Information issuing early warning system and method based on LED screen

Abstract

The invention discloses an information release early warning system and method based on an LED screen, and relates to the technical field of information release, the system comprises a data division module, a model construction module, an early warning cache module and a release distribution module; according to the technical key points, collected multi-source information is divided into a normal data set and an abnormal data set according to short-term, middle-term and long-term labels, different time levels are set accordingly, early warning time of a plurality of priorities under the different time levels is judged, early warning events are quickly and intelligently predicted in combination with early warning signals and an attention mechanism, and the early warning efficiency is improved. The problem of information delay is solved; the distribution level is calculated through the feature information, the distribution box is designed, the distribution adaptation degree is calculated based on the distribution box information, the distribution data packet is generated through processing and uploaded to the LED screen, and the accuracy and fineness of weather forecast are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of information release, and in particular to an information release warning system and method based on a LED screen. Background Art

[0002] In recent years, LED display screens have gradually been used in the field of information release. Their advantages such as high brightness, wide viewing angle, and flexible information update make them an ideal choice for urban public information release. However, the existing information release system based on LED screens still has some limitations in the following aspects: On the one hand, traditional information release is based on known weather reports and cannot reflect the latest weather changes in a timely manner. It lacks intelligent early warning judgment and information push capabilities, resulting in poor early warning effects for emergencies. At the same time, weather forecasts are usually released at fixed times, and frequent weather changes may not be covered in a timely manner. To a certain extent, there are delays in information updates, which cannot reflect the latest situation of emergencies in a timely manner. On the other hand, traditional weather forecasts may not be accurate and detailed enough, which may cause users to obtain outdated data and fail to meet the needs of specific users. Summary of the invention

[0003] 1. Technical issues to be resolved In view of the deficiencies in the prior art, the present invention provides an information release and warning system and method based on an LED screen, the system comprising: a data division module, a model building module, a warning cache module and a release distribution module, which divides the collected multi-source information into a normal data set and an abnormal data set according to time tags, thereby setting different time levels, judging the warning times of several priorities at different time levels, and combining the warning signal and the attention mechanism to quickly and intelligently predict warning events; by setting a release frame to calculate the release adaptability, and correspondingly processing it into a release data packet to upload to the LED screen, to a certain extent, improves the accuracy and detail of the forecast, and solves the problems raised in the background technology.

[0004] (II) Technical solution To achieve the above objectives, the present invention is implemented through the following technical solutions: Data partitioning module, creating normal data set and abnormal data set according to weather reports; The model building module is connected to the data source in real time through the LED screen, senses the multi-source information collected by the data source in real time, inputs the multi-source information collected in real time into the preset event classification model, and outputs the first probability that the collected multi-source information belongs to the normal data set and the second probability that the collected multi-source information belongs to the abnormal data set; a rule engine is built based on the first probability and the second probability, the warning signal is divided, and several warning events corresponding to the priority are generated; wherein the collected multi-source information includes temperature, precipitation and wind speed; The warning cache module extracts feature information from warning events, builds a publishing rule engine, generates publishing data packets, and caches the publishing data packets; The release distribution module obtains the terminal information of the LED screen, determines the release priority based on the release data packet, and generates an information release plan; wherein the terminal information includes the information release queue and the server bandwidth.

[0005] Furthermore, the process of creating a normal data set and an abnormal data set based on the weather report is as follows: The twenty-four hours of each day are divided into several time intervals, and meteorological reports within the time intervals are actively collected. The time stage of the corresponding climate event is determined based on the meteorological report, the frequency characteristics of the climate event are obtained based on the time stage, and normal data sets and abnormal data sets are created based on the frequency characteristics; time tags are inserted into the normal data set and the abnormal data set respectively; among them, climate events are divided into normal events and warning events; time tags include short-term tags, medium-term tags and long-term tags.

[0006] Furthermore, the process of obtaining the first probability and the second probability is: Construct an initial classification model. First, construct a four-dimensional monitoring vector through the collected multi-source information, including temperature parameters, precipitation parameters, wind speed parameters and time. For any time interval Ta, analyze the average and fluctuation values ​​of temperature parameters, precipitation parameters and wind speed parameters in the time interval Ta; generate new features based on the average and fluctuation values; where the fluctuation value represents the average value of the difference between the maximum and minimum values ​​of the corresponding parameters in the time interval Ta; Introducing a training layer into the initial classification model, combining new features with original features to form a new training data set, and providing randomly selected sample data from the collected multi-source information to train and update the training layer to obtain an event classification model; The acquired multi-source information is stratified according to time labels, and the corresponding multi-source information at each time level is input into the event classification model to determine the first probability that the input multi-source information belongs to the normal data set and the second probability that the input multi-source information belongs to the abnormal data set.

[0007] Furthermore, when the first probability or the second probability is greater than a preset threshold, the collected multi-source information is correspondingly allocated to the corresponding data set; and the larger value of the first probability or the second probability is determined, and the collected multi-source information is correspondingly allocated to the data set corresponding to the larger value; when the first probability or the second probability is less than the preset threshold, reclassification is performed based on the first probability or the second probability, and the specific process is: When the first probability or the second probability is greater than the time tag of the normal data set or the abnormal data set, the collected multi-source information is correspondingly allocated to the corresponding data set; When the first probability or the second probability is smaller than the time tag of the normal data set or the abnormal data set, the collected multi-source information is continuously sensed.

[0008] Furthermore, the process of dividing warning events is as follows: In chronological order, within a predetermined time period T, the time points corresponding to the collected multi-source information are marked on the timeline, and a time series analysis method is used in combination with a preset density threshold to identify the time period in which the density corresponding to the abnormal data set exceeds the density threshold; based on the analysis results of the time period, an attention mechanism is triggered, the analysis results are compared and sorted in descending order, a data list is displayed, and a corresponding warning signal is generated to divide warning events corresponding to several categories of priority.

[0009] Furthermore, the extracted characteristic information includes time, type, range, trend and medium method, and the process of generating the release data package is: Build a publishing rule engine to extract types, ranges, and trends from the extracted feature information and define them as key factors. Perform parameter analysis on the values ​​of each key factor to obtain the corresponding parameter values, and keep the parameter values ​​and weights synchronized. The process of parameter analysis is as follows: The value of the key factor is marked as Jb, and the standard interval of the key factor Jb is set to [r1, r2], and the parameter value is calculated based on the formula: ; In the formula, P is the parameter value corresponding to the key factor, σ represents a constant, which is usually 1; Then assign a corresponding weight to each key factor, multiply the parameter value of each factor by its weight, and add up all the results to get the release level; Divide the release level interval based on the release level, and set a first release box smaller than the release level interval, a second release box in the release level interval, and a third release box larger than the release level interval, respectively set corresponding selection interfaces and keyword input interfaces in the first release box, the second release box, and the third release box, select the corresponding warning event based on the release box information, extract the corresponding interface attribute information based on the release box information and perform preprocessing analysis to calculate the release adaptability of the warning event; The warning events are sorted according to the numerical values ​​of the release adaptability to obtain the sorted warning events, and the corresponding parameter information and data format type are identified based on the warning events. The parameter information is parameter data obtained based on the characteristic information. Based on the parameter information and the data format type, a protocol conversion rule is used to convert the format of the warning events, and a release data packet in a unified format is sorted and output; wherein the release data packet includes: warning event, release level, release area, release time and release frequency.

[0010] Furthermore, the attribute information corresponding to the selection interface includes type, priority and event range, and the attribute information corresponding to the keyword interface includes content, relevance, frequency and context; the preprocessing analysis process is: Construct a preprocessing analysis model: mark the input information as a parameter set Q, which contains N0 element values. Mark any element value as Xi, assign corresponding weight factor coefficients to the N0 element values, and the weight factor coefficient of the marked element value Xi is φi. Establish the formula to output the influence coefficient Qy of the parameter set Q: ; Apply the preprocessing analysis model: first integrate the attribute information corresponding to the selection interface and the keyword interface into the selection parameter Q1 and the keyword parameter Q2; where Q1 = {type, priority, event range}, Q2 = {content, relevance, frequency, context}; then input the selection parameters and keyword parameters into the preprocessing analysis model in sequence, and output the selection matching degree tc1 and the keyword matching degree tc2 respectively; The calculation formula for publishing fitness is: ; In the formula, Ad represents the degree of adaptation, s represents the warning level, is the time gain factor; ch represents the difference between the occurrence time of the warning event and the current time, and G represents the time constant.

[0011] Furthermore, the process of determining the release priority is: First, determine whether there is an information release queue in the current LED screen. If an information release queue is currently running, obtain the bandwidth requested at this time and determine whether it is in a saturated state. When the current bandwidth is in a saturated state, determine whether there is a new release request for access. When there is a release request for access, immediately analyze the request release data packet information, reallocate it according to the warning level of the warning event, and generate an information release plan.

[0012] An information release warning method based on an LED screen comprises the following steps: Determine data types: Create normal data sets and abnormal data sets based on weather reports; Determine the event type: connect to the data source in real time through the LED screen, perceive the multi-source information collected by the data source in real time, input the multi-source information collected in real time into the preset event classification model, and output the first probability that the collected multi-source information belongs to the normal data set and the second probability of the abnormal data set; build a rule engine based on the first probability and the second probability, divide the warning signal, and generate several warning events corresponding to the priority levels; Formulate information release rules: extract feature information from warning events, build a release rule engine, generate release data packets, and cache the release data packets; Generate an information release plan: obtain the terminal information of the LED screen, and determine the release priority based on the release data packet to generate an information release plan; the terminal information includes the information release queue and server bandwidth.

[0013] (III) Beneficial effects The present invention provides an information release warning system and method based on LED screen, which has the following beneficial effects: 1. The present invention focuses on the frequency characteristics of climate events and designs an event classification model. It divides the collected multi-source information into normal data sets and abnormal data sets according to short-term, medium-term and long-term labels, and reflects the latest meteorological changes in real time. In this process, by analyzing the first probability and the second probability, the abnormality of the multi-source information can be quickly and clearly divided, and the early warning events can be effectively judged. At the same time, meteorological early warning events often show periodicity, which is convenient for predicting emergencies and ensuring the accuracy of forecasts. 2. The present invention analyzes abnormal data sets at different time levels by setting time levels, identifies warning events of different priorities at each time level, and combines warning signals and attention mechanisms to improve the response speed and ability to warning events, thereby reducing the impact and losses caused by warning events; 3. The present invention is connected to the data source through the LED screen, and can process the collected multi-source information to timely reflect the latest meteorological changes, quickly and intelligently push warning events, and to a certain extent solve the delay of information; calculate the release level through comprehensive feature information, design the release frame, and calculate the release adaptability based on the release frame information, generate a release data packet and upload it to the LED screen to accurately forecast the weather. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a module schematic diagram showing an information release warning system based on an LED screen according to an exemplary embodiment; Figure 2 The present invention is a flow chart showing a method for issuing early warning information based on an LED screen according to an exemplary embodiment. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] Embodiment 1: This embodiment provides an information release warning system based on LED screen. Figure 1 is a schematic diagram of a module of an information release warning system based on an LED screen according to an exemplary embodiment. Figure 1 ,The information release and early warning system includes: a data partition module, a model building module, an early warning cache module and a release and allocation module; ,the data partition module, the model building module, the early warning cache module and the release and allocation module are ,communicated with each module; The following are specific descriptions and explanations of each module or component: Data partitioning module: Create normal data sets and abnormal data sets based on weather reports; The specific process is: Divide the twenty-four hours of each day into several time intervals, actively collect weather reports within the time intervals, determine the time stage of the corresponding climate event based on the weather reports, obtain the frequency characteristics of the climate event based on the time stage, and create a normal data set and an abnormal data set based on the frequency characteristics; insert time tags into the normal data set and the abnormal data set respectively; wherein the time tags include short-term tags, medium-term tags and long-term tags; It should be noted that climate events generated by meteorological reports are generally related to trends or seasons. Climate events are divided into normal events and warning events. By dividing the time scale, we can clearly understand the corresponding information data under short-term warning, medium-term warning and long-term warning, and provide different time labels to divide normal data sets and abnormal data sets; for short-term labels, it usually refers to weather forecasts within 24 to 72 hours; for medium-term labels, it usually refers to weather forecasts within 3 to 10 days; for long-term labels, it usually refers to weather forecasts for more than 10 days, or even to a month or season forecast; it should be noted that abnormal data sets are often manifested as abnormal temperature, abnormal precipitation and abnormal wind speed. Through subsequent monitoring behavior, the corresponding warning events can be analyzed; Model building modules: Connect to the data source in real time through the LED screen, perceive the multi-source information collected by the data source in real time, input the multi-source information collected in real time into the preset event classification model, and output the first probability that the collected multi-source information belongs to the normal data set and the second probability of the abnormal data set; build a rule engine based on the first probability and the second probability, divide the warning signal, and generate several types of warning events corresponding to the priority; the data source refers to the place where the original data is provided, which can be a database, sensor, API, online service, etc. The collected multi-source information includes temperature, precipitation and wind speed; The process of obtaining the first probability and the second probability is: Construct an initial classification model. First, construct a four-dimensional monitoring vector through the collected multi-source information, including temperature parameters, precipitation parameters, wind speed parameters and time. For any time interval Ta, analyze the average and fluctuation values ​​of temperature parameters, precipitation parameters and wind speed parameters in the time interval Ta; generate new features based on the average and fluctuation values; where the fluctuation value represents the average value of the difference between the maximum and minimum values ​​of the corresponding parameters in the time interval Ta; A training layer is introduced into the initial classification model, and the new features are combined with the original features to form a new training data set. Sample data is randomly selected from the collected multi-source information, and the training layer is trained and updated to obtain an event classification model. It should be noted that the training process is based on a convolutional neural network architecture and is classified based on a support vector machine. The feature space is expanded by adding new features to form a data set containing original features and new features. The introduction of a training layer means that the training of the model is not only based on a static data set, but can be dynamically updated by continuously introducing new samples (or randomly selecting samples from multi-source data). After several rounds of training and model updates, a classification model is finally obtained, which can predict the category of an event based on input features (a combination of original features and newly added features). The acquired multi-source information is layered according to the time label, and the corresponding multi-source information at each time layer is input into the event classification model to determine the first probability that the input multi-source information belongs to the normal data set and the second probability that the input multi-source information belongs to the abnormal data set; When the first probability or the second probability is greater than a preset threshold, the collected multi-source information is correspondingly allocated to the corresponding data set; and the larger value of the first probability or the second probability is determined, and the collected multi-source information is correspondingly allocated to the data set corresponding to the larger value; When the first probability or the second probability is less than a preset threshold, reclassification is performed based on the first probability or the second probability; wherein the preset threshold is generally 50%. For example, when the first probability is 20% and the second probability is 80%, the collected multi-source information is classified as an abnormal data set. The specific process of reclassification is as follows: When the first probability or the second probability is greater than the time tag of the normal data set or the abnormal data set, the collected multi-source information is correspondingly allocated to the corresponding data set; When the first probability or the second probability is less than the time label of the normal data set or the abnormal data set, the multi-source information collected is continuously sensed; It should be noted that the time tag contains frequency characteristics. By recording the first probability or second probability of the collected multi-source information in the time series, usually using a line graph to draw, the corresponding trend and fluctuation can be seen; when the collected multi-source information meets the time tag, the collected information is classified into the corresponding data set; For example: assuming that in a short period of time, the time labels of the normal data set and the abnormal data set each account for 50%, then the corresponding time ratio is also 50%. When the first probability of belonging to the abnormal data set calculated by the initial classification model is 70%, and the second probability of belonging to the normal data set is 30%, since 70% is greater than 50%, the collected multi-source information is classified into the abnormal data set; the second probability of belonging to the normal data set calculated by the initial classification model is 30%, and the second probability of belonging to the abnormal data set is 30%. Since both 30% and 30% are less than 50%, the collected multi-source information belongs neither to the normal data set nor to the abnormal data set. The collected multi-source information is continuously sensed according to the time series until the classification is completed; The process of classifying warning events is as follows: In chronological order, within a predetermined time period T, the time points corresponding to the collected multi-source information are marked on the timeline, and the time series analysis method is combined with a preset density threshold to identify the time period when the density of the abnormal data set exceeds the density threshold; according to the analysis results of the time period, the attention mechanism is triggered, the analysis results are compared, and sorted in descending order, the data list is displayed, and the corresponding warning signals are generated to be divided into several types of warning events corresponding to the priority level; The specific process is: The collected multi-source information is marked on the time series, and the density corresponding to the first probability or the second probability generated in each time period is calculated by using the time series analysis method. For different time levels, the density threshold corresponding to the abnormal data set is set, and compared with the density threshold to identify the time period whose corresponding density exceeds the density threshold; when M0 high-density time periods are identified, the first probability generated in each high-density time period is evaluated, for example: by comparing the average value of the first probability of each high-density time period, the attention priority of different high-density time periods is positively correlated with the size of the corresponding average value, the corresponding high-density time period with the largest average value is paid attention to first, and the corresponding high-density time period with the smallest average value is paid attention to last, and high, medium and low priority warning events are set in turn according to the order of attention; Warning cache module: Extract feature information from warning events, build a publishing rule engine, generate a publishing data package, and cache the publishing data package; the extracted feature information includes time, type, range, trend, and medium method, and the process of generating a publishing data package is as follows: Build a publishing rule engine to extract types, ranges, and trends from the extracted feature information and define them as key factors. Perform parameter analysis on the values ​​of each key factor to obtain the corresponding parameter values, and keep the parameter values ​​and weights synchronized. The process of parameter analysis is as follows: The value of the key factor is marked as Jb, and the standard interval of the key factor Jb is set to [r1, r2], and the parameter value is calculated based on the formula: ; In the formula, P is the parameter value corresponding to the key factor, σ represents a constant, which is usually 1; It should be noted that when Jb is within the standard interval, the larger the denominator, the smaller the corresponding reference value, and the smaller the weight value assigned; when Jb is not within the standard interval, the smaller the denominator, the larger the corresponding reference value, the larger the weight value assigned, and the higher the warning level; Type: Red warning, weighted 4 > Orange warning, weighted 3 > Yellow warning, weighted 2 > Blue warning, weighted 1; Scope: A wide impact area is assigned a weight of 4 > A relatively wide impact area is assigned a weight of 3 > Impact on coastal areas is assigned a weight of 2 > Impact on specific areas is assigned a weight of 1; Trend: Long-term, weighted 3 > medium-term, weighted 2 > short-term, weighted 1; Then assign a corresponding weight to each key factor, multiply the parameter value of each factor by its weight, and add up all the results to get the release level; Assume that there are data 1 and data 2: Data 1: Type: Red Alert (weight 4): 4*4=16; Scope: Wide Impact (weight 4): 4*4=16; Trend: Long-term (weight 3): 3*3=9; The release level is: 16+16+9=41; Data 2: Type: Orange warning (weight 3): 3*3=9; Scope: Wide impact (weight 4): 4*4=16; Trend: Long-term (weight 3): 3*3=9; The release level is: 9+16+9=34; Through the above calculation, it can be found that the release level of data 1 is significantly greater than that of data 2; Divide the release level interval based on the release level, set a first release box that is smaller than the release level interval, a second release box in the release level interval, and a third release box that is larger than the release level interval, respectively set corresponding selection interfaces and keyword input interfaces in the first release box, the second release box, and the third release box, select the corresponding warning event based on the release box information, extract the corresponding interface attribute information based on the release box information and perform preprocessing analysis to calculate the release adaptability of the warning event; It should be noted that the release level interval is set to count the historical release levels, and the corresponding average and standard deviation are calculated, and the sum of the average and twice the standard deviation is marked as the standard release level; The formula for publishing the level interval is: ; In the formula, By is the standard release level, k is the preset fluctuation ratio, and θ is the conversion coefficient between the release level and the standard release level; The attribute information corresponding to the interface includes type, priority, and event range, and the attribute information corresponding to the keyword interface includes content, relevance, frequency, and context; For example, using json format: #Set the selection interface: { "interface":[ { "priority": "high priority", "Trend": "Long Term", "Type": "Red Alert", "Event scope": "Wide impact" }, { "Priority": "Medium Priority", "Trend": "Medium term", "Type": "Orange Warning", "Event scope": "Wide impact" }, { "priority": "low priority", "Trend": "Short Term", "Type": "Blue Warning", "Event scope": "Affects a specific area" } ] } #Set keyword interface: { "Keywords": { "Inside Allow ":"flood", "correlation": 0.9, "Frequency": "15, "context": ["weather warning", "emergency response"] } } The process of preprocessing analysis is: Construct a preprocessing analysis model: mark the input information as a parameter set Q, which contains N0 element values. Mark any element value as Xi, assign corresponding weight factor coefficients to the N0 element values, and the weight factor coefficient of the marked element value Xi is φi. Establish the formula to output the influence coefficient Qy of the parameter set Q: ; Apply the preprocessing analysis model: first integrate the attribute information corresponding to the selection interface and the keyword interface into the selection parameter Q1 and the keyword parameter Q2; where Q1 = {type, priority, event range}, Q2 = {content, relevance, frequency, context}; then input the selection parameters and keyword parameters into the preprocessing analysis model in sequence, and output the selection matching degree tc1 and the keyword matching degree tc2 respectively; Select the matching degree: ; In the formula, tc1 represents the selection matching degree, b1, b2 and b3 represent the values ​​corresponding to the type, priority and event range respectively; ε1, ε2 and ε3 represent the weight coefficients corresponding to the type, priority and event range respectively, and ε1, ε2 and ε3 are all greater than 0; Keyword matching: ; Where tc2 represents the keyword interface matching degree, d1, d2, d3 and d4 represent the values ​​corresponding to content, relevance, frequency and context respectively; σ1, σ2, σ3 and σ4 represent the weight coefficients corresponding to content, relevance, frequency and context respectively, and σ1, σ2, σ3 and σ4 are all greater than 0; The calculation formula for publishing fitness is: ; In the formula, Ad represents the degree of adaptation, s represents the warning level, is the time gain factor, which means that the longer the warning event occurs, the greater the impact it causes, so time gain is needed; ch represents the difference between the warning event occurrence time and the current time, and G represents the time constant; usually the setting value of the time constant is set according to the time label, when it is short-term, it is 24 hours to 72 hours; when it is medium-term, it is 3 days to 10 days; when it is long-term, it is 10 days to 30 days, which is convenient for subsequent calculations; Sorting the warning events according to the numerical values ​​of the release adaptability, obtaining the sorted warning events, identifying the corresponding parameter information and data format type based on the warning events, wherein the parameter information is parameter data obtained based on the characteristic information; based on the parameter information and the data format type, format conversion is performed on the warning events according to the protocol conversion rules, and the release data packets in a unified format are sorted and output; wherein the release data packets include: warning events, release levels, release areas, release times, and release frequencies; Release allocation module: Obtain the terminal information of the LED screen, and determine the priority of the release in combination with the release data packet to generate an information release plan; wherein the terminal information includes the information release queue and server bandwidth; The process of determining release priority is: First, determine whether there is an information release queue in the current LED screen. If there is an information release queue currently running, obtain the bandwidth requested at this time and determine whether it is in a saturated state. When the current bandwidth is in a saturated state, determine whether there is a new release request for access. When there is a release request for access, immediately analyze the request release data packet information, reallocate it according to the warning level of the warning event, and generate an information release plan; It should be noted that LED screens can be set up in high-traffic areas (such as traffic intersections, shopping malls, squares, etc.) to ensure that the information covers a wide range and reaches more people. When a warning event occurs, the LED screen can quickly issue warning information to remind people to take corresponding actions.

[0017] In summary, the system includes a data partitioning module, a model building module, an early warning cache module and a release distribution module. The key technical points are as follows: by dividing the collected multi-source information into normal data sets and abnormal data sets according to short-term, medium-term and long-term labels, and setting different time levels, judging the early warning times of several priorities under different time levels, and combining early warning signals and attention mechanisms, quickly and intelligently predicting early warning events, thus solving the problem of information delay; calculating the release level through feature information, designing a release frame, calculating the release adaptability based on the release frame information, and processing and generating release data packets to upload to the LED screen, thus improving the accuracy and detail of weather forecasts.

[0018] Embodiment 2: This embodiment provides an information release warning method based on LED screen. Figure 2 The figure is a flow chart showing a method for issuing information warning based on an LED screen according to an exemplary embodiment. Figure 2 , the information release warning method includes the following steps: Determine data types: Create normal data sets and abnormal data sets based on weather reports; Determine the event type: connect to the data source in real time through the LED screen, perceive the multi-source information collected by the data source in real time, input the multi-source information collected in real time into the preset event classification model, and output the first probability that the collected multi-source information belongs to the normal data set and the second probability of the abnormal data set; build a rule engine based on the first probability and the second probability, divide the warning signal, and generate several warning events corresponding to the priority levels; Formulate information release rules: extract feature information from warning events, build a release rule engine, generate release data packets, and cache the release data packets; Generate information release plan: obtain the terminal information of the LED screen, and determine the release priority based on the release data packet to generate an information release plan.

[0019] In the application, the several formulas involved are all calculated by removing dimensions and taking their numerical values, and the formula is a formula obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The formula is set by technical personnel in this field according to actual conditions.

[0020] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. A person of ordinary skill in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented in electronic hardware, or in a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution.

[0021] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0022] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An information release warning system based on LED screen, characterized in that: The system includes a data partitioning module, a model building module, an early warning cache module and a publishing and allocation module, and the data partitioning module, the model building module, the early warning cache module and the publishing and allocation module are communicatively connected; Data partitioning module, creating normal data set and abnormal data set according to weather reports; The model building module is connected to the data source in real time through the LED screen, senses the multi-source information collected by the data source in real time, inputs the multi-source information collected in real time into the preset event classification model, and outputs the first probability that the collected multi-source information belongs to the normal data set and the second probability that the collected multi-source information belongs to the abnormal data set; a rule engine is built based on the first probability and the second probability, the warning signal is divided, and several warning events corresponding to the priority are generated; wherein the collected multi-source information includes temperature, precipitation and wind speed; The warning cache module extracts feature information from warning events, builds a publishing rule engine, generates publishing data packets, and caches the publishing data packets; The publishing distribution module obtains the terminal information of the LED screen, determines the publishing priority based on the published data packet, and generates an information publishing plan; wherein the terminal information includes the information publishing queue and the server bandwidth.

2. According to claim 1, the information release warning system based on LED screen is characterized in that: The process of creating normal data sets and abnormal data sets based on weather reports is as follows: The twenty-four hours of each day are divided into several time intervals, and meteorological reports within the time intervals are actively collected. The time stage of the corresponding climate event is determined based on the meteorological report, the frequency characteristics of the climate event are obtained based on the time stage, and normal data sets and abnormal data sets are created based on the frequency characteristics; time tags are inserted into the normal data set and the abnormal data set respectively; among them, climate events are divided into normal events and warning events; time tags include short-term tags, medium-term tags and long-term tags.

3. The information release warning system based on LED screen according to claim 2 is characterized by: The process of obtaining the first probability and the second probability is: Construct an initial classification model. First, construct a four-dimensional monitoring vector through the collected multi-source information, including temperature parameters, precipitation parameters, wind speed parameters and time. For any time interval Ta, analyze the average and fluctuation values ​​of temperature parameters, precipitation parameters and wind speed parameters in the time interval Ta; generate new features based on the average and fluctuation values; where the fluctuation value represents the average value of the difference between the maximum and minimum values ​​of the corresponding parameters in the time interval Ta; Introducing a training layer into the initial classification model, combining new features with original features to form a new training data set, and providing randomly selected sample data from the collected multi-source information to train and update the training layer to obtain an event classification model; The acquired multi-source information is stratified according to time labels, and the corresponding multi-source information at each time level is input into the event classification model to determine the first probability that the input multi-source information belongs to the normal data set and the second probability that the input multi-source information belongs to the abnormal data set.

4. The information release warning system based on LED screen according to claim 3 is characterized by: When the first probability or the second probability is greater than the preset threshold, the collected multi-source information is correspondingly allocated to the corresponding data set; and the larger value of the first probability or the second probability is determined, and the collected multi-source information is correspondingly allocated to the data set corresponding to the larger value; when the first probability or the second probability is less than the preset threshold, reclassification is performed based on the first probability or the second probability, and the specific process is: When the first probability or the second probability is greater than the time tag of the normal data set or the abnormal data set, the collected multi-source information is correspondingly allocated to the corresponding data set; When the first probability or the second probability is smaller than the time tag of the normal data set or the abnormal data set, the collected multi-source information is continuously sensed.

5. The information release warning system based on LED screen according to claim 4 is characterized by: The process of classifying warning events is as follows: In chronological order, within a predetermined time period T, the time points corresponding to the collected multi-source information are marked on the timeline, and a time series analysis method is used in combination with a preset density threshold to identify the time period in which the density corresponding to the abnormal data set exceeds the density threshold; based on the analysis results of the time period, an attention mechanism is triggered, the analysis results are compared and sorted in descending order, a data list is displayed, and a corresponding warning signal is generated to divide warning events corresponding to several categories of priority.

6. The information release warning system based on LED screen according to claim 5 is characterized by: The extracted characteristic information includes time, type, range, trend, and medium method. The process of generating the release data package is as follows: Build a publishing rule engine to extract types, ranges, and trends from the extracted feature information and define them as key factors. Perform parameter analysis on the values ​​of each key factor to obtain the corresponding parameter values, and keep the parameter values ​​and weights synchronized. The process of parameter analysis is as follows: The value of the key factor is marked as Jb, and the standard interval of the key factor Jb is set to [r1, r2], and the parameter value is calculated based on the formula: ； In the formula, P is the parameter value corresponding to the key factor, σ represents a constant, which is usually 1; Then assign a corresponding weight to each key factor, multiply the parameter value of each factor by its weight, and add up all the results to get the release level; Divide the release level interval based on the release level, and set a first release box smaller than the release level interval, a second release box in the release level interval, and a third release box larger than the release level interval, respectively set corresponding selection interfaces and keyword input interfaces in the first release box, the second release box, and the third release box, select the corresponding warning event based on the release box information, extract the corresponding interface attribute information based on the release box information and perform preprocessing analysis to calculate the release adaptability of the warning event; The warning events are sorted according to the numerical values ​​of the release adaptability to obtain the sorted warning events, and the corresponding parameter information and data format type are identified based on the warning events. The parameter information is parameter data obtained based on the characteristic information. Based on the parameter information and the data format type, a protocol conversion rule is used to convert the format of the warning events, and a release data packet in a unified format is sorted and output; wherein the release data packet includes: warning event, release level, release area, release time and release frequency.

7. The information release warning system based on LED screen according to claim 6 is characterized by: The attribute information corresponding to the selection interface includes type, priority and event range, and the attribute information corresponding to the keyword interface includes content, relevance, frequency and context; the preprocessing analysis process is as follows: Construct a preprocessing analysis model: mark the input information as a parameter set Q, which contains N0 element values. Mark any element value as Xi, assign corresponding weight factor coefficients to the N0 element values, and the weight factor coefficient of the marked element value Xi is φi. Establish the formula to output the influence coefficient Qy of the parameter set Q: ； Apply the preprocessing analysis model: first integrate the attribute information corresponding to the selection interface and the keyword interface into the selection parameter Q1 and the keyword parameter Q2; where Q1 = {type, priority, event range}, Q2 = {content, relevance, frequency, context}; then input the selection parameters and keyword parameters into the preprocessing analysis model in sequence, and output the selection matching degree tc1 and the keyword matching degree tc2 respectively; The calculation formula for publishing fitness is: ； In the formula, Ad represents the degree of adaptation, s represents the warning level, is the time gain factor; ch represents the difference between the occurrence time of the warning event and the current time, and G represents the time constant.

8. The information release warning system based on LED screen according to claim 1 is characterized by: The process of determining release priority is: First, determine whether there is an information release queue in the current LED screen. If an information release queue is currently running, obtain the bandwidth requested at this time and determine whether it is in a saturated state. When the current bandwidth is in a saturated state, determine whether there is a new release request for access. When there is a release request for access, immediately analyze the request release data packet information, reallocate it according to the warning level of the warning event, and generate an information release plan.

9. An information release and early warning method based on LED screen, applied to the information release and early warning system according to any one of claims 1 to 8, characterized in that: The following steps are involved: Determine data types: Create normal data sets and abnormal data sets based on weather reports; Determine the event type: connect to the data source in real time through the LED screen, perceive the multi-source information collected by the data source in real time, input the multi-source information collected in real time into the preset event classification model, and output the first probability that the collected multi-source information belongs to the normal data set and the second probability of the abnormal data set; build a rule engine based on the first probability and the second probability, divide the warning signal, and generate several warning events corresponding to the priority levels; Formulate information release rules: extract feature information from warning events, build a release rule engine, generate release data packets, and cache the release data packets; Generate an information release plan: obtain the terminal information of the LED screen, and determine the release priority based on the release data packet to generate an information release plan; the terminal information includes the information release queue and server bandwidth.

Images