EMS Disaster Logistics Dataset

This dataset contains a large-scale, spatio-temporal record of emergency medical service (EMS) operations under disaster and high-stress conditions, designed to support research in predictive logistics, disaster healthcare analytics, intelligent transportation systems, and emergency response optimization. The dataset captures the complex interaction between emergency demand, disaster severity, transportation accessibility, healthcare resource availability, and operational delays, reflecting real-world EMS decision environments rather than idealized or balanced conditions. The data are structured as a multivariate time series with a 5-minute temporal resolution, enabling fine-grained analysis of rapid system dynamics that occur during disaster escalation and recovery phases. Temporal and Spatial Coverage Time span: January 1, 2023 – October 31, 2025 [Half Year sample is uploaded] Temporal resolution: 5 minutes Spatial granularity: Regional EMS service zones (24 regions) Geographical attributes: Latitude and longitude provided per region to enable spatial modeling and visualization All records are anonymized and aggregated at the regional level to preserve operational confidentiality while maintaining analytical fidelity. Target Variable expected_response_time_min (minutes):Continuous response time from emergency call initiation to EMS arrival. This variable serves as the primary dependent variable for regression-based forecasting of emergency response performance, particularly under disaster-induced stress. Feature Categories and Descriptions 1. Spatio-Temporal Context These features capture routine temporal patterns and regional context: region_id: Unique identifier for each EMS response region timestamp: ISO-formatted date and time (5-minute alignment) latitude, longitude: Fixed spatial coordinates per region hour_of_day: Hour index (0–23) day_of_week: Day index (0–6) is_weekend: Binary indicator time_since_disaster_onset_min: Minutes elapsed since disaster onset in the region 2. Disaster and Infrastructure Conditions These variables describe disaster intensity and infrastructure degradation: disaster_type: Dominant disaster category affecting the region disaster_severity_score: Continuous severity index (0–1) affected_population_est: Estimated impacted population road_accessibility_ratio: Fraction of usable road infrastructure power_availability_ratio: Power grid availability ratio communication_availability_ratio: Communication network availability 3. Emergency Healthcare Demand These features reflect EMS workload and case severity: total_emergency_calls: Number of emergency calls critical_case_ratio: Proportion of life-threatening cases trauma_case_count: Trauma-related emergencies medical_case_count: Non-trauma medical emergencies pediatric_case_ratio: Pediatric case proportion elderly_case_ratio: Elderly case proportion 4. Logistics and Mobility State These attributes capture transportation and mobility constraints: avg_travel_time_min: Average ambulance travel time traffic_congestion_index: Normalized congestion level road_closure_count: Number of road closures fuel_supply_ratio: Availability of fuel resources vehicle_operational_ratio: Fraction of functional EMS vehicles 5. Medical Resource Availability These variables describe healthcare capacity and constraints: available_ambulances: Ambulances currently available busy_ambulances: Ambulances engaged in service hospital_beds_available: Available hospital beds blood_units_available: Available blood supply units medical_kits_available: Emergency medical kits staff_availability_ratio: Fraction of available medical staff 6. System Load and Stress Indicators Derived indicators capturing operational strain: demand_supply_ratio: Emergency demand relative to ambulance availability ambulance_utilization_rate: Proportion of ambulances in use hospital_capacity_utilization: Fraction of occupied hospital capacity resource_replenishment_eta_min: Estimated resource replenishment delay coordination_delay_min: Inter-agency coordination delay 7. Historical Performance Context Temporal memory features enabling predictive modeling: avg_response_time_prev_window: Previous response time (lag-1) missed_cases_prev_window: Estimated unmet demand allocation_delay_prev_window: Prior dispatch allocation delay rolling_demand_mean: Rolling mean of emergency demand rolling_demand_std: Rolling demand variability Dataset Characteristics Highly imbalanced distributions: Reflecting real EMS operations where extreme delays and disaster conditions are rare but critical Non-stationary behavior: Temporal dynamics change rapidly during disasters Heavy-tailed response times: Capturing worst-case operational stress Chronologically ordered: Suitable for time-series forecasting without information leakage The dataset intentionally preserves these properties rather than artificially smoothing or rebalancing them, ensuring realism for disaster-aware predictive modeling.本数据集包含大规模灾害与高压力场景下的应急医疗服务（Emergency Medical Service, EMS）运行时空记录，旨在为预测物流、灾害医疗分析、智能交通系统与应急响应优化相关研究提供支撑。该数据集捕捉了应急需求、灾害烈度、交通可达性、医疗资源可获得性与运行延迟之间的复杂交互关系，还原了真实的EMS决策场景，而非理想化或均衡化的条件。 该数据集采用多变量时间序列结构，时间分辨率为5分钟，可支持对灾害升级与恢复阶段中快速系统动态的精细化分析。 ## 时空覆盖范围 时间跨度：2023年1月1日—2025年10月31日 [已上传半年样本数据] 时间分辨率：5分钟 空间粒度：EMS服务区域（共24个区域） 地理属性：每个区域均提供经纬度坐标，可用于空间建模与可视化 所有记录均已完成匿名化处理，并按区域级别聚合，在保障运行保密性的同时保留了分析保真度。 ## 目标变量 **预期响应时长（expected_response_time_min，单位：分钟）**：从应急呼叫发起至EMS车辆抵达的连续时长。该变量是基于回归的应急响应性能预测的核心因变量，尤其适用于灾害诱导的高压力场景下的预测。 ## 特征类别与说明 ### 1. 时空上下文特征 此类特征用于捕捉常规时间模式与区域上下文信息： - 区域ID（region_id）：每个EMS响应区域的唯一标识符 - 时间戳（timestamp）：ISO格式的日期与时间（已对齐至5分钟间隔） - 纬度（latitude）、经度（longitude）：每个区域的固定空间坐标 - 当日小时（hour_of_day）：小时索引（0~23） - 当日星期数（day_of_week）：星期索引（0~6） - 是否为周末（is_weekend）：二元指示变量 - 灾害发生后时长（time_since_disaster_onset_min）：该区域自灾害发生以来经过的分钟数 ### 2. 灾害与基础设施状况 此类变量用于描述灾害烈度与基础设施损毁情况： - 灾害类型（disaster_type）：影响该区域的主要灾害类别 - 灾害烈度评分（disaster_severity_score）：连续型烈度指数（0~1） - 受影响人口估算值（affected_population_est）：受影响人口的估算值 - 道路可达性比率（road_accessibility_ratio）：可用道路基础设施占比 - 电力可用性比率（power_availability_ratio）：电网可用率 - 通信可用性比率（communication_availability_ratio）：通信网络可用率 ### 3. 应急医疗需求 此类特征用于反映EMS工作负荷与病例严重程度： - 应急呼叫总量（total_emergency_calls）：应急呼叫总次数 - 危重病例占比（critical_case_ratio）：危重症病例的占比 - 创伤病例数（trauma_case_count）：创伤类应急病例数 - 内科病例数（medical_case_count）：非创伤性医疗应急病例数 - 儿科病例占比（pediatric_case_ratio）：儿科病例占比 - 老年病例占比（elderly_case_ratio）：老年病例占比 ### 4. 物流与运行状态 此类属性用于捕捉交通与移动性约束： - 平均出行时长（avg_travel_time_min，单位：分钟）：救护车平均出行时长 - 交通拥堵指数（traffic_congestion_index）：归一化后的拥堵水平 - 道路封闭数量（road_closure_count）：道路封闭处数 - 燃油供应比率（fuel_supply_ratio）：燃油资源可获得率 - 车辆运行比率（vehicle_operational_ratio）：可用EMS车辆占比 ### 5. 医疗资源可获得性 此类变量用于描述医疗服务能力与约束条件： - 可用救护车数（available_ambulances）：当前可用的救护车数量 - 繁忙救护车数（busy_ambulances）：正在执行任务的救护车数量 - 可用医院床位（hospital_beds_available）：医院可用床位数量 - 可用血液单位数（blood_units_available）：可用血液供给单位量 - 应急医疗包数量（medical_kits_available）：应急医疗包储备量 - 医护人员可用比率（staff_availability_ratio）：可用医护人员占比 ### 6. 系统负荷与压力指标 此类为衍生指标，用于捕捉运行压力： - 供需比率（demand_supply_ratio）：应急需求与救护车可用量的比值 - 救护车使用率（ambulance_utilization_rate）：正在使用的救护车占比 - 医院床位利用率（hospital_capacity_utilization）：医院已占用床位占比 - 资源补给预计时长（resource_replenishment_eta_min，单位：分钟）：资源补给的预估延迟时长 - 跨机构协调延迟（coordination_delay_min，单位：分钟）：跨部门协调的延迟时长 ### 7. 历史性能上下文特征 此类为时间记忆特征，可用于预测建模： - 前一窗口平均响应时长（avg_response_time_prev_window）：前一时刻的响应时长（滞后1阶） - 前一窗口未满足病例数（missed_cases_prev_window）：预估的未满足需求数 - 前一窗口调度分配延迟（allocation_delay_prev_window）：此前的调度分配延迟时长 - 需求滚动均值（rolling_demand_mean）：应急需求的滚动平均值 - 需求滚动标准差（rolling_demand_std）：应急需求的滚动变异性 ## 数据集特性 - 分布高度不平衡：还原了真实的EMS运行场景——极端延迟与灾害场景虽罕见但至关重要 - 非平稳特性：灾害期间系统的时间动态变化迅速 - 响应时长服从厚尾分布：可捕捉极端运行压力场景 - 按时间顺序排序：可直接用于时间序列预测，无信息泄露风险 本数据集有意保留了上述特性，未进行人为平滑或重平衡处理，以保障灾害感知预测建模的真实性。