a dynamic weighted multi-objective optimization algorithm, the optimal scheduling scheme is solved, and the dispatch paths and task assignment results of each emergency rescue resource are generated. Output the scheduling results and update and adjust them dynamically based on real-time traffic and ambulance status information.

[0006] Preferably, a random forest model is used to output the urgency score. The specific steps are as follows: Step 1: Receive the preliminary description of the patient's condition from the call request, construct a feature dictionary, extract several features as model input, vectorize the features, and unify them into a fixed-length input feature array as input to the random forest model; Step 2: Based on historical emergency call data, manually label or use hospital emergency triage results as urgency level labels, and form a structured training sample set; Step 3: Train the random forest model using the training dataset; Step 4: Input the feature vector of each real-time call request into the trained random forest model; the model outputs the voting results of multiple decision trees, the system calculates the probability distribution of urgency based on the voting distribution, and outputs a comprehensive score: the voting results are mapped to urgency score values.

[0007] Preferably, the obtained urgency score is compared with a gradient standard threshold, which includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. The urgency score is then compared with the first standard threshold and the second standard threshold respectively. If the urgency score is greater than the second standard threshold, it is judged as a high urgency level, indicating that the patient's condition is serious and that emergency resources with high response capabilities should be dispatched first and responded in the shortest possible time. If the urgency score is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is judged as medium urgency, indicating that the condition has a certain risk. Response should be made within the limits of available resources, and resources should be allocated appropriately. If the urgency score is less than the first standard threshold, it is judged as a low urgency level, indicating that the condition is relatively mild and the response can be delayed or non-emergency medical resources can be arranged for treatment.

[0008] Preferably, when the dispatch center receives different emergency call requests simultaneously, it first extracts keywords from the description of the illness through the speech recognition and natural language processing module, and combines them with patient background information, location features and historical case tags. Then, it calls the trained random forest model to score the urgency and starts the multi-objective dispatch optimization model. The four initial optimization objectives are: minimize response time; minimize dispatch cost; maximize resource utilization balance; maximize the matching degree of illness urgency; and dynamically adjust the weights of the four objectives based on the scoring results.

[0009] Preferably, the following are included: minimizing response time: the shortest time required for each ambulance to reach the assigned call point from its current location; minimizing dispatch costs: including minimizing vehicle idling distance, fuel consumption, and manpower costs; maximizing resource utilization balance: including calculating the maximum value of the balance score; and maximizing the matching degree of the urgency of the illness: referring to the urgency score value that is greater than the second standard threshold.

[0010] Preferably, the balance score is obtained by setting the scheduling task set as follows: m represents the total number of scheduled tasks, and each task represents an emergency call request; the resource set is set as follows: Where n is the total number of resources, and each resource represents an ambulance or emergency medical station; each ant simulates a matching path from a task to a resource, i.e., a complete scheduling scheme; for each task-resource pair Initialize pheromone concentration Set as a constant value Simultaneously initialize the heuristic function. The heuristic function, taking into account both response time and current resource load, is defined as follows: In the formula, For resources To the mission The expected response time For resources The number of tasks currently undertaken The maximum number of tasks among all resources is ϵ, which is a minimal constant. The path selection probability is calculated using both pheromone and heuristic function. After one iteration, the ant colony forms a scheduling scheme and calculates the task load of each resource. Calculate the balance score B using the following expression: In the formula, The standard deviation of the workload for each resource task. This represents the average task load.

[0011] Preferably, if there are patients with a score higher than 0.9 in the current scheduled events, the weight of their matching degree and response time targets will be automatically increased.

[0012] The present invention also provides a pre-hospital emergency resource scheduling optimization system, including a data acquisition module, a condition assessment and scoring module, a weight configuration module, a scheduling optimization module, and a dynamic adjustment module; Data acquisition module: Receives multiple pre-hospital emergency call requests and extracts basic information for each call request, including call location, call time, and preliminary description of the patient's condition; Condition assessment and scoring module: Through an integrated condition urgency assessment model, the urgency of each call request is scored. The model is based on historical cases, emergency rules, natural language processing and machine learning algorithms, and outputs a score that reflects the clinical severity of the patient. Weight configuration module: Constructs a multi-objective optimization scheduling model, which includes four optimization objectives: response time, scheduling cost, resource balance and matching degree of urgency of illness, and dynamically adjusts the weight parameters of each optimization objective according to the urgency score of illness; Scheduling optimization module: Based on a dynamic weighted multi-objective optimization algorithm, it solves the optimal scheduling scheme and generates dispatch paths and task assignment results for each emergency rescue resource; Dynamic adjustment module: Outputs scheduling results and performs rolling updates and dynamic adjustments based on real-time traffic and ambulance status information.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention introduces a medical urgency assessment model that integrates natural language processing and machine learning to dynamically identify the actual medical urgency of each emergency request. This breaks through the limitations of existing dispatch systems, which suffer from rigid weight settings and neglect of differences in patient conditions, significantly improving the adaptability and scientific rigor of dispatch decisions in clinical scenarios. Through a multi-objective optimization model driven by urgency scores, the system can flexibly adjust the trade-offs between response time, cost, resource balance, and patient matching, achieving a truly intelligent dispatch strategy centered on "patient priority."

[0014] 2. This invention introduces an improved ant colony optimization algorithm, effectively integrating resource load balancing indicators during the scheduling path search process to avoid excessive resource concentration and improve the overall sustainability and robustness of the system. Combined with a rolling update mechanism for real-time traffic data and ambulance status, the scheduling results remain dynamically optimal under complex environments such as emergencies and peak hours, ensuring priority response for high-risk patients while achieving comprehensive scheduling optimization under multiple points, multiple resources, and multiple constraints. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0016] Figure 1 This is a mind map of the method of the present invention.

[0017] Figure 2 This is a system module diagram of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 As shown in this embodiment, a pre-hospital emergency resource scheduling optimization method includes: It receives multiple pre-hospital emergency call requests and extracts basic information for each call request, including the call location, call time, and preliminary description of the patient's condition. An integrated emergency assessment model is used to score the urgency of each call request. The model is based on historical cases, emergency rules, natural language processing, and machine learning algorithms, and outputs a score that reflects the clinical severity of the patient. A multi-objective optimization scheduling model is constructed, which includes four optimization objectives: response time, scheduling cost, resource balance, and matching degree of disease urgency. The weight parameters of each optimization objective are dynamically adjusted according to the disease urgency score. Based on a dynamic weighted multi-objective optimization algorithm, the optimal scheduling scheme is solved, and the dispatch paths and task assignment results of each emergency rescue resource are generated. Output the scheduling results and update and adjust them dynamically based on real-time traffic and ambulance status information.

[0020] This includes receiving multiple pre-hospital emergency care calls and extracting basic information from each call, specifically including the following: Call request reception methods: The system connects to multiple pre-hospital emergency call sources, commonly including: 120 hotline telephone system; emergency apps, WeChat mini programs, urban emergency information platforms; voice recognition and automatic alarm terminals (such as AI voice triage systems); wearable devices or smart terminal automatic alarm interfaces (such as smartwatch fall alarms, vehicle collision detection automatic calls); the system connects these call information to the dispatch platform in real time based on a unified API or protocol.

[0021] Call Basic Information Extraction: Upon receiving a call request, the system extracts structured data from each request record, specifically including the following fields: Call Location: Telephone users input their address via voice or keypad; smart devices automatically obtain GPS location; mobile apps access location permissions and upload coordinates; accuracy standard: prioritizes parsing to latitude and longitude coordinates, and combines with the GIS system to convert to a standard address; if the address is unclear, the system prompts the dispatcher for manual confirmation or a callback. Call Time: The system automatically records the absolute timestamp of the call (accurate to milliseconds); it also records the time difference with the previous request (used for subsequent conflict detection and resource early warning); if the calling device has a time synchronization mechanism (such as NTP or GPS time synchronization), the high-precision time source is used first.

[0022] Preliminary description of the patient's condition: A brief description of the patient's condition collected by the police officer via voice, text, or structured options; if it is a voice call, the system uses Automatic Speech Recognition (ASR) technology to transcribe the call content into text in real time; Preliminary illness types include, but are not limited to, the following: The system handles various ailments including difficulty breathing, cardiac arrest, external bleeding, coma, poisoning, burns, fractures, and childbirth. It utilizes a Natural Language Processing (NLP) model to extract and standardize key information, including: symptom keyword extraction (e.g., "convulsions," "chest tightness," "fall"); initial assessment of acuteness (e.g., "immediate," "severe," "mild"); and intent recognition (e.g., "Does an ambulance need?" "Are there multiple injured persons?"). If information is incomplete, the system automatically prompts the dispatcher to supplement specific fields. The extracted call request data is then encapsulated into a standard data structure, such as a JSON or XML object.

[0023] If there are missing information in the call information (such as inability to locate or missing description of the illness), the system will automatically enter the manual confirmation process; if there are multiple duplicate or similar requests (such as multiple people calling for the same event), the system will use a similarity matching algorithm to aggregate the events and avoid duplicate dispatching; all call requests have a unique identifier (UUID) to facilitate subsequent dispatch process tracking and backtracking.

[0024] An integrated emergency assessment model is used to score the urgency of each call request. This model uses the following data as input features: Preliminary description of symptoms: Subjective statements entered by users or dispatchers, such as "chest pain for 3 hours", "elderly person falls and falls unconscious", "child has persistent high fever", etc. If it is voice input, it is first transcribed into text through automatic speech recognition (ASR); then the natural language processing (NLP) module performs word extraction and syntactic analysis to identify keywords, symptom categories, onset time and other content.

[0025] Patient background information (if any): Age (children / elderly preferred); Past medical history (e.g., heart disease, stroke, diabetes); Special populations (pregnant women, infants, post-operative patients); Call time and geographical characteristics: whether it is late at night or on a holiday (when medical resources are less available); whether it is located in a remote or inaccessible area; the number and distance of nearby medical institutions, etc.

[0026] Based on the clinical severity labeling results corresponding to similar disease descriptions in the historical emergency data of this region, a training sample set was established to provide a data foundation for subsequent supervised learning models.

[0027] Construct a rule base for mapping disease keywords to urgency levels, for example: "Respiratory arrest" "Unconscious" → Emergency level: Extremely high; "Chest pain" and "Numbness in the left upper limb" → Emergency level: High; "Dizziness" or "minor abrasions" → Emergency level: medium or low.

[0028] The rules are referenced from sources including: WHO pre-hospital emergency assessment guidelines; domestic emergency procedure standards; and three-tier emergency triage systems (such as ESI and CTAS).

[0029] The text description is encoded using BERT, Word2Vec, or a medically-specific pre-trained model (such as BioBERT); extracted features include: symptom word frequency, sentiment polarity; verb descriptions (such as "convulsions", "vomiting", "coma"); and combined features (such as "elderly + fall + headache"); the model can be random forest, XGBoost, support vector machine (SVM), or a neural network model; the output is an urgency score, typically ranging from 0 to 1, with higher values ​​indicating greater urgency; supervised learning is performed using local or external emergency medical case data, labeled with historical triage decisions by professional doctors or dispatchers.

[0030] For example, this invention uses a random forest model to output an urgency score, and the specific steps are as follows: Step 1: Text preprocessing and feature extraction: 1.1 Receive the initial description of the patient's condition in the call request (e.g., "The elderly person suddenly collapsed, lost consciousness, and had convulsions"). 1.2 Use Chinese word segmentation tools (such as jieba) or medical text preprocessing models to segment and clean the text, remove stop words, and retain medical-related keywords; 1.3 By constructing a feature dictionary, the following types of features are extracted as model input: Symptom keyword vectors (such as convulsions, loss of consciousness, chest tightness, etc.); Patient attributes: age, gender, medical history; Description of emotion or severity: such as "serious", "persistent", "sudden", etc.; Location and time factors (such as late at night, high temperature, holidays); 1.4 Vectorize the above features into a fixed-length input feature array so that it can be used as input to the random forest model.

[0031] Step 2: Construct the training dataset: 2.1 Based on historical emergency call data, urgency levels are tagged manually or using hospital emergency triage results (such as whether the patient is critically ill, whether they were admitted to the ICU, etc.). The tag scores are generally as follows: 0: Mild (e.g., superficial injury, slight dizziness); 1: Moderate (e.g., high fever, non-coma-related trauma); 2: High level of emergency (e.g., difficulty breathing, loss of consciousness); 3: Extreme emergency (such as sudden cardiac arrest or respiratory arrest).

[0032] 2.2 Form a structured training sample set.

[0033] Step 3: Model Training and Tuning 3.1 Train the random forest model using the training dataset. Key parameters include: Number of decision trees: e.g., 100-500; Maximum depth: Set according to feature complexity; Minimum number of splits in a sample, minimum number of samples in a leaf node; Use the Gini coefficient or information gain ratio as the criterion for node splitting; 3.2 Use cross-validation methods (such as 5-fold cross-validation) to evaluate the model's stability and accuracy; 3.3 Improve model performance by optimizing hyperparameter combinations through grid search or random search.

[0034] Step 4: Model Prediction and Scoring Output: 4.1 Input the feature vector of each real-time call request into the trained random forest model; 4.2 The model outputs voting results from multiple decision trees. The system calculates the probability distribution of urgency based on the voting distribution and outputs a comprehensive score: 4.3 Map the voting results to an urgency score, for example: UrgencyScore = (Number of votes for Level 1 × 0.3 + Number of votes for Level 2 × 0.6 + Number of votes for Level 3 × 1.0) / Total number of votes; If the result is UrgencyScore = 0.86, it indicates that the condition corresponding to the call belongs to the high-risk level.

[0035] The model outputs an urgency score, typically between 0 and 1, with higher scores indicating more urgent conditions. This score can be further mapped to high, medium, and low urgency level labels, providing a reference for the dispatch system. Specifically: The obtained urgency score is compared with the gradient standard threshold, which includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. The urgency score is then compared with the first standard threshold and the second standard threshold respectively. If the urgency score is greater than the second standard threshold, it is judged as a high urgency level, indicating that the patient's condition is serious and that emergency resources with high response capabilities should be dispatched first and responded in the shortest possible time. If the urgency score is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is judged as medium urgency level, indicating that the condition has a certain risk, but is not life-threatening at the moment, and can be responded to and resources allocated appropriately if resources allow; If the urgency score is less than the first standard threshold, it is judged as a low urgency level, indicating that the condition is relatively mild and the response can be delayed, or non-emergency medical resources can be arranged for treatment.

[0036] In this embodiment, when the dispatch center receives three emergency call requests simultaneously, they are as follows: Request A: Location is in the city center business district, described as "a young person fell, suffered a head abrasion, and is conscious"; Request B: From the suburbs, described as "an elderly person suddenly has difficulty breathing and cannot speak"; Request C: Location is near an industrial park, described as "Someone has collapsed at the factory gate and is unresponsive".

[0037] The system first extracts keywords from the description of the illness using speech recognition and natural language processing modules. Then, it combines this with patient background information (such as gender and age), location features, and historical case tags to call a pre-trained random forest model to score the urgency level. For example, the scoring results are as follows: Request A urgency score: 0.32 (low); Request B urgency score: 0.81 (high); Request C urgency score: 0.93 (extremely high); Subsequently, the system initiates a multi-objective scheduling optimization model, with the following four initial optimization objectives: Minimize response time; Minimize scheduling cost; Maximize the balance of resource utilization; Maximize the matching degree of the urgency of the illness.

[0038] The system dynamically adjusts the weights of the four objectives based on the scoring results. Because request C received an extremely high score, the system increased the weights of response time and patient condition matching to 40% and 35%, respectively, while relatively reducing the weights of cost and resource balance to 15% and 10%. Therefore, the scheduling algorithm prioritizes dispatching the nearest high-level emergency vehicle to respond to request C, even if that vehicle comes from a busy area, to ensure rapid treatment for critically ill patients.

[0039] Request B was assigned to the next nearest available ambulance, accompanied by an intermediate-level emergency medical technician. Request A, due to the minor nature of the illness, was arranged to be picked up by a vehicle that was about to return, with a slight delay in dispatch time, but within an acceptable range.

[0040] The final system output scheduling scheme takes into account both the urgency of medical care and the rationality of system resource allocation. In actual simulation scenarios, it was verified that the response time was reduced by an average of 12%, and the average arrival time of high-risk patients before treatment was reduced by about 4 minutes, effectively improving the scheduling effect.

[0041] After completing the urgency score and constructing a multi-objective optimization model, this invention further employs a dynamically weighted multi-objective optimization algorithm to comprehensively consider the importance of each scheduling objective and solve for the optimal pre-hospital emergency resource scheduling scheme.

[0042] In this embodiment, for example, an improved weighted multi-objective genetic algorithm is used to fuse the following four objective functions into a comprehensive evaluation index: Minimize response time: that is, the shortest time required for each ambulance to travel from its current location to its assigned call point; Minimize dispatch costs: including minimizing vehicle idle distance fuel consumption and labor costs; Maximizing the balance of resource utilization: This includes calculating the maximum value of the balance score; Maximizing the matching degree of urgency of the condition: refers to the urgency score that is greater than the second standard threshold.

[0043] The balance score is obtained by setting the scheduling task set as follows: m represents the total number of scheduled tasks, and each task represents an emergency call request; the resource set is set as follows: , where n is the total number of resources; each ant simulates a matching path from task to resource, that is, a complete scheduling scheme.

[0044] For each pair of task resources Initialize pheromone concentration , is usually set to a constant value ; Simultaneously initialize the heuristic function. This indicates the attractiveness of the resources to which the current task is assigned.

[0045] The heuristic function, taking into account both response time and current resource load, is defined as follows: In the formula, For resources To the mission The expected response time For resources The number of tasks currently undertaken ϵ represents the maximum number of tasks currently available across all resources; ϵ is a very small constant to prevent division by zero errors.

[0046] The path selection probability is determined by both the pheromone and the heuristic function: ; The task will be assigned to the kth ant. Allocate resources The probability; α, β are weight parameters that adjust the influence of pheromones and heuristic functions, and Available is the set of currently allocable resources.

[0047] After the ant colony completes one iteration, it formulates a scheduling scheme and calculates the task load of each resource. Calculate the balance score B using the following expression: In the formula, The standard deviation of the workload for each resource task. This represents the average task load.

[0048] The weights of the objective function described above are dynamically driven by the urgency score of the patient's condition. For example, if there are patients with a score higher than 0.9 in the current scheduling events, the system automatically increases the weights of the matching degree and response time objectives to ensure priority response to critically ill patients.

[0049] During the optimization algorithm execution, the system constructs a schedulable combination of all available ambulances and currently pending emergency call requests. Through iterative algorithmic search, different assignment methods and path allocations are continuously tried, and each solution is comprehensively scored across the four dimensions mentioned above.

[0050] The output of the optimization algorithm includes: The task assignment result for each ambulance, i.e., the call number it needs to respond to; The corresponding travel route (the shortest reachable route calculated based on real-time traffic data); Estimated arrival time; Priority description of dispatched vehicles and tasks; The overall optimization score of the task scheduling scheme (used for system self-assessment and manual review).

[0051] After generating the optimal solution, the system automatically pushes the dispatch plan to the emergency vehicle terminal and the dispatcher platform, and allows for manual intervention to adjust the plan in special circumstances.

[0052] After solving the multi-objective optimization algorithm based on dynamic weighting, the system outputs the generated optimal scheduling scheme as the initial instruction and distributes it in real time to relevant emergency medical resource terminals through the scheduling platform. The scheduling results include the following: The target mission number and patient location for each ambulance; route planning (generated based on the latest traffic conditions); estimated arrival time and mission priority; personnel configuration suggestions for the current mission (e.g., whether an advanced life support team is needed); and an overall optimization score for the dispatch results (for back-end management analysis or manual intervention reference).

[0053] To ensure the effectiveness of the scheduling scheme in the complex and dynamic urban traffic environment, this invention introduces a rolling update and dynamic adjustment mechanism to adapt to the ever-changing external environment and resource status.

[0054] The system connects to urban traffic information platforms or high-precision map services to obtain the following data in real time: Road conditions (whether congested, closed, under construction, etc.); real-time speed estimates; temporary emergencies (such as traffic accidents, road closures); based on real-time traffic changes, the system continuously updates the estimated travel routes and arrival times for each task and re-evaluates the scheduling effectiveness.

[0055] The system continuously collects status information of each ambulance through vehicle positioning equipment and emergency personnel's operating terminals, including: current geographical location (GPS positioning); whether it is idle or performing a mission; traffic flow on the road segment; whether there are any abnormal situations such as mechanical failure, delay, or inability to reach the destination; the system is set up with a status trigger mechanism, once a critical change occurs in the status of an ambulance (such as deviating from the route, traffic jam exceeding the set time threshold, mission failure, etc.), the system immediately activates the mission reassignment mechanism.

[0056] The scheduling system is based on a rolling window mechanism, which re-executes a local optimization calculation every set time period (e.g., 30 seconds or 1 minute) or after receiving a state change event. The characteristics of this mechanism are: it does not reassign completed or immutable tasks to avoid system instability; it only optimizes currently executing or upcoming tasks; and it prioritizes the stability of tasks for high-risk patients, while allowing for appropriate delays or replacement of vehicle resources for low-priority tasks.

[0057] For example, if an ambulance scheduled to go to a high-risk patient encounters severe traffic congestion en route, extending its estimated arrival time from 8 minutes to 18 minutes, the system will identify this abnormal situation in real time, quickly assess nearby vehicle resources, and reallocate an idle vehicle with better responsiveness to the patient's location, while transferring the original vehicle's task to another low-to-medium priority request.

[0058] Example 2, please refer to Figure 2 As shown in the figure, the pre-hospital emergency resource scheduling optimization system described in this embodiment includes a data acquisition module, a disease assessment and scoring module, a weight configuration module, a scheduling optimization module, and a dynamic adjustment module; Data acquisition module: Receives multiple pre-hospital emergency call requests and extracts basic information for each call request, including call location, call time, and preliminary description of the patient's condition; Condition assessment and scoring module: Through an integrated condition urgency assessment model, the urgency of each call request is scored. The model is based on historical cases, emergency rules, natural language processing and machine learning algorithms, and outputs a score that reflects the clinical severity of the patient. Weight configuration module: Constructs a multi-objective optimization scheduling model, which includes four optimization objectives: response time, scheduling cost, resource balance and matching degree of urgency of illness, and dynamically adjusts the weight parameters of each optimization objective according to the urgency score of illness; Scheduling optimization module: Based on a dynamic weighted multi-objective optimization algorithm, it solves the optimal scheduling scheme and generates dispatch paths and task assignment results for each emergency rescue resource; Dynamic adjustment module: Outputs scheduling results and performs rolling updates and dynamic adjustments based on real-time traffic and ambulance status information.

[0059] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0060] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0061] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0062] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for optimizing the scheduling of pre-hospital emergency medical resources, characterized in that: include: Receive multiple pre-hospital emergency call requests and extract basic information for each call request, including call location, call time, and preliminary description of the patient's condition; An integrated emergency assessment model is used to score the urgency of each call request. The model is based on historical cases, emergency rules, natural language processing, and machine learning algorithms, and outputs a score that reflects the clinical severity of the patient. A multi-objective optimization scheduling model is constructed, which includes four optimization objectives: response time, scheduling cost, resource balance, and matching degree of disease urgency. The weight parameters of each optimization objective are dynamically adjusted according to the disease urgency score. Based on a dynamic weighted multi-objective optimization algorithm, the optimal scheduling scheme is solved, and the dispatch paths and task assignment results of each emergency rescue resource are generated. Output the scheduling results and update and adjust them dynamically based on real-time traffic and ambulance status information.

2. The pre-hospital emergency resource scheduling optimization method according to claim 1, characterized in that: The urgency score is output using a random forest model. The specific steps are as follows: Step 1: Receive the preliminary description of the patient's condition from the call request, construct a feature dictionary, extract several features as model input, vectorize the features, and unify them into a fixed-length input feature array as input to the random forest model; Step 2: Based on historical emergency call data, manually label or use hospital emergency triage results as urgency level labels, and form a structured training sample set; Step 3: Train the random forest model using the training dataset; Step 4: Input the feature vector of each real-time call request into the trained random forest model; the model outputs the voting results of multiple decision trees, the system calculates the probability distribution of urgency based on the voting distribution, and outputs a comprehensive score: the voting results are mapped to urgency score values.

3. The pre-hospital emergency resource scheduling optimization method according to claim 2, characterized in that: The obtained urgency score is compared with the gradient standard threshold, which includes a first standard threshold and a second standard threshold, and the first standard threshold is less than the second standard threshold. The urgency score is then compared with the first standard threshold and the second standard threshold respectively. If the urgency score is greater than the second standard threshold, it is judged as a high urgency level, indicating that the patient's condition is serious and that emergency resources with high response capabilities should be dispatched first and responded in the shortest possible time. If the urgency score is greater than or equal to the first standard threshold and less than or equal to the second standard threshold, it is judged as medium urgency, indicating that the condition has a certain risk. Response should be made within the limits of available resources, and resources should be allocated appropriately. If the urgency score is less than the first standard threshold, it is judged as a low urgency level, indicating that the condition is relatively mild and the response can be delayed or non-emergency medical resources can be arranged for treatment.

4. The pre-hospital emergency resource scheduling optimization method according to claim 3, characterized in that: When the dispatch center receives different emergency call requests simultaneously, it first extracts keywords from the description of the illness through the speech recognition and natural language processing module. Then, it combines the patient's background information, location features, and historical case tags, calls the trained random forest model to score the urgency, and starts the multi-objective dispatch optimization model. The four initial optimization objectives are: minimize response time; minimize dispatch cost; maximize resource utilization balance; maximize the matching degree of illness urgency; and dynamically adjust the weights of the four objectives based on the scoring results.

5. The pre-hospital emergency resource scheduling optimization method according to claim 4, characterized in that: in, Minimize response time: that is, the shortest time required for each ambulance to travel from its current location to its assigned call point; Minimize dispatch costs: including the minimum values ​​of vehicle idle distance fuel consumption and labor costs; maximize resource utilization balance: including the maximum value of the balance score; maximize the matching degree of urgency of illness: refers to the urgency score value that is greater than the second standard threshold.

6. The pre-hospital emergency resource scheduling optimization method according to claim 5, characterized in that: The balance score is obtained by setting the scheduling task set as follows: m represents the total number of scheduled tasks, and each task represents an emergency call request; the resource set is set as follows: Where n is the total number of resources, and each resource represents an ambulance or emergency medical station; each ant simulates a matching path from a task to a resource, i.e., a complete scheduling scheme; for each task-resource pair Initialize pheromone concentration Set as a constant value Simultaneously initialize the heuristic function. The heuristic function, taking into account both response time and current resource load, is defined as follows: In the formula, For resources To the mission The expected response time For resources The number of tasks currently undertaken, The maximum number of tasks among all resources is ϵ, which is a minimal constant. The path selection probability is calculated using both pheromone and heuristic function. After one iteration, the ant colony forms a scheduling scheme and calculates the task load of each resource. Calculate the balance score B using the following expression: In the formula, The standard deviation of the workload for each resource task. This represents the average task load.

7. The pre-hospital emergency resource scheduling optimization method according to claim 6, characterized in that: If there are patients with a score higher than 0.9 in the current pending events, the weight of their matching degree and response time targets will be automatically increased.

8. A pre-hospital emergency medical resource scheduling optimization system, used to implement the pre-hospital emergency medical resource scheduling optimization method according to any one of claims 1-7, characterized in that: It includes a data acquisition module, a disease assessment and scoring module, a weight configuration module, a scheduling optimization module, and a dynamic adjustment module; Data acquisition module: Receives multiple pre-hospital emergency call requests and extracts basic information for each call request, including call location, call time, and preliminary description of the patient's condition; Condition assessment and scoring module: Through an integrated condition urgency assessment model, the urgency of each call request is scored. The model is based on historical cases, emergency rules, natural language processing and machine learning algorithms, and outputs a score that reflects the clinical severity of the patient. Weight configuration module: Constructs a multi-objective optimization scheduling model, which includes four optimization objectives: response time, scheduling cost, resource balance and matching degree of urgency of illness, and dynamically adjusts the weight parameters of each optimization objective according to the urgency score of illness; Scheduling optimization module: Based on a dynamic weighted multi-objective optimization algorithm, it solves the optimal scheduling scheme and generates dispatch paths and task assignment results for each emergency rescue resource; Dynamic adjustment module: Outputs scheduling results and performs rolling updates and dynamic adjustments based on real-time traffic and ambulance status information.

Images