A method and system for generating time series data of emergency events that integrates long-term and short-term time patterns
By constructing background intensity and self-excitation process models, emergency event time series are generated, solving the problem that existing technologies cannot simultaneously account for long-term and short-term time patterns, and achieving more realistic and stable emergency event simulation.
Patent Information
- Application Number
- CN202611201763.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-07
- Publication Date
- 2026-11-10
AI Technical Summary
Existing technologies cannot simultaneously characterize the long-term periodicity and short-term clustering effects of emergency events, resulting in inconsistencies between the generated event sequences and the actual timing patterns of emergency alarms.
By constructing a background intensity model and a self-excitation process model, and combining them with historical emergency event data, a comprehensive conditional intensity function is generated to produce emergency event time series.
The generated event sequences not only reflect the statistical consistency of long-term time distribution, but also present continuous events and local bursts, thus improving the realism and stability of the simulation results.
Smart Images

Figure CN123071124A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of public safety data simulation technology, and in particular to a method and system for generating emergency event time series that integrates long-term and short-term time patterns. Background Technology
[0002] With the development of intelligent alarm receiving, emergency dispatch, and public safety systems, emergency response systems increasingly require large-scale, high-quality data with real-time patterns for alarm receiver training, dispatch algorithm testing, system stress testing, and emergency response process evaluation. However, real emergency alarm data typically involves sensitive information such as alarm caller information, geographical location, event content, and handling records, and direct use poses risks related to privacy protection, data anonymization, and compliance. Furthermore, low-frequency, high-risk scenarios such as major fires, short-term alarm surges, and continuous emergencies are relatively scarce in real-world data, making it difficult to meet training and testing needs.
[0003] Therefore, constructing a data synthesis platform for emergency alarm flow simulation and intelligent alarm receiving training is of great significance. In this platform, the event triggering module determines the occurrence time of emergency events on the simulation timeline, serving as the foundation for subsequent event instantiation, state evolution, and alarm flow generation. If the event triggering time does not conform to real-world patterns, even if the subsequently generated alarm text is relatively realistic, the overall alarm flow may still lack temporal plausibility and scenario credibility.
[0004] Real-world emergency events typically do not occur uniformly but exhibit distinct temporal variations. On one hand, the frequency of events is influenced by factors such as hours, weeks, months, and seasons, exhibiting periodicity and non-stationarity. On the other hand, after an event occurs, there may be consecutive events, localized clusters, or alarm peaks within a short period. Therefore, an event triggering method is needed that can simultaneously characterize long-term temporal patterns and short-term clustering characteristics, providing a more realistic time-driven basis for synthesizing emergency alarm flows.
[0005] Traditional event triggering methods mainly include the following categories.
[0006] The first type is a manually configured event triggering scheme. This scheme typically requires the user to manually set the event occurrence time, number of events, or event interval. For example, an event may be generated at fixed intervals, or a number of events may be manually configured within a certain time period. This scheme is simple to implement, but it mainly relies on human experience.
[0007] The second type is a random triggering scheme with fixed probability or fixed rate. This scheme typically assumes that emergency events occur with a fixed probability or fixed occurrence rate, and uses a common Poisson process or simple random sampling method to generate event times. This scheme has a certain degree of randomness and is suitable for basic simulation scenarios.
[0008] The third type is triggering schemes based on historical statistical frequencies. This scheme calculates the frequency of events at different hours, weeks, or months based on historical data, and then generates event times according to the statistical distribution. This scheme can reflect the frequency differences across different time periods to a certain extent.
[0009] The fourth type is the triggering scheme based on the self-excitation process. This scheme assumes that an event that has already occurred will increase the probability of subsequent events occurring in a short period of time. Therefore, it can be used to describe the phenomenon of a dense occurrence of events or a surge in alarms in a short period of time.
[0010] While existing technologies can generate event times to a certain extent, they still have the following shortcomings: First, manually configured solutions rely on human experience and lack data-driven support. When simulation times are long or event scales are large, manual configuration becomes costly and it is difficult to guarantee the stability and reproducibility of the generated results.
[0011] Second, fixed probability or fixed rate schemes are difficult to reflect the non-stationarity of emergency events. The frequency of real events often varies between day and night, weekdays and weekends, and different months or seasons, while fixed rate models tend to generate overly uniform event sequences.
[0012] Third, while schemes based on historical statistical frequencies can reflect some periodic patterns, they usually only focus on the average frequency of occurrence and are difficult to describe the short-term impact of a certain event on subsequent events. Therefore, they are not capable of simulating continuous events, local clusters, and sudden alarm peaks.
[0013] Fourth, while a single self-excitation process scheme can characterize short-term aggregation features, it is difficult to maintain true daily, weekly, or seasonal variations if background time patterns are not modeled. This may overemphasize local suddenness while ignoring long-term time patterns.
[0014] Therefore, it is necessary to propose an emergency event time series generation method that can combine historical emergency event data to generate emergency event occurrence time series with periodicity, non-stationarity, short-term clustering, and controllability within a continuous time range, so as to provide a more realistic and stable event triggering basis for the data synthesis platform for emergency alarm flow simulation and intelligent alarm receiving training. Summary of the Invention
[0015] This invention proposes a method and system for generating emergency event time series that integrates long-term and short-term time patterns. It solves the problem that existing emergency event simulation triggering methods cannot simultaneously depict the long-term periodicity and short-term clustering effect of events, resulting in inconsistencies between the generated event sequences and the actual emergency alarm time patterns.
[0016] The technical solution of this invention is implemented as follows: The first aspect of this invention provides a method for generating time series data of emergency events that integrates long-term and short-term time patterns, comprising the following steps: S1: Acquire real historical emergency event data and preprocess it to extract the time information of the emergency events and form a standardized event time series. S2: Statistically standardize the frequency of emergency events in time series and construct a background intensity model to describe the occurrence pattern of emergency events under long-term time variations; S3: Using the background intensity model and standardized event time series, construct a self-excited process model to describe the short-term clustering effect of past events on the probability of subsequent events, and estimate the parameters of the self-excited process model. S4: Superimpose the parameters of the background intensity model and the self-excitation process model to generate a comprehensive conditional intensity function; S5: Sampling of events is performed based on the comprehensive condition intensity function to generate a time series of emergency events.
[0017] Specifically, in step S2, the method for constructing the background intensity model is as follows: Extract multiple time factors from standardized event time series and statistically analyze the frequency of historical events under different time factors; Based on the relationship between the occurrence frequency and the influence of the time factor on the intensity of the event, the background intensity model is obtained: ; in, express Background intensity at any given moment Indicates the global base strength; , and They represent The values of the three different time factors corresponding to the time; , and These represent the adjustment coefficients for the hourly factor, monthly factor, and weekly factor, respectively.
[0018] Specifically, step S3 includes the following steps: The self-excitation effect of historical events on the current moment is expressed as the sum of the excitation effects generated by all historical events, with the excitation effect decaying exponentially over time. The excitation amplitude and decay coefficient are obtained through parameter estimation methods. The expression for the self-excitation effect is as follows: ; in, This indicates that all historical events occurred in The intensity of self-excitation generated at any given moment. Indicates the first The time of occurrence of each historical event; This indicates the extent to which historical events increase the probability of the current event occurring. This indicates the rate at which the self-excitation effect of historical events on the current moment decays over time.
[0019] Furthermore, the parameter estimation method includes: An objective function is constructed using the parameters of the self-excitation process model as optimization variables. Initial values and search ranges for the parameters are set. A numerical optimization algorithm is used to iteratively optimize the objective function until it reaches its maximum value in the parameter space. The parameter value corresponding to this maximum value is used as the estimate of the excitation amplitude and decay coefficient. The objective function is a log-likelihood function. ; in, Represents the set of model parameters The log-likelihood function and model parameters are shown below. The set includes the parameters of the background intensity model, the excitation amplitude, and the attenuation coefficient; Represents the total number of historical events. Indicates the observation time range; Indicates in The collection of all historical events that have occurred before a certain point in time. Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given moment; Indicates in The collection of all historical events that have occurred before a certain point in time. Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given moment.
[0020] Specifically, in step S4, the expression for the comprehensive condition strength function is: ; in, Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given time. Indicates in The collection of all historical events that have occurred before a certain point in time; express Background intensity at any given moment; This indicates that all historical events occurred in The intensity of self-excitation generated at any given moment.
[0021] Specifically, step S5 includes the following steps: S51: Set the simulation time range [0, T], and initialize the current simulation time. and initialize the historical events to empty; S52: At the current simulation moment, a pre-defined upper limit for intensity is set. To ensure that it is not less than the comprehensive condition intensity within the current sampling interval; S53: Extract candidate waiting times from the exponential distribution Obtain candidate event times ; S54: Calculate candidate times using the comprehensive conditional intensity function. The overall condition strength is determined, and the acceptance probability is calculated: ; in, This represents the probability of accepting a candidate event as a real triggering event; Indicates candidate time The overall strength of the conditions; S55: Generate a random number that follows a uniform distribution in the range [0,1]. ,like If so, then accept the candidate event and... Record the actual triggered event and add it to the event history; if If the candidate event is not found, then the event history will not be updated. S56: Set the current simulation time Updated to Repeat steps S52 to S55 until the current simulation time exceeds the termination time T, and finally obtain a set of emergency event occurrence time series arranged in chronological order.
[0022] Preferably, the method further includes: S6: Compare and verify the generated emergency event occurrence time series with real historical emergency event data. If the verification result does not meet the preset requirements, adjust the parameters of the background intensity model or self-excitation process model, and repeat steps S2 to S5 until the verification result meets the preset requirements. The preset requirements include that the statistical difference between the generated event time series and the real historical event data in at least one of the following aspects—total number of events, event interval distribution, hourly distribution, weekly distribution, monthly distribution, and short-term clustering characteristics—is less than a preset threshold.
[0023] A second aspect of the present invention provides an emergency event time series generation system that integrates long-term and short-term time patterns, comprising: The background intensity construction module is used to build a background intensity model that describes the long-term time variation pattern based on real historical emergency event data. The self-excitation process parameter estimation module is used to estimate the parameters describing the short-term clustering impact based on the background intensity model and real historical emergency event data. The simulation and verification module is used to superimpose the background intensity model and self-excitation process parameters to generate a comprehensive conditional intensity function, generate an emergency event occurrence time series through sampling, and compare and verify it with real historical data.
[0024] Specifically, the background intensity construction module includes: The data loading and preprocessing unit is used to read historical emergency event data, perform format standardization, outlier processing and duplicate record filtering, and output standardized event time series. The time factor fitting unit is used to extract time factors from standardized event time series, statistically analyze the frequency of event occurrence under different time factors, fit the influence relationship of time factors on the intensity of event occurrence, and obtain the background intensity function. The background intensity output unit is used to output the background intensity function and its parameters for subsequent modules to call. The self-excitation parameter estimation module includes: Model definition unit, used to define the functional form of the self-excitation process; The recursive calculation unit is used to recursively calculate the self-excitation intensity at different times based on the event history. Whenever a new event occurs, this unit continuously updates the impact value of the historical event on the current time as the event time series progresses. The objective function calculation unit is used to calculate the objective function value based on the historical event time series and the current model parameters. The objective function is used to measure the degree of fit of the model to the actual event occurrence time. The parameter optimization unit is used to iteratively optimize the objective function using numerical optimization algorithms, and output the excitation amplitude and attenuation coefficient. The parameter output unit is used to output the self-excitation parameter results, including excitation amplitude, attenuation coefficient and model fitting value; The simulation and verification module includes: The rapid simulation unit is used to calculate the comprehensive condition intensity of emergency events based on background intensity and self-excitation parameters; The sparse sampling unit is used to generate candidate event times based on the comprehensive condition strength and to determine whether a candidate event is accepted. The results comparison unit is used to statistically compare the generated event sequence with real historical data and output the comparison results. The simulation result output unit is used to output the final simulation results.
[0025] Preferably, the system further includes a parameter adjustment module, which is used to trigger the background intensity construction module or the self-excitation process parameter estimation module to adjust the corresponding model parameters when the verification result does not meet the preset requirements, thereby forming iterative optimization.
[0026] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) This invention first constructs a background intensity model describing long-term periodic patterns and a self-excited process model describing short-term clustering effects based on real historical emergency event data. Then, the two are superimposed to form a comprehensive conditional intensity function. Finally, event sampling is performed based on this function to generate an emergency event occurrence time series. Since the background intensity model describes the non-stationarity and periodic change patterns of emergency events at time scales such as hours, weeks, and months, while the self-excited process model describes the short-term enhancement effect of the occurrence of events on the probability of subsequent events, the event series generated in the end has both the statistical consistency of the long-term time distribution of real emergency events and can present short-term clustering phenomena such as continuous events, local outbreaks, and alarm peaks. This solves the problem that a single model in the prior art cannot simultaneously take into account the two types of time patterns. (2) This invention extracts multiple time factors such as hours, months, and weeks from standardized event time series, statistically analyzes the frequency of events under each time factor, and uses a log-linear form to fit the influence relationship between time factors and event intensity to construct a background intensity model. This allows the background intensity to accurately reflect the differences in event frequency under different time periods, dates, and months. Compared with existing fixed-rate models or simple statistical frequency schemes, this method can effectively characterize the non-stationary and periodic changes of emergency events on long-term time scales, avoiding the problem of overly uniform generated results and lack of time variation characteristics, making the synthesized event sequence closer to real data in long-term statistical characteristics. (3) This invention expresses the self-excitation effect of historical events on the current moment as the sum of the excitation effects of all historical events, and the excitation effect decays exponentially over time. It constructs a log-likelihood function with the self-excitation process model parameters as optimization variables as the objective function, and uses a numerical optimization algorithm to make the objective function reach its maximum value in the parameter space. The parameter value corresponding to this maximum value is used as the estimation result of the excitation amplitude and decay coefficient. This allows the estimated self-excitation parameters to accurately characterize the short-term aggregation intensity and decay rate in real emergency events. Thus, it effectively reproduces the probability enhancement phenomenon after the event occurs and its decay law over time during the simulation generation process, making up for the shortcomings of existing schemes in simulating continuous events and sudden peaks. Attached Figure Description
[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0028] Figure 1 This is a flowchart illustrating a method for generating emergency event time series that integrates long-term and short-term time patterns according to the present invention.
[0029] Figure 2 This is an architecture diagram of an emergency event time series generation system that integrates long-term and short-term time patterns according to the present invention. Detailed Implementation
[0030] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0031] Reference Figure 1 The first aspect of this invention provides a method for generating time series data of emergency events that integrates long-term and short-term time patterns, comprising the following steps: Step S1: Obtain real historical emergency event data and preprocess it to extract the time information of the emergency events and form a standardized event time series; This step transforms the raw data into standardized event time series that can be used for modeling through data preprocessing, providing a standardized data foundation for subsequent background intensity construction and self-excitation parameter estimation.
[0032] The historical emergency event data includes at least the event occurrence time, and may further include auxiliary fields such as event type, event location, and number of alarms. Preprocessing includes formatting the raw data, handling outliers, filtering duplicate records, and parsing the time field.
[0033] Step S2: Statistically analyze the frequency of emergency events in standardized event time series, and construct a background intensity model to describe the long-term patterns of emergency events. This includes the following steps: Extract multiple time factors from standardized event time series and statistically analyze the frequency of historical events under different time factors; Based on the relationship between the occurrence frequency and the influence of the time factor on the intensity of the event, the background intensity model is obtained: ; in, express Background intensity at any given moment Indicates the global base strength; , and They represent The values of the three different time factors corresponding to the time; , and These represent the adjustment coefficients for the hourly factor, monthly factor, and weekly factor, respectively.
[0034] This step extracts multiple time factors such as hour, month, and weekday, statistically analyzes the frequency of events under each time factor, and constructs a background intensity model by fitting the influence of time factors on event intensity using a log-linear form. This ensures that the background intensity accurately reflects the differences in event frequency across different time periods, dates, and months. Compared to existing fixed-rate models or simple statistical frequency schemes, this approach effectively characterizes the non-stationary and periodic variations of emergency events over long-term time scales, avoiding the problems of overly uniform generated results and lack of temporal variation characteristics. This makes the synthesized event sequence closer to real data in terms of long-term statistical properties.
[0035] The time factors are not limited to hours, months, and weeks of the week, but may also include seasons, years, holidays, weather, temperature, regional risk levels, population density, major events, and other factors. Different application scenarios can select different combinations of time factors based on data availability to improve the authenticity of event-triggered results.
[0036] The calculation process of the background intensity model is as follows: First, the number of events in each hour of a 24-hour day, the number of events in each day of a 7-day week, and the number of events in each month of a 12-month year are counted; then, the influence relationship between each time factor and the intensity of event occurrence is fitted based on these statistical frequencies to obtain the background intensity function.
[0037] As an alternative implementation method, background intensity can also be constructed using segmented statistical tables, spline functions, kernel density estimation, time series prediction models, or neural network models, as long as the background intensity that changes over time can be obtained based on historical emergency event data.
[0038] Step S3: Using the background intensity model and standardized event time series, construct a self-excited process model to describe the short-term clustering effect of past events on the probability of subsequent events, and estimate the parameters of the self-excited process model; specifically including the following steps: The self-excitation effect of historical events on the current moment is expressed as the sum of the excitation effects generated by all historical events, with the excitation effect decaying exponentially over time; the expression for the self-excitation effect is: ; in, This indicates that all historical events occurred in The intensity of self-excitation generated at any given moment. Indicates the first The time of occurrence of each historical event; This indicates the extent to which historical events increase the probability of the current event occurring. This indicates the rate at which the self-excitation effect of historical events on the current moment decays over time.
[0039] The excitation amplitude and attenuation coefficient are obtained through parameter estimation methods, including: An objective function is constructed using the parameters of the self-excitation process model as optimization variables. Initial values and search ranges for the parameters are set. A numerical optimization algorithm is used to iteratively optimize the objective function until it reaches its maximum value in the parameter space. The parameter value corresponding to this maximum value is used as the estimate of the excitation amplitude and decay coefficient. The objective function is a log-likelihood function. ; in, Represents the set of model parameters The log-likelihood function and model parameters are shown below. The set includes the parameters of the background intensity model, the excitation amplitude, and the attenuation coefficient; Represents the total number of historical events. Indicates the observation time range; Indicates in The collection of all historical events that have occurred before a certain point in time. Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given moment; Indicates in The collection of all historical events that have occurred before a certain point in time. Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given moment.
[0040] This step represents the self-excitation impact of historical events on the current moment as the sum of the excitation effects generated by all historical events, with the excitation effects decaying exponentially over time. A log-likelihood function is constructed with the self-excitation process model parameters as optimization variables, serving as the objective function. A numerical optimization algorithm is used to maximize the objective function in the parameter space, and the parameter value corresponding to this maximum value is used as the estimate of the excitation amplitude and decay coefficient. This step enables the estimated self-excitation parameters to accurately characterize the short-term aggregation intensity and decay rate in real emergency events. Therefore, it effectively reproduces the probability enhancement phenomenon after an event occurs and its decay over time during the simulation generation process, overcoming the shortcomings of existing schemes in simulating continuous events and sudden peaks.
[0041] Each time a new event occurs, it affects the intensity of subsequent events. In engineering implementation, to optimize computational efficiency, a recursive approach is used to update the self-excitation intensity. That is, when a new event is generated, only the contribution of that event needs to be added to the self-excitation intensity accumulator. The self-excitation intensity at subsequent times can be recursively updated based on the current accumulated value and the time of the new event, avoiding the need to traverse all historical events from the beginning each time.
[0042] As an optional implementation, the functional form of the self-excitation effect is not limited to the exponential decay function. It can also use the power-law decay function, the piecewise decay function, the Gaussian kernel function, the rectangular window function, or other nonlinear decay functions, as long as they can represent the short-term influence of the already occurred event on the intensity of subsequent events.
[0043] As an optional implementation method, the self-excitation parameter is not limited to maximum likelihood estimation. It can also be estimated using other numerical optimization methods such as minimum error fitting, gradient descent, grid search, Bayesian estimation, genetic algorithm or particle swarm optimization, as long as the model parameters such as background intensity parameter, self-excitation amplitude and attenuation coefficient can be obtained from historical event data.
[0044] Step S4: Superimpose the parameters of the background intensity model and the self-excited process model to generate the comprehensive conditional intensity function, the expression of which is: ; in, Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given time. Indicates in The collection of all historical events that have occurred before a certain point in time; express Background intensity at any given moment; This indicates that all historical events occurred in The intensity of self-excitation generated at any given moment.
[0045] This step superimposes the background intensity and the self-excitation intensity to form a comprehensive conditional intensity. This comprehensive conditional intensity function simultaneously reflects long-term temporal variation patterns and short-term event clustering characteristics, providing a unified intensity driving function for subsequent event sampling. This comprehensive conditional intensity simultaneously includes two types of temporal patterns: one is the long-term temporal variation pattern caused by time factors such as hours, weeks, and months; the other is the short-term event clustering pattern caused by already occurred events.
[0046] Step S5: Sampling of events based on the comprehensive conditional intensity function to generate a time series of emergency events, specifically including the following steps: Step S51: Set the simulation time range [0, T] and initialize the current simulation time. and initialize the historical events to empty; Step S52: At the current simulation moment, preset an upper limit for intensity. To ensure that it is not less than the comprehensive condition intensity within the current sampling interval; Step S53: Extract candidate waiting times from the exponential distribution Obtain candidate event times ; Step S54: Calculate candidate time using the comprehensive conditional intensity function. The overall condition strength is determined, and the acceptance probability is calculated: ; in, This represents the probability of accepting a candidate event as a real triggering event; Indicates candidate time The overall strength of the conditions; Step S55: Generate a random number that follows a uniform distribution in the range [0,1]. ,like If so, then accept the candidate event and... Record the actual triggered event and add it to the event history; if If the candidate event is not found, then the event history will not be updated. Step S56: Set the current simulation time Updated to Repeat steps S52 to S55 until the current simulation time exceeds the termination time T, and finally obtain a set of emergency event occurrence time series arranged in chronological order.
[0047] As an optional implementation, the sampling method for the event occurrence time is not limited to the sparse sampling method (i.e., accept-reject sampling) described in this embodiment. Other point process simulation algorithms such as inverse function sampling, discrete time approximation sampling, or Gillespie sampling can also be used to generate the event time series, as long as the occurrence time of the emergency event can be generated based on the condition intensity that changes over time.
[0048] Step S6: Compare and verify the generated emergency event occurrence time series with real historical emergency event data. If the verification result does not meet the preset requirements, adjust the parameters of the background intensity model or self-excitation process model, and repeat steps S2 to S5 until the verification result meets the preset requirements. The preset requirements include that the statistical difference between the generated event time series and the real historical event data in at least one of the following aspects—total number of events, event interval distribution, hourly distribution, weekly distribution, monthly distribution, and short-term clustering characteristics—is less than a preset threshold.
[0049] The closed-loop verification mechanism in this step transforms the event triggering process from a one-way, one-time generation into an iterative process of modeling, generation, verification, feedback, and optimization. This continuously improves the consistency between the generated results and real data in multiple statistical characteristics, ensuring the comprehensive authenticity of the synthesized event sequence across different time scales and aggregation characteristics.
[0050] like Figure 2 As shown, a second aspect of the present invention provides an emergency event time series generation system that integrates long-term and short-term time patterns, comprising: 1) Background intensity construction module, used to construct a background intensity model describing long-term time variation patterns based on real historical emergency event data; the background intensity construction module includes: The data loading and preprocessing unit is used to read historical emergency event data, perform format standardization, outlier processing and duplicate record filtering, and output standardized event time series. The time factor fitting unit is used to extract time factors from standardized event time series, statistically analyze the frequency of event occurrence under different time factors, fit the influence relationship of time factors on the intensity of event occurrence, and obtain the background intensity function. The background intensity output unit is used to output the background intensity function and its parameters for subsequent modules to call. 2) A self-excitation process parameter estimation module, used to estimate self-excitation process parameters describing short-term clustering effects based on a background intensity model and real historical emergency event data; the self-excitation parameter estimation module includes: Model definition unit, used to define the functional form of the self-excitation process; The recursive calculation unit is used to recursively calculate the self-excitation intensity at different times based on the event history. Whenever a new event occurs, this unit continuously updates the impact value of the historical event on the current time as the event time series progresses. The objective function calculation unit is used to calculate the objective function value based on the historical event time series and the current model parameters. The objective function is used to measure the degree of fit of the model to the actual event occurrence time. The parameter optimization unit is used to iteratively optimize the objective function using numerical optimization algorithms, and output the excitation amplitude and attenuation coefficient. The parameter output unit is used to output the self-excitation parameter results, including excitation amplitude, attenuation coefficient and model fitting value; 3) A simulation and verification module, used to superimpose the background intensity model and self-excitation process parameters to generate a comprehensive conditional intensity function, generate an emergency event occurrence time series through sampling, and compare and verify it with real historical data; the simulation and verification module includes: The rapid simulation unit is used to calculate the comprehensive condition intensity of emergency events based on background intensity and self-excitation parameters; The sparse sampling unit is used to generate candidate event times based on the comprehensive condition strength and to determine whether a candidate event is accepted. The results comparison unit is used to statistically compare the generated event sequence with real historical data and output the comparison results. The simulation result output unit is used to output the final simulation results.
[0051] 4) Parameter adjustment module, which is used to trigger the background intensity construction module or the self-excitation process parameter estimation module to adjust the corresponding model parameters when the verification results do not meet the preset requirements, so as to form iterative optimization.
[0052] The event time series generated by this invention does not directly contain sensitive information such as the identity, contact information, precise address, or specific alarm content of the actual alarm caller. It can serve as the time-driven basis for subsequent synthetic data generation, applicable to scenarios such as dispatcher training, intelligent dispatch algorithm testing, system stress testing, and emergency response plan simulation. This invention can be preferentially applied to fire alarm flow simulation scenarios, but can also be extended to other emergency event types such as gas leaks, traffic accidents, people trapped, flood drainage, medical assistance, and natural disasters. For different event types, corresponding background intensity and self-excitation parameters can be established separately, or a unified multi-type event triggering model can be established.
[0053] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for generating time series data of emergency events that integrates long-term and short-term time patterns, characterized in that, Includes the following steps: S1: Acquire real historical emergency event data and preprocess it to extract the time information of the emergency events and form a standardized event time series. S2: Statistically standardize the frequency of emergency events in time series and construct a background intensity model to describe the occurrence pattern of emergency events under long-term time variations; S3: Using the background intensity model and standardized event time series, construct a self-excited process model to describe the short-term clustering effect of past events on the probability of subsequent events, and estimate the parameters of the self-excited process model. S4: Superimpose the parameters of the background intensity model and the self-excitation process model to generate a comprehensive conditional intensity function; S5: Sampling of events is performed based on the comprehensive condition intensity function to generate a time series of emergency events.
2. The emergency event time series generation method that integrates long-term and short-term time patterns as described in claim 1, characterized in that, In step S2, the method for constructing the background intensity model is as follows: Extract multiple time factors from standardized event time series and statistically analyze the frequency of historical events under different time factors; Based on the relationship between the occurrence frequency and the influence of the time factor on the intensity of the event, the background intensity model is obtained: ; in, express Background intensity at any given moment Indicates the global base strength; , and They represent The values of the three different time factors corresponding to the time; , and These represent the adjustment coefficients for the hourly factor, monthly factor, and weekly factor, respectively.
3. The method for generating emergency event time series by integrating long-term and short-term time patterns as described in claim 1, characterized in that, Step S3 includes the following steps: The self-excitation effect of historical events on the current moment is expressed as the sum of the excitation effects generated by all historical events, with the excitation effect decaying exponentially over time. The excitation amplitude and decay coefficient are obtained through parameter estimation methods. The expression for the self-excitation effect is as follows: ; in, This indicates that all historical events occurred in The intensity of self-excitation generated at any given moment. Indicates the first The time of occurrence of each historical event; This indicates the extent to which historical events increase the probability of the current event occurring. This indicates the rate at which the self-excitation effect of historical events on the current moment decays over time.
4. The emergency event time series generation method that integrates long-term and short-term time patterns as described in claim 3, characterized in that, The parameter estimation method includes: An objective function is constructed using the parameters of the self-excitation process model as optimization variables. Initial values and search ranges for the parameters are set. A numerical optimization algorithm is used to iteratively optimize the objective function until it reaches its maximum value in the parameter space. The parameter value corresponding to this maximum value is used as the estimate of the excitation amplitude and decay coefficient. The objective function is a log-likelihood function. ; in, Represents the set of model parameters The log-likelihood function and model parameters are shown below. The set includes the parameters of the background intensity model, the excitation amplitude, and the attenuation coefficient; Represents the total number of historical events. Indicates the observation time range; Indicates in The collection of all historical events that have occurred before a certain point in time. Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given moment; Indicates in The collection of all historical events that have occurred before a certain point in time. Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given moment.
5. The emergency event time series generation method that integrates long-term and short-term time patterns as described in claim 1, characterized in that, In step S4, the expression for the comprehensive condition strength function is: ; in, Indicates the set of known historical events Under the conditions, The instantaneous intensity of new events occurring at any given time. Indicates in The collection of all historical events that have occurred before a certain point in time; express Background intensity at any given moment; This indicates that all historical events occurred in The intensity of self-excitation generated at any given moment.
6. The method for generating emergency event time series by integrating long-term and short-term time patterns as described in claim 1, characterized in that, Step S5 includes the following steps: S51: Set the simulation time range [0, T], and initialize the current simulation time. and initialize the historical events to empty; S52: At the current simulation moment, a pre-defined upper limit for intensity is set. To ensure that it is not less than the comprehensive condition intensity within the current sampling interval; S53: Extract candidate waiting times from the exponential distribution Obtain candidate event times ; S54: Calculate candidate times using the comprehensive conditional intensity function. The overall condition strength is determined, and the acceptance probability is calculated: ; in, This represents the probability of accepting a candidate event as a real triggering event; Indicates candidate time The overall strength of the conditions; S55: Generate a random number that follows a uniform distribution in the range [0,1]. ,like If so, then accept the candidate event and... Record the actual triggered event and add it to the event history; if If the candidate event is not found, then the event history will not be updated. S56: Set the current simulation time Updated to Repeat steps S52 to S55 until the current simulation time exceeds the termination time T, and finally obtain a set of emergency event occurrence time series arranged in chronological order.
7. The emergency event time series generation method that integrates long-term and short-term time patterns as described in claim 1, characterized in that, The method further includes: S6: Compare and verify the generated emergency event occurrence time series with real historical emergency event data. If the verification result does not meet the preset requirements, adjust the parameters of the background intensity model or self-excitation process model, and repeat steps S2 to S5 until the verification result meets the preset requirements. The preset requirements include that the statistical difference between the generated event time series and the real historical event data in at least one of the following aspects—total number of events, event interval distribution, hourly distribution, weekly distribution, monthly distribution, and short-term clustering characteristics—is less than a preset threshold.
8. An emergency event time series generation system integrating long-term and short-term time patterns, used to implement the emergency event time series generation method according to any one of claims 1-7, characterized in that, include: The background intensity construction module is used to build a background intensity model that describes the long-term time variation pattern based on real historical emergency event data. The self-excitation process parameter estimation module is used to estimate the parameters describing the short-term clustering impact based on the background intensity model and real historical emergency event data. The simulation and verification module is used to superimpose the background intensity model and self-excitation process parameters to generate a comprehensive conditional intensity function, generate an emergency event occurrence time series through sampling, and compare and verify it with real historical data.
9. The emergency event time series generation system that integrates long-term and short-term time patterns as described in claim 8, characterized in that, The background intensity construction module includes: The data loading and preprocessing unit is used to read historical emergency event data, perform format standardization, outlier processing and duplicate record filtering, and output standardized event time series. The time factor fitting unit is used to extract time factors from standardized event time series, statistically analyze the frequency of event occurrence under different time factors, fit the influence relationship of time factors on the intensity of event occurrence, and obtain the background intensity function. The background intensity output unit is used to output the background intensity function and its parameters for subsequent modules to call. The self-excitation parameter estimation module includes: Model definition unit, used to define the functional form of the self-excitation process; The recursive calculation unit is used to recursively calculate the self-excitation intensity at different times based on the event history. Whenever a new event occurs, this unit continuously updates the impact value of the historical event on the current time as the event time series progresses. The objective function calculation unit is used to calculate the objective function value based on the historical event time series and the current model parameters. The objective function is used to measure the degree of fit of the model to the actual event occurrence time. The parameter optimization unit is used to iteratively optimize the objective function using numerical optimization algorithms, and output the excitation amplitude and attenuation coefficient. The parameter output unit is used to output the self-excitation parameter results, including excitation amplitude, attenuation coefficient and model fitting value; The simulation and verification module includes: The rapid simulation unit is used to calculate the comprehensive condition intensity of emergency events based on background intensity and self-excitation parameters; The sparse sampling unit is used to generate candidate event times based on the comprehensive condition strength and to determine whether a candidate event is accepted. The results comparison unit is used to statistically compare the generated event sequence with real historical data and output the comparison results. The simulation result output unit is used to output the final simulation results.
10. The emergency event time series generation system that integrates long-term and short-term time patterns as described in claim 8, characterized in that, The system also includes a parameter adjustment module, which is used to trigger the background intensity construction module or the self-excitation process parameter estimation module to adjust the corresponding model parameters when the verification result does not meet the preset requirements, thereby forming iterative optimization.
