Artificial intelligence-based brucellosis spatiotemporal prediction method and system

Abstract

The present application relates to the technical field of space-time prediction, in particular to a brucellosis space-time prediction method and system based on artificial intelligence, comprising the following steps: acquiring case information to calculate correlation, extracting path sequence to screen trend direction, and matching new cases to generate prediction data. In the present application, the transmission correlation between cases is constructed and a spatial directionality index is introduced to realize the continuity recognition of the transmission path. The angle average value and variance are used to screen out non-trend transmission processes, thereby enhancing the spatial consistency of path screening. The spatio-temporal characteristics of new cases are matched with existing transmission trends to improve the sensitivity of the prediction to the trend attribution of new cases. The trend path is reversely projected and the regional positioning landing frequency is analyzed to enhance the spatial directionality of potential disease areas. Through the linkage processing of multi-level path extraction, trend judgment and regional coding, the prediction data can better capture the transmission and evolution characteristics of brucellosis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of spatiotemporal prediction, and in particular to a brucellosis spatiotemporal prediction method and system based on artificial intelligence. BACKGROUND

[0002] The technical field of spatiotemporal prediction involves modeling and analyzing the distribution patterns of specific events or phenomena in the time and space dimensions, including multi-source data collection, geographic information modeling, time series construction and analysis, prediction model construction and verification, etc., and is widely used in the fields of disease transmission, traffic flow, environmental pollution, etc. The brucellosis spatiotemporal prediction method refers to the prediction modeling technology for the transmission trend and distribution pattern of brucellosis, and is aimed at analyzing and predicting the transmission path, time node and geographical distribution characteristics of brucellosis between humans and animals. After the epidemiological survey data is geocoded, a prediction model is established using a statistical regression model combined with the confirmed time and geographic coordinate information in the historical incidence data to estimate the potential incidence area and time trend in the future.

[0003] The existing brucellosis spatiotemporal prediction technology mainly relies on statistical modeling of the spatiotemporal distribution characteristics of historical cases when dealing with the transmission trend of brucellosis. However, it is limited by the lack of analysis capability for the transmission sequence and direction characteristics between cases, making it difficult to accurately depict the evolution pattern of the transmission chain in the face of complex transmission paths and case aggregation fluctuations, resulting in the model's difficulty in path attribution or trend prediction when new cases occur. For example, in areas with frequent human flow or high concentration of cases, traditional models are prone to trend misjudgment and regional positioning ambiguity, which further affects the accurate identification of potential risk sections and resource allocation efficiency. SUMMARY

[0004] The purpose of the present application is to solve the problems existing in the prior art, and to provide a brucellosis spatiotemporal prediction method based on artificial intelligence.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme: a brucellosis spatiotemporal prediction method based on artificial intelligence, comprising the following steps:

[0006] S1: obtaining the time label and geographic coordinates of confirmed brucellosis cases, calculating the time interval and Euclidean distance between case pairs, labeling the case pairs that meet the brucellosis transmission correlation condition as transmission correlation units, and generating case correlation node records;

[0007] S2: based on the transmission correlation units established in the case correlation node records, extracting the time sequence between cases and calculating the spatial direction angle sequence, merging the continuous transmission path sequence, sorting all paths by starting point number, and generating a path sorting sequence index set;

[0008] S3: According to the path sorting sequence index set, the average value of the included angle of all direction vectors in the path and the variance of the included angle are calculated, the path that does not meet the direction consistency is screened out, the screening passed path number is marked, and a unified trend vector index is established, and a path trend index record is generated;

[0009] S4: Obtain the time label and geographic coordinates of the newly diagnosed cases, and jointly judge the direction angle of the path vector and the time interval of the path end node in the path trend index record, screen the trend adaptation path and record the corresponding number, and generate a case matching path record;

[0010] S5: According to the path number in the case matching path record, the direction line segment is extracted from the original path and the direction is reversed, the area coding mark is combined with the backtracking trajectory of the direction line segment, and the brucellosis space-time prediction data is obtained.

[0011] As a further scheme of the application, the case association node record includes a transmission relationship identifier, an associated case number, a spatial distance label, and a time interval label. The path sorting sequence index set includes a path starting point number, a transmission path length, a time sequence index, and a spatial direction sequence. The path trend index record includes a path number, a direction consistency score, a trend vector index, and a screening state identifier. The case matching path record includes a case number and an adaptation path number. The brucellosis space-time prediction data includes backtracking trajectory coordinates, regional positioning frequency data, spatial coding results, and predicted hot spot segments.

[0012] As a further scheme of the application, the case association node record acquisition step specifically comprises:

[0013] S111: Obtain the time label and geographic coordinates of the brucellosis diagnosed cases, and construct a case pair combination for all diagnosed cases. Calculate the time interval parameter of all case pairs by difference calculation of the time labels of the two cases in each case pair combination. Calculate the two-dimensional Euclidean distance value of the geographic coordinates of each group of cases, and generate a case pair time interval and spatial distance matrix;

[0014] S112: Based on the case pair time interval and spatial distance matrix, read the time interval parameter and spatial distance parameter respectively, and perform difference calculation and non-negativity judgment on the time interval parameter and the preset brucellosis transmission time threshold, and simultaneously perform difference calculation and non-negativity judgment on the spatial distance parameter and the brucellosis spatial threshold. Extract the index position of the case pair that simultaneously satisfies the two non-negativity judgment conditions, and generate a case pair transmission association mark set;

[0015] S113: According to the case pair index position recorded in the transmission correlation label set, the corresponding case pair data is extracted from the case pair combination, the unique number information of any two cases is aggregated and encoded, a time label and geographical coordinates are attached to construct a node structure, and a case correlation node record is generated.

[0016] As a further scheme of the application, the path order sequence index set obtaining step specifically comprises:

[0017] S211: Based on the node information in the case correlation node record, the time label fields of two nodes in each transmission unit are extracted, and the nodes are sorted in ascending order of diagnosis time. The geographical coordinate fields of the sorted node pairs are extracted as coordinate sequences in turn, and a space vector is constructed for each pair of adjacent nodes in the coordinate sequence. The angle value between the first and last nodes in a continuous three-point group is calculated, all angle values are mapped and encoded according to the transmission time sequence, and a direction angle sequence between nodes is generated.

[0018] S212: According to the angle value data in the direction angle sequence between nodes, the corresponding angle value of each three-node group is read and compared with the transmission path judgment threshold. If the angle value is less than the transmission path judgment threshold, the three nodes in the corresponding three-node group are path aggregated, and the nodes that continuously meet the condition are merged into a sequence segment to generate a continuous path aggregation node group.

[0019] S213: Based on the start node number in the continuous path aggregation node group, the path segments are preliminarily indexed and rearranged in ascending order of the number, and the path segment number mapping is re-established in combination with the start number sequence. All aggregated path segments arranged in order of number are generated to form a path order sequence index set.

[0020] As a further scheme of the application, the path trend index record obtaining step specifically comprises:

[0021] S311: Based on each path segment number in the path order sequence index set, the continuous node coordinate points in the path segment are extracted one by one to construct a space vector sequence. The direction angle between each two adjacent vectors is calculated, and all angle values are summarized. The arithmetic mean operation and variance operation are performed on the angle sequence to obtain the angle mean value and the angle variance corresponding to each path segment, and a path direction consistency parameter set is generated.

[0022] S312: According to the angle mean value and the angle variance in the path direction consistency parameter set, double judgment is performed on the angle mean value and the angle variance respectively in combination with the set judgment threshold. The path segment numbers that do not meet the condition are filtered out, the path segment number sequence that meets the condition is extracted, and a direction consistent path number set is established.

[0023] S313: Based on all path numbers in the direction consistent path number set, all direction vectors of corresponding path segments are extracted in turn, and single-axis uniform merging processing is performed on each path segment direction vector, and a one-to-one mapping is established with all path numbers to generate a path trend index record.

[0024] As a further scheme of the present application, the case matching path record acquisition step is specifically:

[0025] S411: Obtain the time label and geographical coordinates of the newly added confirmed cases, extract the diagnosis time stamp of each case and uniformly convert it into coordinated time format, extract the geographical position coordinates of the cases and construct them into a two-dimensional coordinate value group, construct an index structure for all case data, and arrange them in ascending order according to time sequence to establish a new case coordinate and time data set;

[0026] S412: According to the new case coordinate and time data set, the case coordinates and the trend vectors corresponding to each path segment in the path trend index record are extracted respectively to construct two-dimensional angle values, and the difference between the end node time stamp of each path segment and the case time label is calculated to screen the trend adaptation path and record the trend adaptation path number set;

[0027] S413: According to all path segment numbers in the trend adaptation path number set, a mapping structure is established for each new case number, the case number, matching path number, angle value, and time interval value are counted, and the case matching path record is obtained.

[0028] As a further scheme of the present application, the brucellosis spatiotemporal prediction data acquisition step is specifically:

[0029] S511: According to the path number recorded in the case matching path record, the corresponding path segment in the original path structure body is retrieved in turn, the end segment vector constituting the direction in the path segment is extracted, the starting node and end node coordinates are recorded, the direction vector is constructed and the reverse mapping operation is performed to generate a unit reverse vector, and the path reverse direction vector set is obtained;

[0030] S512: Read each reverse vector in the path reverse direction vector set, perform discrete stepping in the reverse direction along the corresponding vector starting coordinate within the two-dimensional region boundary, calculate whether the current landing point is located inside the region boundary, if the landing point is legal, record the point as a landing point trajectory, repeat the stepping operation until it is out of the boundary, record all landing point coordinates in each reverse path, calculate the region landing point offset metric value, and generate a landing point trajectory distribution value set;

[0031] S513: According to each landing point coordinate in the landing point trajectory distribution value set, the corresponding coding area is located in the map grid index, each landing point is regionally coded and accumulated according to the block of the coordinate, all path segments are traversed, and brucellosis space-time prediction data is established.

[0032] As a further scheme of the application, the calculation formula of the regional landing point offset measure value is:

[0033] ;

[0034] Wherein, represents the first landing point coordinate component, represents the first landing point coordinate component, is the average value of the horizontal and vertical coordinates of all landing points of the path segment, is the total number of landing points corresponding to a single path segment, represents the first landing point coordinate component, represents the first landing point coordinate component.

[0035] The brucellosis space-time prediction system based on artificial intelligence comprises:

[0036] The case correlation construction module is used to perform S1: obtaining the time label and geographical coordinates of the confirmed cases of brucellosis, calculating the time interval and Euclidean distance between the case pairs, marking the case pairs meeting the brucellosis transmission correlation condition as transmission correlation units, and generating case correlation node records;

[0037] The path sequence recombination module is used to perform S2: based on the transmission correlation units established in the case correlation node records, extracting the time sequence and calculating the spatial direction angle sequence between the cases, merging the continuous transmission path sequence, sorting all paths according to the starting point number, generating a path sorting sequence index set, and generating a path sorting sequence index set;

[0038] The trend path screening module is used to perform S3: according to each path in the path sorting sequence index set, calculating the average value and variance of the angle of all direction vectors in the path, screening out the paths that do not meet the direction consistency, marking the screening through path number and establishing a unified trend vector index, generating a path trend index record;

[0039] The case path discrimination module is used to perform S4: obtaining the time label and geographical coordinates of the newly added confirmed cases, jointly judging the direction angle of the path vector and the time interval of the path end node in the path trend index record, screening the trend adaptation path and recording the corresponding number, and generating a case matching path record;

[0040] The space-time prediction backtracking module is used for performing S5: according to the path number in the case matching path record, extracting a directional line segment in the original path and performing directional reverse projection, combining the backtracking trajectory of the directional line segment to perform area coding marking, and obtaining brucellosis space-time prediction data.

[0041] Compared with the prior art, the advantages and positive effects of the present application are that:

[0042] In the present application, by constructing the transmission correlation between cases and introducing a spatial directionality index, the continuity of the transmission path is recognized, the non-trend transmission process is screened out through the angle average and variance, the spatial consistency of path screening is enhanced, the sensitivity of the prediction to the trend attribution of the newly occurring case is improved through the joint matching of the space-time characteristics of the new case and the existing transmission trend, the spatial directionality to the potential disease area is enhanced through the reverse projection of the trend path and the analysis of the regional positioning drop point frequency, and the prediction data is improved in the ability to capture the transmission and evolution characteristics of brucellosis through the linkage processing of multi-level path extraction, trend judgment and regional coding. BRIEF DESCRIPTION OF DRAWINGS

[0043] Figure 1 The present application is a main step flowchart;

[0044] Figure 2 The present application is a case correlation node record acquisition flowchart;

[0045] Figure 3 The present application is a path sorting sequence index set acquisition flowchart;

[0046] Figure 4 The present application is a path trend index record acquisition flowchart;

[0047] Figure 5 The present application is a case matching path record acquisition flowchart;

[0048] Figure 6 The present application is a brucellosis space-time prediction data acquisition flowchart. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application is further described in detail below in combination with the drawings and examples. It should be understood that the specific examples described herein are only used to explain the present application and do not limit the present application.

[0050] In the description of the present application, it should be understood that the terms "length", "width", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, in the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise specifically limited.

[0051] Please refer to Figure 1 The artificial intelligence-based brucellosis spatiotemporal prediction method comprises the following steps:

[0052] S1: Obtain the time label and geographic coordinates of confirmed brucellosis cases, calculate the time interval and Euclidean distance between the case pairs, and determine whether the case pairs meet the time threshold (72 hours) and spatial threshold (25 kilometers) of brucellosis transmission correlation condition. The case pairs that meet the condition are labeled as transmission correlation units, and a case correlation node record is generated;

[0053] S2: Based on the transmission correlation units established in the case correlation node record, extract the time sequence between the cases and calculate the spatial direction angle sequence. Determine whether the direction angle of the first and last nodes in any three-point group is less than the judgment threshold. If it meets the condition, it is merged into a continuous transmission path sequence. All paths that meet the condition are sorted by starting point number, and a path sorting sequence index set is generated;

[0054] S3: According to each path in the path sorting sequence index set, calculate the average value and variance of the angle of all direction vectors in the path. Combined with the screening standard (the average value of the angle is less than 30° and the variance is less than 20), the path that does not meet the direction consistency is screened out. Mark the path number that passes the screening, establish a unified trend vector index, and generate a path trend index record;

[0055] S4: Obtain the time label and geographic coordinates of the newly diagnosed cases. According to the direction angle between the new case coordinate point and each path vector in the path trend index record and the time interval of the path end node, jointly judge the path segment with an angle less than 30° and a time interval less than 36 hours as a trend adaptation path. Record the corresponding number of the adaptation path, and generate a case matching path record;

[0056] S5: According to the path number in the case matching path record, extract the direction line segment in the original path and perform direction reverse projection. Combined with the positioning drop point frequency of the backtracking trajectory of the direction line segment in the regional boundary, and marked with regional code, the brucellosis spatiotemporal prediction data is obtained.

[0057] The case association node record includes a propagation relationship identifier, an associated case number, a spatial distance label, and a time interval label. The path order sequence index set includes a path starting point number, a propagation path length, a time sequence index, and a spatial direction sequence. The path trend index record includes a path number, a direction consistency score, a trend vector index, and a screening state identifier. The case matching path record includes a case number and an adaptive path number. The brucellosis spatiotemporal prediction data includes a backtracking trajectory coordinate, a regional positioning frequency data, a spatial encoding result, and a predicted hot spot section.

[0058] Please refer to Figure 2 , S1 step is:

[0059] S111: Obtain the time label and geographic coordinates of the brucellosis confirmed cases, and construct a case pair combination for all confirmed cases. Calculate the difference between the time labels of the two cases in each case pair combination, extract the time interval parameters of all case pairs, calculate the two-dimensional Euclidean distance value of the geographic coordinates of each group of cases, and generate a case pair time interval and spatial distance matrix;

[0060] To obtain the time label and geographic coordinates of the brucellosis confirmed cases, the unique number of each case must be extracted first, and the diagnosis time and geographic location information must be retrieved in the case database. The time label is converted into a time stamp in the format of "year-month-day hour: minute: second", and the geographic coordinates are represented by the longitude and latitude values in the WGS-84 coordinate system, for example, the diagnosis time of case A is 2024-07-15 14:30:00, and the longitude and latitude are 34.2356°N, 108.9527°E. The diagnosis time of case B is 2024-07-17 10:15:00, and the longitude and latitude are 34.2400°N, 108.9600°E. Then, the case numbers are recorded as 001 and 002, respectively. All case numbers are combined to form a case pair set. When the total number of cases is N, the number of combinations is When N=4, 6 case pair combinations are formed, which are {001, 002}, {001, 003}, {001, 004}, {002, 003}, {002, 004}, and {003, 004}. Then, the time labels of each case pair are processed by difference, and the time interval is calculated. The operation steps are to subtract the time stamp of the previous case from the time stamp of the next case, and the unit is unified as hours. Taking cases A and B as an example, the time difference is (2024-07-17 10:15:00)-(2024-07-15 14:30:00), which is 43 hours and 45 minutes, and after conversion, it is 43.75 hours. Next, the two-dimensional Euclidean distance of the geographic coordinates of each case pair is calculated. Let the coordinates of point A be , and the coordinates of point B be The spherical distance formula is used:

[0061] ;

[0062] wherein km is the average radius of the earth, is the latitude of the two points (in radians), is the longitude of the two points, calculated as follows using the example data above:

[0063] rad;

[0064] rad;

[0065] rad;

[0066] rad;

[0067] Substituting into the formula gives:

[0068] ;

[0069] The time interval and spatial distance results are stored in a two-dimensional matrix, with the matrix rows and columns corresponding to the case pair combination index, and the elements being the time interval value and the Euclidean distance value of the corresponding case pair, to obtain the following example matrix:

[0070] Table 1 Case pair time interval and spatial distance table

[0071]

[0072] As shown in Table 1, the time interval and spatial distance between different case pairs differ significantly and are uniformly distributed between 0-70 hours and 0-25 kilometers, providing basic parameter support for subsequent screening of case transmission correlations, and generating a case pair time interval and spatial distance matrix.

[0073] S112: Based on the case pair time interval and spatial distance matrix, the time interval parameter and the spatial distance parameter are read respectively, the difference calculation and non-negativity judgment are performed on the time interval parameter and the preset brucellosis transmission time threshold, and the difference calculation and non-negativity judgment are performed on the spatial distance parameter and the brucellosis spatial threshold, the case pair index position that simultaneously satisfies the two non-negativity judgment conditions is extracted, and a case pair transmission correlation marker set is generated;

[0074] Based on each set of data in the time interval and spatial distance matrix of case pairs, the time interval field and spatial distance field need to be read line by line, and the time interval value and spatial distance value of each case pair combination are called respectively, and the difference is judged with the preset threshold value. The time propagation threshold is 72 hours, and the spatial propagation threshold is 25 kilometers. Both thresholds are derived from the historical transmission path analysis results of brucellosis epidemiology investigation, based on the transmission incubation period and host activity radius. The specific value can be set according to the following empirical data: the maximum interval of brucellosis transmission from one case to another does not exceed 3 days, which is converted to 72 hours according to 24 hours / day; and the spatial transmission range is mostly concentrated within 25 kilometers of the source of the epidemic. It has been verified that the correlation of case pairs above this range decreases significantly, so the spatial transmission threshold is set to 25 kilometers. Taking case pair 001-004 in Table 1 as an example, the time interval is 67.25 hours, and the Euclidean distance is 18.45 kilometers. The difference is calculated with the transmission time threshold 72 hours and the spatial threshold 25 kilometers respectively:

[0075] ;

[0076] It indicates that the case pair meets the transmission space-time range condition; otherwise, for case pair 003-004, if the time interval is 75 hours and the Euclidean distance is 26 kilometers, then:

[0077] ;

[0078] It does not meet the double judgment conditions and should be excluded. In actual operation, all case pair combinations are traversed, and the time judgment function and the spatial judgment function are set as follows:

[0079] ;

[0080] The final transmission correlation judgment is , only when , it indicates that the case pair is a transmission correlation pair, and its index position in the case pair combination and the corresponding case number are recorded. Referring to the data in Table 1, the final case pair index set that meets the conditions is the first five groups in {001-002, 001-003, 001-004, 002-003, 002-004, 003-004}, numbered as index [1, 2, 3, 4, 5]. Based on this set, the case pair number is extracted and its index value in the original data table is recorded to generate the case pair transmission correlation marker set.

[0081] S113: According to the case pair index position recorded in the case pair propagation association mark set, extract the corresponding case pair data from the case pair combination, and perform aggregation coding operation on the unique number information of any two cases. Attach time label and geographic coordinates to build node structure body, and generate case association node record;

[0082] According to the case pair index position recorded in the case pair propagation association mark set, read the numbers of the corresponding two cases from the original case pair combination one by one, and extract their respective time labels and geographic coordinates at the same time. Then, the time labels in the two records are generated by string splicing to generate a case pair propagation time node, which is used to represent the start and end time points of the case pair propagation. The expression is "case A diagnosis time-case B diagnosis time", for example, the diagnosis time of case 001 is 2024-07-15 14:30:00, and the diagnosis time of case 003 is 2024-07-16 11:45:00, then the time node is "2024-07-15 14:30:00-2024-07-16 11:45:00", the time span of the corresponding case pair is 21 hours and 15 minutes. This field will be one of the key time description fields of the node record, at the same time, the geographic coordinates of the two cases are constructed into a two-dimensional array, arranged in time order. If the coordinates of case 001 are latitude 34.2356 and longitude 108.9527, and the coordinates of case 003 are latitude 34.2480 and longitude 108.9580, then the constructed coordinate array is [[34.2356, 108.9527], [34.2480, 108.9580]]. This coordinate information is used for subsequent propagation path visualization processing. In the node record structure, the case pair number, time node and coordinate array are respectively configured as the three attribute fields of the record entry. Each record is indexed by the case pair number, and the time node and coordinate array are additional attributes. This further ensures the integrity and uniformity of the data structure in the propagation chain construction process. Then, all the case pair structure records that meet the conditions are merged and written into the record table. The record table structure defines four fields: case pair number, time node, start point coordinate and end point coordinate, so as to facilitate subsequent column retrieval and network graph construction. The record table sample is as follows:

[0083] Table 2 Case Association Node Record Table

[0084]

[0085] As shown in Table 2, the structure record clearly represents the propagation start and end information and geographic path between cases in a standardized field format. After the record table is constructed, the entire table structure needs to be serialized to convert the data structure into a JSON string or CSV structure text, so as to realize cross-platform data calling and visualization operation deployment. After the data structure is encapsulated, the case association node record is generated.

[0086] Referring to Figure 3 , the S2 step is:

[0087] S211: Based on the node information in the case-associated node record, the time tag field of two nodes in each propagation unit is extracted, and the nodes are sorted in ascending order of diagnosis time. The geographic coordinate field of the sorted node pair is extracted as a coordinate sequence, and a space vector is constructed for each pair of coordinates between two adjacent nodes in the coordinate sequence. The angle value between the first and last nodes in a continuous three-point group is calculated. All angle values are mapped and encoded according to the propagation time sequence to generate a direction angle sequence between nodes;

[0088] Based on the propagation unit data structure in the case-associated node record, the diagnosis time tag field in each node pair is first extracted, and the year, month, day, hour, minute, and second information of the time string is used as the basis for sorting. The time sequence is rearranged according to the time sequence, and the time stamp is expressed to the second level to avoid the situation that the same case group has the same time and causes the sorting to fail. Then, the geographic coordinate values of adjacent nodes are extracted one by one in the sorted node sequence. Let the node A coordinate be , the node B coordinate be , and the node C coordinate be , respectively construct the vector , , and perform vector angle calculation for each three-node group. The vector dot product calculation method is used to construct the angle formula:

[0089] ;

[0090] Take node A, B, and C as [34.2300, 108.9500], [34.2400, 108.9600], and [34.2500, 108.9700] respectively for example calculation:

[0091] ;

[0092] ;

[0093] The dot product is ;

[0094] The lengths are , ;

[0095] Substituting into the formula gives:

[0096] ;

[0097] represents the three-point direction consistency, the included angle is 0 degrees, and the included angle value is stored in the included angle sequence. The complete included angle sequence is constructed by taking all three-node combinations as traversal units. Each group of included angle values is indexed according to the time label mapping. Thus, the direction included angle sequence between nodes is obtained. Among them, represents the spatial displacement vector formed by any two consecutive nodes, represents the direction turning angle between the first and last nodes formed by three points, in radians. The included angle value is used for subsequent direction consistency judgment and reflects whether the path has obvious turning changes.

[0098] S212: According to the included angle value data in the direction included angle sequence between nodes, the included angle value corresponding to each group of three nodes is read and compared with the propagation path judgment threshold. If the included angle value is less than the propagation path judgment threshold, the three nodes in the corresponding three-node group are subjected to path aggregation processing, and the nodes that continuously satisfy the conditions are merged into a sequence segment to generate a continuous path aggregation node group.

[0099] According to the three-node group index corresponding to each included angle value in the direction included angle sequence between nodes, the included angle value is called, and the set judgment threshold is set. radians, derived from the reasonable direction turning limit angle range in the Brucellosis transmission path. The threshold is set based on the statistical analysis of the typical path turning angle distribution in the case diffusion network. The upper limit of the angle interval covered by 85% of the case path direction changes is selected to ensure the direction coherence and executability of the aggregated path. The comparison operation is to judge whether each is less than , for example, when a certain included angle is radians, the aggregation condition is met. Otherwise, if radians, the condition is not met. Then, for the three-point group that meets the condition, the three-node sequence is aggregated into a single path segment to construct a node aggregation group. If multiple groups continuously meet the included angle condition, the last node of the path segment and the first node of the next segment are further connected and aggregated to construct a continuous transmission path. The original time sequence and spatial path structure are preserved during the aggregation process. The node group number range before and after merging is recorded and a new path segment coding rule is generated. Finally, the path segments composed of all node groups that meet the angle judgment standard are output, and the continuous path aggregation node group is obtained.

[0100] S213: Based on the start node number in the continuous path aggregation node group, the path segments are preliminarily indexed and rearranged in ascending order of the number, and the path segment number mapping is re-established in combination with the start number sequence. All aggregated path segments arranged in order of number are generated to form a path sorting sequence index set.

[0101] Based on the path segment information in each group of continuous path aggregation nodes, the case number of the first node is extracted as the path segment index basis, sorted in ascending order, and the sorting operation is performed in ascending order with the numerical part of the string number as the key. If the path segment starting number is 003, 008, 002, 011, the sorted order is 002, 003, 008, 011. After sorting, the path segment sequence is renumbered, and an index mapping structure is established, where the key-value form is {path segment number: starting node number}, and the mapping structure is saved in the structure table. The structure table has the functions of path segment number retrievability and path relationship traceability, which is used to support subsequent path visualization construction and propagation trajectory tracking process. The table structure includes fields: path segment number, path sequence starting point number, path segment length, and part of the structure is as follows:

[0102] Table 3 Path segment sorting structure table

[0103]

[0104] As shown in Table 3, the path segments are rearranged and coded according to the starting point number, and the numbering index system of the propagation path is established, generating the path sorting sequence index set.

[0105] Please refer to Figure 4 , the S3 step is:

[0106] S311: Based on each path segment number in the path sorting sequence index set, the continuous node coordinate points in the path segment are extracted one by one to construct a spatial vector sequence. The direction included angle between each two adjacent vectors is calculated, and all the included angle values are summarized. The arithmetic average operation and the variance operation are performed on the included angle sequence to obtain the average value and the variance of the included angle corresponding to each path segment, and the path direction consistency parameter set is generated.

[0107] Based on each path segment number in the path sorting sequence index set, the geographic coordinates of all adjacent nodes in the path segment are extracted in turn, and a two-dimensional coordinate difference vector is constructed with the latitude and longitude values. The included angle between each pair of continuous vectors is calculated to obtain the direction turning angle value, forming the direction change sequence in the path segment. Then two operations are performed on the direction change sequence. First, the arithmetic average is calculated. After accumulating all the included angle values, the included angle number is divided to obtain the average direction. Then the variance is calculated. The square difference value of the included angle and the average value is accumulated and then divided by the total number. Let the included angle values in a path be 、 、 、 、 , the average value is:

[0108] ;

[0109] The corresponding variance is:

[0110] ;

[0111] All path segments perform the above calculation operation in turn, store the calculation result associated with the path segment number, form the direction mean and dispersion matrix record, and construct the structured parameter set for the screening module to call, establish the path direction consistency parameter set.

[0112] S312: According to the angle mean and angle variance in the path direction consistency parameter set, double judgment is carried out respectively combined with the angle mean and angle variance and the set judgment threshold, the path segment number that does not meet the condition is screened out, the path segment number sequence that meets the condition is extracted, and the direction consistent path number set is established;

[0113] According to the mean and variance recorded in the path direction consistency parameter set, the angle mean and the angle variance of each group are extracted, and double condition judgment is carried out with the set screening reference value in turn, and the screening conditions are:

[0114] ;

[0115] Among them radian, radian, the screening logic adopts the Boolean product judgment method, and only when the two conditions are met at the same time can the path number enter the screening result set, for example, path segment The corresponding parameters are , , which meet the conditions, and path segment The corresponding parameters are , , only the first condition is not met, then path segment should be removed, after traversing the entire parameter set, record all the path segment numbers that pass the judgment as a sequence array, and output as a number list, finally get the direction consistent path number set.

[0116] S313: Based on all path numbers in the direction consistent path number set, and sequentially extracting the direction vector of all vectors in the corresponding path segment, performing single-axis unified merging processing on each path segment direction vector, and establishing one-to-one mapping with all path numbers, generating path trend index record;

[0117] Based on each path segment number in the direction consistent path number set, extract all space vectors in the corresponding path segment, take the direction angle of each vector as the input parameter, perform trigonometric function decomposition on the vectors in the same path segment according to the angle direction, decompose into , , and perform weighted merging on all components in each path segment, the path segment total vector is:

[0118] ;

[0119] The unit direction vector after merging is:

[0120] ;

[0121] For example, in a certain path segment , the corresponding component is:

[0122] 、 、 ;

[0123] 、 、 ;

[0124] Then:

[0125] ;

[0126] After normalization:

[0127] ;

[0128] Assemble the path segment number and direction vector into a mapping index entry, establish a structured index table, and generate a path trend index record.

[0129] Table 4 Path Trend Index Table

[0130]

[0131] As shown in Table 4, the different path segment trend vector structure clearly records the direction distribution results, providing input basis for trend comparison.

[0132] Please refer to Figure 5 , the S4 step is:

[0133] S411: Obtain the time label and geographic coordinates of the newly diagnosed cases, extract the diagnosis time stamp of each case and convert it to coordinated time format, extract the geographic coordinates of the case and construct a two-dimensional coordinate value group, construct a numbered index structure for all case data, and arrange them in ascending order according to time sequence, and establish a dataset of new case coordinates and time data;

[0134] Obtain the time label and geographic coordinates of the newly diagnosed cases, extract the diagnosis time information of each case data and convert it to timestamp format, set the time stamp precision to second level and record it uniformly in Beijing time, set the case Z time as October 14, 2024 8:30:15, the corresponding timestamp is 1697243415, and extract the geographic coordinates of case Z as , and record items are constructed with number A, and then all case records are constructed as a structure array with fields including case number, timestamp, horizontal coordinate, and vertical coordinate. The structure array is arranged in ascending order with timestamp as the key. If the input records are Z, B, C, and D, the corresponding timestamps are 1697243415, 1697230000, 1697251000, and 1697225000, and after sorting, they are D, B, Z, and C. Further, an index mapping dictionary is generated for the sorted structure array and stored in a data frame. In this structure, a one-to-one correspondence between coordinates and time is established, which is used for subsequent spatial matching and time difference judgment of the input data source, and finally, a new case coordinate and time dataset is generated.

[0135] S412: According to the new case coordinate and time dataset, the trend vector corresponding to each path segment in the path trend index record is constructed to extract the two-dimensional angle value, the difference between the end node timestamp of each path segment and the case time label is calculated, the trend adaptation path is selected, and the trend adaptation path number set is recorded.

[0136] According to the coordinate information recorded in the new case coordinate and time dataset, the case coordinates are extracted one by one and unit vectors are constructed. The vector angle calculation is performed with the direction vector of each path segment in the path trend index record. Let the new case vector be , the direction vector of the path segment be , and the angle be calculated by the cosine formula . After unitization, the inner product result is directly used as the cosine value of the angle to participate in the inverse cosine function operation. If the calculated angle between a case and a path segment is radians, converted to degrees, it is , which meets the threshold limit condition. Further, the time difference between the case time label and the path segment end node timestamp is calculated. Let the case time be 1697243415 and the end node time be 1697235415, then the time interval is 9000 seconds, which is converted to 2.5 hours. The judgment threshold is set to 36 hours, which is converted to 129600 seconds. If the angle is less than and the time interval is less than 129600 seconds, the path segment number is recorded in the matching sequence. After the two-way matching determination of all cases and path segments is completed, all path segment numbers that meet the conditions are accumulated in sequence to obtain the trend adaptation path number set.

[0137] S413: According to all path segment numbers in the trend adaptation path number set, a mapping structure is established for each new case number, the case number, matching path number, angle value, and time interval value are counted, and the case matching path record is obtained.

[0138] Based on the trend fitting path number set, the case information matched by each numbered path is matched and recorded in sequence. The record content includes case number, corresponding path number, direction angle value, and time interval value. After arranging in ascending order according to the case number, a structure list is constructed and formatted into a data table form. The fields are case number, path number, angle value (unit °), and time interval (unit h). Taking part of the record as an example, if the case number is C002, the corresponding path number is P015, the angle value is 27.3, and the time interval is 18.2 hours, then the record behavior [C002, P015, 27.3, 18.2] is constructed. After sequentially constructing all record entries, a structured data frame is formed, and output in table form. Finally, the case matching path record is obtained.

[0139] Table 5 Case matching path record table

[0140]

[0141] As shown in Table 5, the matching path record clearly shows the direction and time matching between the case and the path segment, which is used for subsequent path updating processing.

[0142] Please refer to Figure 6 , the S5 step is:

[0143] S511: According to the path number recorded in the case matching path record, the corresponding path segment in the original path structure is retrieved in sequence. The end segment vector constituting the direction in the path segment is extracted, the starting node and end node coordinates are recorded, the direction vector is constructed and the reverse mapping operation is performed to generate the unit reverse vector, and the path reverse direction vector set is obtained.

[0144] According to the path number in the case matching path record, the corresponding numbered path segment is retrieved from the original path structure set, and its end direction segment is obtained. By reading the coordinate values of the starting node and the terminal node of the segment, the two-dimensional vector representation is calculated , then the vector is taken in the opposite direction to construct , and the unit vector normalization operation is performed to obtain , for example, if the coordinates of a path end segment are , , the original vector is , the unit reverse vector is , and after all path segments are completed, the reverse direction vector array is constructed, where each item structure contains path number, direction vector component, and unitization result. In actual application, this array can be used to simulate the reverse expansion trend field of the Brucella spread direction, providing a spatial vector basis for the next step of drop point trajectory deduction, and finally generating the path reverse direction vector set.

[0145] S512: Read each reverse vector in the reverse direction vector set, perform discrete stepping in the reverse direction within the boundary of the two-dimensional region according to the starting coordinates of the corresponding vector, calculate whether the current landing point is inside the region boundary, if the landing point is legal, record the point as a landing point trajectory, repeat the stepping operation until it is out of the boundary, record all landing point coordinates in each reverse path, and use the formula:

[0146] ;

[0147] Calculate the region landing point offset metric value, generate the landing point trajectory distribution value set; wherein, represents the th landing point coordinate component, is the average value of the horizontal and vertical coordinates of all landing points in the path segment, is the total number of landing points corresponding to a single path segment, represents the normalized offset value of the th path segment;

[0148] Read each unit direction vector in the reverse direction vector set, get its corresponding starting position coordinates, and then use the fixed-step continuous projection strategy to perform discrete stepping, with each step distance set to , sequentially generate coordinate sequence points by adding along the unit vector direction, and form the path segment backtracking trajectory. In each projection, calculate whether the current coordinate point is inside the region boundary. Set the boundary latitude and longitude range as , . If the current point satisfies and , it is considered as a legal landing point and is included in the statistics. Otherwise, the path segment is terminated. In this process, each path is projected at most to a total distance . For each path segment, organize and store the landing point set according to the path number to form a two-dimensional array structure landing point set , and calculate the dispersion degree of the landing point sequence. Use the average value of the horizontal and vertical coordinates of each landing point in the path to perform absolute difference weighted summation. When the path number is P007 and the landing point coordinates are as follows , the average value is . Replace it with the calculation of each difference value and then normalize. Replace it with the formula to obtain the normalized offset value . Perform this calculation on all path segments to finally generate the landing point trajectory distribution value set.

[0149] The operation logic of the formula is to measure the dispersion degree of the backtracking landing point of a certain path segment in space by normalizing the absolute offset, where the numerator represents the th landing point and the average point of all landing points of the path segment The cumulative linear deviation amount in two directions is calculated by taking the absolute value, so that the positive and negative deviations are both included in the total amount; the denominator part The position amplitude of the landing point itself in the region space is measured, and the normalized factor is formed by combining the horizontal coordinate mean square of all landing points, the purpose being to eliminate the offset amplification caused by inconsistent space scales, so that the offset degrees of different path segments in the same interval are comparable, and the summation symbol is used to accumulate the relative dispersion of all landing points, and finally output a dimensionless normalized index representing the trend of the landing point set of the path segment , and the smaller the value, the more concentrated the trajectory landing points. The entire operation logic embodies the relative measurement between space deviation accumulation and scale reconciliation, thereby realizing the quantifiable characterization of landing point distribution density.

[0150] The region landing offset measurement value is used to represent the concentration degree of all dispersed landing points in the space region during the reverse trajectory deduction of a path segment, and this value reflects the deviation degree of all landing points relative to the mean position of the landing points. In practical applications, if this value tends to zero, it indicates that the landing points are highly concentrated during the path segment reverse deduction, and the spatial dispersion is small, indicating that the potential propagation direction has strong directionality and regional stability in geographical space. Conversely, if the value is large, it indicates that the reverse trajectory landing points present a strong dispersion trend and cannot form an effective regional aggregation signal, indicating that the reverse propagation trend of the path segment lacks spatial convergence. Therefore, this measurement value can be used as an important quantitative basis for the effectiveness of the path trend and the rationality of the target region positioning.

[0151] S513: According to the landing point trajectory distribution value of each landing point coordinate, the belonging coded region is located in the map grid index, and each landing point is regionally coded and accumulated according to the coordinate belonging block. All path segments are traversed to establish the Brucellosis spatiotemporal prediction data;

[0152] According to the coordinates of each point belonging to the region code, the geographical space is divided into 0.01 degree grid units according to the latitude and longitude region index mechanism, and the block number of the single point belonging to the block number is calculated= , where N is the number of longitude direction blocks. The frequency of each path landing point belonging to the region is accumulated in the landing point set. If the number of landing points in the unit with region code a142 is 6, and the frequency threshold is set to , it is considered to meet the marking requirements. The binding relationship between the path segment number and the region code is established, and a one-to-many index structure is established between all path segments meeting the conditions and the region number. The path number, landing point region code, and landing point frequency are recorded. An encoding matrix is formed to represent the spatial propagation tendency, and finally the Brucellosis spatiotemporal prediction data is obtained.

[0153] Table 6 Brucellosis path landing frequency table

[0154]

[0155] As shown in Table 6, the region code forms a one-to-one or one-to-many matching relationship with the path number, which provides basic data support for determining the spatial transmission trend of brucellosis.

[0156] The artificial intelligence-based brucellosis space-time prediction system comprises:

[0157] The case correlation construction module is used to perform S1: obtaining the time label and geographic coordinates of the confirmed cases of brucellosis, calculating the time interval and Euclidean distance between the case pairs, marking the case pairs that meet the brucellosis transmission correlation conditions as transmission correlation units, and generating case correlation node records;

[0158] The path sequence recombination module is used to perform S2: based on the transmission correlation units established in the case correlation node records, extracting the time sequence between each case and calculating the spatial direction angle sequence, merging the continuous transmission path sequence, sorting all paths according to the starting point number, generating a path sorting sequence index set;

[0159] The trend path screening module is used to perform S3: according to each path in the path sorting sequence index set, calculating the average value and variance of the angle of all direction vectors in the path, screening out the paths that do not meet the direction consistency, marking the screening through path number and establishing a unified trend vector index, and generating a path trend index record;

[0160] The case path discrimination module is used to perform S4: obtaining the time label and geographic coordinates of the newly diagnosed cases, jointly judging the direction angle of the path vector and the time interval of the path end node in the path trend index record, screening the trend adaptation path and recording the corresponding number, and generating a case matching path record;

[0161] The space-time prediction backtracking module is used to perform S5: according to the path number in the case matching path record, extracting the direction line segment in the original path and performing direction reverse projection, combining the backtracking trajectory of the direction line segment to mark the region code, and obtaining the brucellosis space-time prediction data.

[0162] The above is only a preferred embodiment of the present application, and does not limit the form of the present application, and any skilled person in the art can use the disclosed technical content to make changes or modifications to equivalent embodiments applied to other fields, but any simple modification, equivalent change and modification made according to the technical essence of the present application to the above embodiments within the technical solution content of the present application still belongs to the protection scope of the present application.

Claims

1. A spatiotemporal prediction method for brucellosis based on artificial intelligence, characterized in that, Includes the following steps: S1: Obtain the time label and geographic coordinates of confirmed brucellosis cases, calculate the time interval and Euclidean distance between case pairs, mark case pairs that meet the conditions for brucellosis transmission association as transmission association units, and generate case association node records; S2: Based on the established transmission association units in the case association node records, extract the temporal sequence between each case and calculate the spatial direction angle sequence, merge the continuous transmission path sequences, sort all paths by starting point number, and generate a path sorting sequence index set; S3: Based on each path in the path sorting sequence index set, calculate the average angle and variance of the angles of all direction vectors within the path, filter out paths that do not conform to the direction consistency, mark the filtered path numbers and establish a unified trend vector index to generate path trend index records; S4: Obtain the time label and geographical coordinates of newly confirmed cases, and make a joint judgment with the direction angle of the path vector and the time interval of the end node of the path in the path trend index record, filter the trend-fitting path and record the corresponding number, and generate a case matching path record; S5: Based on the path number in the case matching path record, extract the direction line segment from the original path and perform directional reverse projection. Combine the back-trajectory of the direction line segment to perform regional coding and marking to obtain the spatiotemporal prediction data of brucellosis.

2. The artificial intelligence-based spatiotemporal prediction method for brucellosis according to claim 1, characterized in that, The case association node record includes a transmission relationship identifier, associated case number, spatial distance label, and time interval label. The path sorting sequence index set includes a path start number, transmission path length, time order index, and spatial direction sequence. The path trend index record includes a path number, directional consistency score, trend vector index, and filtering status identifier. The case matching path record includes a case number and an adapted path number. The brucellosis spatiotemporal prediction data includes backtracking trajectory coordinates, regional positioning frequency data, spatial coding results, and predicted hotspot segments.

3. The artificial intelligence-based spatiotemporal prediction method for brucellosis according to claim 1, characterized in that, The specific steps for obtaining the case-related node records are as follows: S111: Obtain the time tags and geographic coordinates of confirmed brucellosis cases, construct case pair combinations for all confirmed cases, calculate the difference between the time tags of the two cases in each case pair combination, extract the time interval parameters of all case pairs, calculate the two-dimensional Euclidean distance value for the geographic coordinates of each group of cases, and generate a case pair time interval and spatial distance matrix. S112: Based on the case pair time interval and spatial distance matrix, read the time interval parameter and spatial distance parameter respectively, calculate the difference between the time interval parameter and the preset brucellosis transmission time threshold and perform non-negativity judgment, and at the same time calculate the difference between the spatial distance parameter and the brucellosis spatial threshold and perform non-negativity judgment, extract the index position of the case pair that simultaneously meets the two non-negativity judgment conditions, and generate a case pair transmission association tag set; S113: Based on the case pair index position recorded in the case pair transmission association marker set, extract the corresponding case pair data from the case pair combination, perform aggregation encoding operation on the unique number information of any two cases, attach time tags and geographic coordinates to construct a node structure, and generate case association node records.

4. The artificial intelligence-based spatiotemporal prediction method for brucellosis according to claim 1, characterized in that, The specific steps for obtaining the path sorting sequence index set are as follows: S211: Based on the node information in the case-related node record, extract the time tag field of two nodes in each transmission unit, sort the nodes in ascending order of diagnosis time, extract the geographic coordinate field of the sorted node pairs into a coordinate sequence, construct a spatial vector for the coordinate pairs between each pair of adjacent nodes in the coordinate sequence, calculate the angle between the first and last nodes in a group of three consecutive points, map and encode all angle values ​​according to the transmission time sequence, and generate a sequence of directional angles between nodes; S212: Based on the angle value data in the direction angle sequence between nodes, read the angle value corresponding to each group of three nodes and compare it with the propagation path judgment threshold. If the angle value is less than the propagation path judgment threshold, the three nodes in the corresponding three-node group will be processed for path aggregation, and the node pairs that continuously meet the conditions will be merged into a sequence segment to generate a continuous path aggregation node group. S213: Based on the starting node number in the continuous path aggregation node group, the path segments are rearranged in ascending order of the number, and the path segment number mapping is re-established in combination with the starting number sequence. All aggregated path segments arranged in number order are used to generate a path sorting sequence index set.

5. The artificial intelligence-based spatiotemporal prediction method for brucellosis according to claim 1, characterized in that, The specific steps for obtaining the path trend index record are as follows: S311: Based on the path segment number in the path sorting sequence index set, extract the coordinate points of continuous nodes in the path segment one by one and construct a spatial vector sequence. Calculate the directional angle between each pair of adjacent vectors and summarize all angle values. Perform arithmetic mean and variance operations on the angle sequence to obtain the average angle and variance of the angle corresponding to each path segment, and generate a path direction consistency parameter set. S312: Based on the mean and variance of the included angle in the path direction consistency parameter set, perform a double judgment by combining the mean and variance of the included angle with the set judgment threshold, filter out the path segment numbers that do not meet the conditions, extract the sequence of path segment numbers that meet the conditions, and establish a path number set with consistent direction. S313: Based on all path numbers in the set of path numbers with consistent direction, extract the direction vectors of all vectors in the corresponding path segments in sequence, perform single-axis unified merging processing on the direction vectors of each path segment, establish a one-to-one mapping with all path numbers, and generate path trend index records.

6. The artificial intelligence-based spatiotemporal prediction method for brucellosis according to claim 1, characterized in that, The specific steps for obtaining the case matching path record are as follows: S411: Obtain the time label and geographic coordinates of newly confirmed cases, extract the confirmation timestamp of each case and convert it into the Coordinated Time Format, extract the geographic coordinates of the cases and construct a two-dimensional coordinate value group, construct a numbered index structure for all case data and sort them in ascending order according to time, and establish a dataset of coordinates and time of newly confirmed cases. S412: Based on the newly added case coordinates and time dataset, extract the case coordinates and the corresponding trend vectors of each path segment in the path trend index record to construct a two-dimensional angle value. Perform difference calculation on the end node timestamp and case time label of each path segment, filter trend-adapted paths, and record and generate a set of trend-adapted path numbers. S413: Based on all path segment numbers in the trend-adapted path number set, establish a mapping structure corresponding to the newly added case number, and count the case number, matching path number, angle value, and time interval value to obtain the case matching path record.

7. The artificial intelligence-based spatiotemporal prediction method for brucellosis according to claim 1, characterized in that, The specific steps for acquiring spatiotemporal prediction data for brucellosis are as follows: S511: Based on the path number recorded in the case matching path record, sequentially retrieve the corresponding path segment in the original path structure, extract the end line segment vector that constitutes the direction in the path segment, record the coordinates of the starting node and the end node, construct the direction vector and perform a reverse mapping operation to generate a unit reverse vector, and obtain the path reverse direction vector set. S512: Read each reverse vector in the reverse direction vector set of the path, perform discrete stepping in the reverse direction within the two-dimensional region boundary according to the starting coordinate of the corresponding vector, calculate whether the current landing point is inside the region boundary, if the landing point is legal, record the point as a landing point trajectory, repeat the stepping operation until it leaves the boundary, record all landing point coordinates in each reverse path, calculate and obtain the region landing point offset metric value, and generate a landing point trajectory distribution value set; S513: Based on the coordinates of each landing point in the landing point trajectory distribution value set, locate the corresponding coding area in the map grid index, accumulate the regional codes of each landing point according to the block to which the coordinates belong, traverse all path segments, and establish spatiotemporal prediction data for brucellosis.

8. The artificial intelligence-based spatiotemporal prediction method for brucellosis according to claim 7, characterized in that, The formula for calculating the regional landing point offset metric is as follows: ； in, , This represents the coordinate component of the i-th landing point. is the average of the x and y coordinates of all landing points on the path segment, where n is the total number of landing points for a single path segment. This represents the normalized offset value of the j-th path segment.

9. An artificial intelligence-based spatiotemporal prediction system for brucellosis, characterized in that, The system is used to implement the artificial intelligence-based spatiotemporal prediction method for brucellosis as described in any one of claims 1-8, comprising: The case association construction module is used to execute S1: obtain the time label and geographic coordinates of confirmed brucellosis cases, calculate the time interval and Euclidean distance between case pairs, mark case pairs that meet the conditions for brucellosis transmission association as transmission association units, and generate case association node records; The path sequence recombination module is used to execute S2: based on the established transmission association units in the case association node records, extract the temporal sequence between each case and calculate the spatial direction angle sequence, merge the continuous transmission path sequences, sort all paths by starting point number, and generate a path sorting sequence index set; The trend path screening module is used to perform S3: according to each path in the path sorting sequence index set, calculate the average angle and variance of the included angle of all direction vectors in the path, screen out paths that do not conform to the direction consistency, mark the path numbers that pass the screening and establish a unified trend vector index, and generate path trend index records; The case path identification module is used to execute S4: obtain the time label and geographical coordinates of newly confirmed cases, and make a joint judgment with the direction angle of the path vector in the path trend index record and the time interval of the end node of the path, filter the trend-adapted path and record the corresponding number, and generate a case matching path record; The spatiotemporal prediction backtracking module is used to execute S5: based on the path number in the case matching path record, extract the direction line segment in the original path and perform directional back projection, combine the backtracking trajectory of the direction line segment to perform regional coding and marking, and obtain spatiotemporal prediction data for brucellosis.

Images