Tunnel traffic flow big data mining method and system combined with time sequence

Abstract

The invention provides a tunnel traffic flow big data mining method and system combined with a time series, and relates to the technical field of traffic flow data analys.The tunnel traffic flow big data mining method comprises the steps that firstly, a time series data set corresponding to tunnel traffic flow big data is obtained, and the time series data set comprises a plurality of traffic flow time series fragments collected under the continuous time dimension; constructing time sequence association chains based on a time sequence adjacent relation, performing feature filling on adjacent fragment time slots to obtain a time sequence association chain set, and performing time sequence evolution feature extraction processing on each time sequence association chain to obtain a time sequence evolution feature set; and calling a pre-trained time sequence mode mining model to carry out abnormal evolution mode matching processing on the time sequence evolution characteristic set, generating an abnormal mode mining result, finally generating a traffic flow regulation and control suggestion set based on the abnormal mode mining result, and analyzing and adjusting regulation and control contents in combination with an influence range, so that the regulation and control suggestions have more pertinence and operability. The efficiency and accuracy of tunnel traffic flow regulation and control can be effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of traffic flow data analysis, in particular to a tunnel traffic flow big data mining method and system combined with time series. BACKGROUND

[0002] In tunnel traffic management, accurate analysis and mining of traffic flow data are directly related to the safety and efficiency of tunnel traffic. Currently, most methods for processing tunnel traffic flow data focus on the analysis of static data or only consider simple time period division, ignoring the continuous change of traffic flow over time.

[0003] The prior art usually divides a day into several fixed time periods and respectively counts the traffic flow in each time period. The above processing method cannot capture the dynamic changes of traffic flow in the continuous time dimension. For example, during the morning and evening peak hours, traffic flow will fluctuate rapidly over time, and the flow changes between different time periods may have complex correlations, but traditional methods are difficult to reveal these correlations. Moreover, when traffic anomalies such as traffic accidents and congestion occur, the evolution process of these anomalies in the time series cannot be effectively captured and analyzed. This results in that when facing complex changes in tunnel traffic flow, the existing data analysis methods cannot provide comprehensive and accurate information, and cannot provide effective decision basis for traffic flow regulation, thereby affecting the management level of tunnel traffic. SUMMARY

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present application, the embodiments of the present application provide a tunnel traffic flow big data mining method combined with time series, which comprises: obtaining a time series data set corresponding to tunnel traffic flow big data, the time series data set containing a plurality of traffic flow time series segments collected in a continuous time dimension, each traffic flow time series segment recording the traffic flow change information of a specific monitoring area in the tunnel within a corresponding time interval; constructing a time series correlation chain based on the time series adjacent relationship between the traffic flow time series segments, performing feature filling processing on the time gap of adjacent segments during the construction process to obtain a time series correlation chain set, the time series correlation chain being used to represent the correlation relationship of traffic flow change information in different time intervals, and each time series correlation chain containing a node sequence and an association strength parameter between nodes; performing time series evolution feature extraction processing on each time series correlation chain, respectively extracting features from the time dimension and the correlation dimension and performing interactive fusion to obtain a time series evolution feature set corresponding to each time series correlation chain, the time series evolution feature set containing the change trend feature and the correlation strength change feature of the time series correlation chain in the time dimension; The pre-trained time sequence pattern mining model is called to perform abnormal evolution pattern matching processing on each of the time sequence evolution feature sets, to generate an abnormal pattern mining result, which contains an abnormal evolution pattern type matched and identification information of the time sequence correlation chain; A traffic flow regulation suggestion set is generated based on the abnormal pattern mining result, and the regulation content is adjusted in combination with an influence range of the abnormal pattern, each suggestion in the traffic flow regulation suggestion set corresponds to one abnormal evolution pattern type, and each suggestion contains regulation direction information and regulation implementation step information for the corresponding abnormal evolution pattern.

[0005] In another aspect, the embodiment of the present application also provides a tunnel traffic flow big data mining system combined with time sequence, which comprises a processor and a machine readable storage medium, the machine readable storage medium is connected with the processor, the machine readable storage medium is used for storing programs, instructions or codes, and the processor is used for executing the programs, instructions or codes in the machine readable storage medium to realize the above method.

[0006] Based on the above aspects, by obtaining a time sequence data set containing a plurality of traffic flow time sequence segments in a continuous time dimension, the traffic flow change information of a specific monitoring area in a tunnel in different time intervals is recorded, a time sequence correlation chain is constructed based on time sequence adjacency relationship, and feature filling processing is performed on the time gap of adjacent segments, effectively solving the gap problem that may exist in the time sequence data, the constructed time sequence correlation chain set can accurately represent the correlation relationship of the traffic flow change information in different time intervals, time sequence evolution feature extraction processing is performed on each time sequence correlation chain, features are extracted from the time dimension and the correlation dimension respectively and are interactively fused, the obtained time sequence evolution feature set can comprehensively reflect the change trend feature and the correlation strength change feature of the time sequence correlation chain in the time dimension, the pre-trained time sequence pattern mining model is called to perform abnormal evolution pattern matching processing, which can quickly and accurately generate an abnormal pattern mining result, providing an effective means for timely discovering traffic flow abnormalities. Finally, a traffic flow regulation suggestion set is generated based on the abnormal pattern mining result, and the regulation content is adjusted in combination with the influence range, so that the regulation suggestion is more targeted and operable, which can effectively improve the efficiency and accuracy of tunnel traffic flow regulation. BRIEF DESCRIPTION OF DRAWINGS

[0007] Figure 1 is an execution flow diagram of the tunnel traffic flow big data mining method combined with time sequence provided by the embodiment of the present application.

[0008] Figure 2 is a schematic diagram of exemplary hardware and software components of the tunnel traffic flow big data mining system combined with time sequence provided by the embodiment of the present application. DETAILED DESCRIPTION

[0009] The application will be described in detail below with reference to the accompanying drawings, Figure 1 is a flowchart of a tunnel traffic flow big data mining method combined with time series provided by an embodiment of the application. The tunnel traffic flow big data mining method combined with time series will be described in detail below.

[0010] Step S110: Obtain a time series data set corresponding to tunnel traffic flow big data, the time series data set containing a plurality of traffic flow time series segments collected in continuous time dimensions, each traffic flow time series segment recording traffic flow change information of a specific monitoring area in a tunnel in a corresponding time interval.

[0011] In this embodiment, taking traffic flow monitoring of a tunnel group in a city as an example, the tunnel management system can collect traffic flow data of each monitoring area according to a preset time interval (for example, every 5 minutes). The monitoring area here can be different road segments in the tunnel, such as an entrance segment, a middle driving segment, an exit segment, etc. Each traffic flow time series segment corresponds to the flow change of a monitoring area in a time interval, for example, for the monitoring area A of the tunnel entrance segment, in the time interval from 8:00 to 8:05, the traffic flow gradually increases from the initial 0 to a certain value, and the flow change information in this process constitutes a traffic flow time series segment. Thus, all the required traffic flow data can be retrieved from the database, and arranged into a time series data set. Each element in the time series data set is a traffic flow time series segment, containing monitoring area identifier, time interval, flow change sequence, etc.

[0012] Step S120: Construct a time series correlation chain based on the time series adjacent relationship between the traffic flow time series segments, and perform feature filling processing on the time gap of adjacent segments in the construction process to obtain a time series correlation chain set, the time series correlation chain being used to represent the correlation relationship of traffic flow change information in different time intervals, and each time series correlation chain containing a node sequence and an association strength parameter between nodes.

[0013] Step S121: Extract time interval information of all traffic flow time series segments from the time series data set, the time interval information containing the start time and end time of each traffic flow time series segment.

[0014] In this embodiment, for the time series data set of the tunnel group, each traffic flow time series segment is traversed to extract its start time and end time. For example, one time series segment of monitoring area A has a start time of 8:00 and an end time of 8:05; another time series segment has a start time of 8:05 and an end time of 8:10; and another time series segment of monitoring area B has a start time of 8:03 and an end time of 8:08. The above time interval information is arranged into a list for subsequent analysis.

[0015] Step S122: determining the time series adjacent relationship between any two traffic flow time series segments according to the time interval information, the time series adjacent relationship being used to indicate whether the time intervals of two traffic flow time series segments have a continuous connection relationship, and if there is an interval between the end time of a previous traffic flow time series segment and the start time of a subsequent traffic flow time series segment, it is determined that there is a time series gap between the two traffic flow time series segments.

[0016] In this embodiment, the extracted time interval information is compared pairwise. For example, the 8:00-8:05 segment and the 8:05-8:10 segment of monitoring area A, the end time 8:05 of the former and the start time 8:05 of the latter completely connect, and there is no time series gap; and the 8:05-8:10 segment of monitoring area A and the 8:03-8:08 segment of monitoring area B, the start time 8:05 of the former and the end time 8:08 of the latter, and the end time 8:10 of the former and the start time 8:03 of the latter do not directly connect. Here, between the 8:05-8:10 segment of monitoring area A and the 8:03-8:08 segment of monitoring area B, if the next possible segment of monitoring area A is considered, assuming that there is a segment of monitoring area C with a start time of 8:11, then between the 8:05-8:10 segment of monitoring area A and the 8:11-8:16 segment of monitoring area C, there is a 1-minute interval between the end time 8:10 of the former and the start time 8:11 of the latter, and it is determined that there is a time series gap between the two segments.

[0017] Step S123: for two traffic flow time series segments with a time series gap, using the traffic flow change information of adjacent segments to perform feature interpolation processing to generate a transition feature sequence that fills the time series gap, the time length of the transition feature sequence being consistent with the time length of the time series gap.

[0018] In this embodiment, for the 8:05-8:10 segment of the monitoring area A and the 8:11-8:16 segment of the monitoring area C, the time gap is 1 minute (i.e. 60 seconds, assuming the time unit is second, the time length is 60). First, extract the traffic flow change information of the 8:05-8:10 segment of the monitoring area A, such as the flow a1 at 8:05, a2 at 8:06,..., a6 at 8:10; the traffic flow change information of the 8:11-8:16 segment of the monitoring area C, c1 at 8:11, c2 at 8:12,..., c6 at 8:16. Using the interpolation method, according to the flow values of a6 and c1 and the length of the time gap, the transition feature sequence is generated. For example, assuming the time gap is 1 minute, i.e. one time step (here it is assumed that the time step is 1 minute), then the time length of the transition feature sequence is 1, and its flow value can be obtained by linear interpolation of a6 and c1, such as the flow value of the transition feature sequence in the time interval of 8:10-8:11 is (a6+c1) / 2 (of course other interpolation methods such as polynomial interpolation can also be used, according to the actual demand), thus obtaining the transition feature sequence, and the time length is consistent with the time length of the time gap.

[0019] Step S124: For the two traffic flow time sequence segments with the time sequence adjacent relationship, calculate the similarity degree of the traffic flow change information in the two traffic flow time sequence segments, and take the similarity degree as the time sequence correlation strength between the two traffic flow time sequence segments.

[0020] Step S1241: Extract the traffic flow change information in the two traffic flow time sequence segments with the time sequence adjacent relationship, and if there is a transition feature sequence, combine the transition feature sequence with the traffic flow change information of the subsequent traffic flow time sequence segment to obtain the first traffic flow change sequence and the subsequent traffic flow change sequence respectively.

[0021] In this embodiment, the 8:00-8:05 segment and the 8:05-8:10 segment of the monitoring area A are taken as examples, the two segments have a time sequence adjacent relationship, and there is no transition feature sequence. The traffic flow change information of the 8:00-8:05 segment of the monitoring area A is extracted, assuming that the time step of the segment is 1 minute, then the flow change sequence is [q1, q2, q3, q4, q5] (q1 is the flow at 8:00, q2 is the flow at 8:01, and so on); the traffic flow change sequence of the 8:05-8:10 segment is [q6, q7, q8, q9, q10]. The former is taken as the first traffic flow change sequence, and the latter is taken as the subsequent traffic flow change sequence. For example, the 8:05-8:10 segment of the monitoring area A and the 8:11-8:16 segment of the monitoring area C, a transition feature sequence [q11] is generated through step S123 (assuming that the time step is 1 minute), then the flow change sequence [q6, q7, q8, q9, q10] of the 8:05-8:10 segment of the monitoring area A is taken as the first traffic flow change sequence, the transition feature sequence [q11] is combined with the flow change sequence [c1, c2, c3, c4, c5] of the 8:11-8:16 segment of the monitoring area C to obtain the subsequent traffic flow change sequence [q11, c1, c2, c3, c4, c5].

[0022] Step S1242: Perform length unification processing in the time dimension on the first traffic flow change sequence and the subsequent traffic flow change sequence, and calculate the traffic flow value difference at the corresponding position for each time step based on the first traffic flow change sequence and the subsequent traffic flow change sequence after length unification, to obtain a difference sequence composed of multiple time step difference values.

[0023] In this embodiment, for the 8:00-8:05 segment (sequence length is 5) and the 8:05-8:10 segment (sequence length is 5) of the monitoring area A described above, the lengths of the two are the same, and no length unification processing is required. The difference is calculated for each time step, assuming that the length of the first sequence is extended to 6 by padding, such as adding a value identical to the last element at the end, or by interpolation. Then the difference is calculated for each time step, such as the first sequence after extension [q1, q2, q3, q4, q5, q5], the subsequent sequence [q11, c1, c2, c3, c4, c5], then each element of the difference sequence is q11-q1, c1-q2, c2-q3, c3-q4, c4-q5, c5-q5. In this way, the difference sequence is obtained.

[0024] Step S1243: Calculate the overall fluctuation degree of the difference sequence according to the traffic flow value difference of all time steps, and represent the overall fluctuation degree by statistical distribution of all difference values in the difference sequence.

[0025] In this embodiment, the distribution of all difference values in the obtained difference sequence is counted. The variance or standard deviation of the difference values can be calculated. The smaller the variance, the more concentrated the difference values are, and the lower the overall fluctuation degree is. The larger the variance, the higher the fluctuation degree is. For example, the difference sequence is [d1, d2, d3, d4, d5, d6], the sum of squares of the difference between each difference value and the mean value is calculated, and then divided by the number of difference values (here the number is 6) to obtain the variance. The overall fluctuation degree is represented by the size of the variance.

[0026] Step S1244: determining a basic value of the similarity degree based on the overall fluctuation degree. The lower the overall fluctuation degree is, the higher the basic value of the similarity degree is.

[0027] In this embodiment, a mapping relationship between the fluctuation degree and the basic value of the similarity degree is preset, such as when the variance is 0, the basic value of the similarity degree is 1; when the variance is v1, the basic value of the similarity degree is 0.8; when the variance is v2, the basic value of the similarity degree is 0.5, and so on (here v1 and v2 are threshold values set according to actual conditions). According to the calculated variance of the difference sequence, the corresponding basic value of the similarity degree is found. For example, the calculated variance is v0, if v0 is less than v1, the basic value of the similarity degree is 1; if v1≤v0

[0028] Step S1245: correcting the basic value of the similarity degree in combination with the position correlation degree of the monitoring areas corresponding to the two traffic flow time sequence segments. If the monitoring areas corresponding to the two traffic flow time sequence segments are adjacent, the basic value of the similarity degree is increased, and if the monitoring areas corresponding to the two traffic flow time sequence segments are not adjacent, the basic value of the similarity degree is kept unchanged.

[0029] In this embodiment, the position correlation degrees of the monitoring areas in the tunnel are defined in advance, such as the position correlation degree of adjacent monitoring areas (such as the monitoring areas of the tunnel entrance section and the middle driving section) is 1, the position correlation degree of one monitoring area apart is 0.5, and the position correlation degree of two or more monitoring areas apart is 0. For the monitoring areas corresponding to the two traffic flow time sequence segments, the position correlation degrees are checked. For example, the monitoring area A (entrance section) and the monitoring area B (middle driving section) are adjacent, and the position correlation degree is 1, so the basic value of the similarity degree is increased by a certain percentage, such as 0.1; if the monitoring area A and the monitoring area C (exit section) are not adjacent, the position correlation degree is 0, and the basic value of the similarity degree is kept unchanged.

[0030] Step S1246: taking the corrected similarity degree as the time sequence correlation strength between the two traffic flow time sequence segments.

[0031] In this embodiment, the similarity degree basis value obtained in step S1244 is modified in step S1245, and the obtained value is the time sequence correlation strength between the two traffic flow time sequence segments. For example, if the similarity degree basis value is 0.8, the position correlation degree is 1, and the modified value is increased by 0.1, then the time sequence correlation strength is 0.9.

[0032] Step S125: dynamically calibrating the time sequence correlation strength currently calculated based on the correlation strength distribution of the same type of segments in the historical traffic flow data, and eliminating the influence of abnormal values on the correlation strength.

[0033] In this embodiment, the correlation strength data of the same type of segments (the same type refers to the same type of monitoring area and the same type of time interval, such as the entrance segment monitoring area segments of the morning peak period) as the current two traffic flow time sequence segments are retrieved from the historical traffic flow database, the distribution of these correlation strengths is counted, the mean and standard deviation are calculated. Then compare the currently calculated time sequence correlation strength with the mean and standard deviation, if the difference between the current correlation strength and the mean exceeds 3 times the standard deviation, it is considered to be an abnormal value, and it is calibrated. The calibration method can be to adjust it to the mean plus or minus 3 times the standard deviation (according to the deviation direction), or use other statistical methods for correction, to eliminate the influence of abnormal values.

[0034] Step S126: filtering out the traffic flow time sequence segment pairs whose time sequence correlation strength meets the threshold requirement according to a preset correlation strength threshold, to obtain an effective correlation pair set, each correlation pair in the effective correlation pair set includes two traffic flow time sequence segments and the calibrated time sequence correlation strength.

[0035] In this embodiment, a correlation strength threshold is preset, for example, 0.6. All the calculated and calibrated time sequence correlation strengths are traversed, and the traffic flow time sequence segment pairs whose time sequence correlation strength is greater than or equal to 0.6 are filtered out to form an effective correlation pair set. For example, the time sequence correlation strength of the 8:00-8:05 segment and the 8:05-8:10 segment of monitoring area A is 0.9, which is greater than 0.6, and is filtered out; the time sequence correlation strength of some two segments of monitoring area B is 0.5, which is less than 0.6, and is not filtered out. Each element in the effective correlation pair set includes the identification of the two traffic flow time sequence segments and the calibrated time sequence correlation strength.

[0036] Step S127: taking one traffic flow time sequence segment in each effective correlation pair as a starting node and the other traffic flow time sequence segment as a subsequent node, connecting the multiple effective correlation pairs in turn according to the order of time intervals, if a node is repeated in the connection process, the connection path with higher correlation strength is retained, and an initial time sequence correlation chain is constructed.

[0037] In this embodiment, for each association pair in the set of effective association pairs, the time sequence thereof is determined, with the earlier time segment as the starting node and the later time segment as the subsequent node. Then these association pairs are connected in turn to form a chain structure. For example, effective association pair 1: segment A (8:00-8:05) and segment B (8:05-8:10), association strength 0.9; effective association pair 2: segment B (8:05-8:10) and segment C (8:10-8:15), association strength 0.8; effective association pair 3: segment A (8:00-8:05) and segment D (8:03-8:08), association strength 0.7. When connecting, segment A connects segment B, segment B connects segment C to form a chain; if segment A connects segment D, since segment A has been used as the starting node in the previous chain and the time interval 8:03-8:08 of segment D overlaps with the time interval 8:00-8:05 of segment A, node repetition (segment A repetition) occurs. At this time, the association strength of the two paths is compared, and the association strength total (0.9+0.8=1.7) of segment A-segment B-segment C is greater than the association strength (0.7) of segment A-segment D, so the path of segment A-segment B-segment C is retained to construct the initial time sequence association chain.

[0038] Step S128: performing a redundant node elimination process on the initial time sequence association chain, deleting a node in the initial time sequence association chain whose traffic flow change information has a repetition degree exceeding a preset repetition degree threshold with respect to the traffic flow change information of adjacent nodes, and updating the time sequence association strength between the nodes before and after the deleted node.

[0039] For example, step S1281: starting from the starting node of the initial time sequence association chain, each node is selected in turn as a current node, and a subsequent node adjacent to the current node is determined, which is the node immediately following the current node in time in the time sequence association chain.

[0040] In this embodiment, for the initial time sequence association chain constructed above, for example, the chain is segment A (8:00-8:05)-segment B (8:05-8:10)-segment C (8:10-8:15)-segment D (8:15-8:20), starting from the starting node segment A, the current node is segment A and the subsequent node is segment B; then the current node is segment B and the subsequent node is segment C; the current node is segment C and the subsequent node is segment D; the current node is segment D and there is no subsequent node (assuming that the chain ends here).

[0041] Step S1282: extracting the traffic flow change information of the current node and the traffic flow change information of the subsequent node, which includes the traffic flow value and the traffic flow change trend of each time step.

[0042] In this embodiment, taking the current node as segment B (8:05-8:10) and the subsequent node as segment C (8:10-8:15) as an example, the traffic flow change information of segment B is extracted, including the traffic flow value at 8:05, the traffic flow value at 8:06,..., the traffic flow value at 8:10, and the traffic flow change trend at each time step (for example, rising, falling, or stable); similarly, the traffic flow change information of segment C is extracted, including the traffic flow value at 8:10, the traffic flow value at 8:11,..., the traffic flow value at 8:15, and the corresponding change trend.

[0043] Step S1283: The two kinds of traffic flow change information are compared time step by time step, the proportion of the number of time steps with the same traffic flow value in the total number of time steps is calculated, and the proportion of the number of time steps with the same traffic flow change trend in the total number of time steps is calculated.

[0044] In this embodiment, it is assumed that the time steps of segment B are 5 (8:05-8:10, one time step per minute), and the time steps of segment C are also 5 (8:10-8:15, one time step per minute). The traffic flow values at 8:05 (segment B) and 8:10 (segment C), the traffic flow values at 8:06 (segment B) and 8:11 (segment C), and so on are compared time step by time step. It is assumed that the number of time steps with the same traffic flow value is 3, so the proportion of the same traffic flow value is 3 / 5; similarly, it is assumed that the number of time steps with the same traffic flow change trend is 4, so the proportion of the same traffic flow change trend is 4 / 5.

[0045] Step S1284: Calculate the repetition degree comprehensive value, add the proportion of the same traffic flow value and the proportion of the same traffic flow change trend according to the preset weight, and obtain the traffic flow change information repetition degree of the current node and the subsequent node.

[0046] In this embodiment, the weight of the proportion of the same traffic flow value is 0.6, and the weight of the proportion of the same traffic flow change trend is 0.4. Then the repetition degree comprehensive value = 0.6x(3 / 5) + 0.4x(4 / 5) = 0.6x0.6 + 0.4x0.8 = 0.36 + 0.32 = 0.68.

[0047] Step S1285: Compare the calculated repetition degree with the preset repetition degree threshold value, if the repetition degree exceeds the preset repetition degree threshold value, mark the subsequent node as a redundant node.

[0048] In this embodiment, the preset repetition degree threshold value is 0.7. Since the calculated repetition degree is 0.68, which is less than 0.7, the subsequent node segment C is not marked as a redundant node. For example, if the calculated repetition degree is 0.75, which exceeds 0.7, the subsequent node is marked as a redundant node.

[0049] Step S1286: If the redundancy does not exceed the preset redundancy threshold, the subsequent node is reserved, and the next node is judged.

[0050] In this embodiment, since the redundancy 0.68 of the segment B and the segment C does not exceed the threshold 0.7, the segment C is reserved, and the next node segment C is judged to determine the subsequent node segment D, and the operations of the steps S1281-S1285 are repeated.

[0051] Step S1287: All nodes in the initial time sequence correlation chain are traversed, and after the marking of all redundant nodes is completed, all marked redundant nodes are deleted.

[0052] In this embodiment, all nodes in the initial time sequence correlation chain are traversed, and the redundancy of each node and the subsequent node is calculated and marked. Assuming that in another chain, a subsequent node is marked as a redundant node, after the traversal is completed, all marked redundant nodes are deleted from the initial time sequence correlation chain. For example, the initial time sequence correlation chain is segment E-segment F-segment G-segment H, wherein the redundancy of the segment F and the segment G exceeds the threshold, the segment G is marked as a redundant node, and then the segment G is deleted, and the chain becomes segment E-segment F-segment H.

[0053] Step S1288: The adjacent relationship of the remaining nodes after the redundant nodes are deleted is determined, and if the deleted redundant node is located between the previous reserved node and the subsequent reserved node, the previous reserved node and the subsequent reserved node become new adjacent nodes.

[0054] In this embodiment, for the chain segment E-segment F-segment H after the segment G is deleted, since the segment G is deleted, the subsequent node of the segment F is the segment G, and the subsequent node of the segment G is the segment H, and now the segment G is deleted, so the previous reserved node segment F and the subsequent reserved node segment H become new adjacent nodes.

[0055] Step S1289: The time sequence correlation strength between the previous reserved node and the subsequent reserved node is calculated, and the new time sequence correlation strength is obtained by combining the correlation strength of the previous reserved node and the original redundant node, the correlation strength of the original redundant node and the subsequent reserved node, and the similarity of the traffic flow change information of the previous reserved node and the subsequent reserved node.

[0056] In this embodiment, it is assumed that the association strength of the preceding reserved node segment F and the original redundant node segment G is s1, the association strength of the original redundant node segment G and the subsequent reserved node segment H is s2, and the similarity degree of the traffic flow change information of the preceding reserved node segment F and the subsequent reserved node segment H is s3. The new time sequence association strength can be calculated by a certain formula, such as s_new=(s1+s2)×s3 / 2 (the formula here can be adjusted according to actual conditions, such as considering weights, etc.). For example, s1=0.8, s2=0.7, and s3=0.6, then s_new=(0.8+0.7)×0.6 / 2=1.5×0.6 / 2=0.45.

[0057] Step S12810: re-connect the remaining nodes after deleting the redundant nodes according to the original time interval order, and add a new time sequence association strength to form an optimized time sequence association chain.

[0058] In this embodiment, the remaining nodes after deleting the redundant nodes are rearranged according to the time interval order, such as segment E-segment F-segment H, and then the new time sequence association strength is added between adjacent nodes, such as the association strength between segment E and segment F remains unchanged (assuming not deleted), and the association strength between segment F and segment H is s_new=0.45 calculated, thereby forming an optimized time sequence association chain.

[0059] Step S129: integrate all the optimized time sequence association chains, record the node number, total time span, and average association strength of each time sequence association chain, and form a time sequence association chain set.

[0060] In this embodiment, all the optimized time sequence association chains are collected, and statistics are performed on each chain. For example, an optimized time sequence association chain contains 3 nodes, the time interval is from 8:00-8:05, 8:05-8:10, and 8:10-8:15, the total time span is 15 minutes (from 8:00 to 8:15), and the association strength between nodes is 0.9, 0.8, respectively, and the average association strength is (0.9+0.8) / 2=0.85. The above information is recorded, together with other optimized time sequence association chains, to form a time sequence association chain set.

[0061] Step S130: perform time sequence evolution feature extraction processing on each time sequence association chain, extract features from time dimension and association dimension respectively, and interactively fuse to obtain a time sequence evolution feature set corresponding to each time sequence association chain, and the time sequence evolution feature set contains the change trend feature and the association strength change feature of the time sequence association chain in the time dimension.

[0062] Step S131: dividing each of the time sequence correlation chains into a plurality of continuous evolution stages in the order of time intervals, each of the evolution stages containing at least two adjacent traffic flow time sequence segment nodes, and the time span of each of the evolution stages being consistent.

[0063] In this embodiment, each of the time sequence correlation chains in the set of time sequence correlation chains is divided in the order of time intervals. For example, the time intervals of a time sequence correlation chain are 8:00-8:05, 8:05-8:10, 8:10-8:15, and 8:15-8:20, and the total time span is 20 minutes. The time sequence correlation chain is divided into two evolution stages, each of which has a time span of 10 minutes. The first evolution stage contains the first two nodes (8:00-8:05 and 8:05-8:10), and the second evolution stage contains the last two nodes (8:10-8:15 and 8:15-8:20). The time span of each evolution stage is consistent, and each stage contains at least two adjacent nodes.

[0064] Step S132: extracting the traffic flow change information of all traffic flow time sequence segment nodes in each of the evolution stages, calculating the average change rate of the traffic flow change information in each of the evolution stages, and taking the average change rate as the intra-stage basic feature, wherein the average change rate is obtained by dividing the total change amount of the traffic flow values in the stage by the stage time span.

[0065] In this embodiment, the first evolution stage (8:00-8:05 and 8:05-8:10) is taken as an example, and the traffic flow change information of the two nodes is extracted. Assuming that the traffic flow change sequence of 8:00-8:05 is [q1, q2, q3, q4, q5], and the traffic flow change sequence of 8:05-8:10 is [q6, q7, q8, q9, q10]. The total change amount of the traffic flow values in the stage is (q5-q1)+(q10-q6), and the stage time span is 10 minutes (from 8:00 to 8:10). The average change rate is equal to the total change amount divided by 10, and the average change rate is taken as the intra-stage basic feature of the evolution stage.

[0066] Step S133: analyzing the transition of the traffic flow change information between two adjacent evolution stages, calculating the difference between the average change rate of the previous evolution stage and the average change rate of the next evolution stage, and simultaneously counting the number of fluctuations of the traffic flow values in the transition process, wherein the difference and the number of fluctuations are taken as independent components of the inter-stage transition feature.

[0067] In this embodiment, taking the two evolution stages described above as examples, the average change rate of the first evolution stage is v1, the average change rate of the second evolution stage is v2, and the difference is v2-v1. Then the number of traffic flow value fluctuations in the transition process (i.e. the traffic flow change between the last time step of the first evolution stage and the first time step of the second evolution stage) is analyzed. For example, the flow value at the last time step (8:10) of the first evolution stage is q10, the flow value at the first time step (8:11) of the second evolution stage is q11, then 8:12 is q12, …, the number of times the flow value rises and then falls or falls and then rises in the transition process (here it is assumed that the time range of the transition process is 8:10-8:15, i.e. the time interval of the second evolution stage) is counted, i.e. the number of fluctuations. The difference v2-v1 and the number of fluctuations are taken as two independent components of the inter-stage transition feature.

[0068] Step S134: Based on the intra-stage basic feature and the inter-stage transition feature of all evolution stages in each of the time sequence association chains, an evolution trend curve of the time sequence association chain is constructed, with the middle time point of each evolution stage as the horizontal axis coordinate and the intra-stage basic feature as the vertical axis coordinate, to draw the evolution trend curve.

[0069] Step S1341: The start time and the end time are extracted from the time interval information of each of the evolution stages, the middle value of the start time and the end time is calculated, and the middle value is taken as the middle time point of the evolution stage.

[0070] In this embodiment, taking the time interval 8:00-8:10 of the first evolution stage as an example, the start time is 8:00, the end time is 8:10, and the middle value is 8:05. 8:05 is taken as the middle time point of the evolution stage; the time interval of the second evolution stage is 8:10-8:20, the start time is 8:10, the end time is 8:20, and the middle value is 8:15. 8:15 is taken as the middle time point of the evolution stage.

[0071] Step S1342: Taking the middle time point of each of the evolution stages as the horizontal axis coordinate value and the intra-stage basic feature of each of the evolution stages as the vertical axis coordinate value, the coordinate point corresponding to each of the evolution stages is marked in a two-dimensional coordinate system, and each coordinate point includes the horizontal axis coordinate value and the vertical axis coordinate value.

[0072] In this embodiment, the middle time point of the first evolution stage is 8:05, and the intra-stage basic feature (average change rate) is v1, so the coordinate point is (8:05, v1); the middle time point of the second evolution stage is 8:15, and the intra-stage basic feature is v2, so the coordinate point is (8:15, v2). The two coordinate points are marked in a two-dimensional coordinate system.

[0073] Step S1343: Check whether the marked coordinate points have abnormal deviation, and if the difference between the longitudinal coordinate value of any coordinate point and the longitudinal coordinate values of the adjacent two coordinate points exceeds the preset deviation threshold, determine that the coordinate point is an abnormal coordinate point.

[0074] In this embodiment, it is assumed that there is a third evolution stage, the intermediate time point is 8:25, the basic feature in the stage is v3, and the coordinate point is (8:25, v3). Now check each coordinate point, for example, for the coordinate point (8:15, v2), calculate the difference between its longitudinal coordinate value and the longitudinal coordinate values of the adjacent two coordinate points (8:05, v1) and (8:25, v3), that is, |v2-v1| and |v2-v3|. If both of the two differences exceed the preset deviation threshold (for example, the preset deviation threshold is Δv), it is determined that (8:15, v2) is an abnormal coordinate point.

[0075] Step S1344: Correct the abnormal coordinate point, replace the longitudinal coordinate value of the abnormal coordinate point with the average of the longitudinal coordinate values of the adjacent two normal coordinate points, keep the horizontal coordinate value of the abnormal coordinate point unchanged, and obtain the corrected coordinate point set.

[0076] In this embodiment, it is assumed that the coordinate point (8:15, v2) is determined to be an abnormal coordinate point, and its adjacent normal coordinate points are (8:05, v1) and (8:25, v3). The corrected longitudinal coordinate value is (v1+v3) / 2, and the horizontal coordinate value is still 8:15, obtaining the corrected coordinate point (8:15, (v1+v3) / 2). After correcting all the coordinate points, the corrected coordinate point set is obtained.

[0077] Step S1345: Connect all the coordinate points in the corrected coordinate point set in turn by using a smooth curve connection method to avoid sharp broken lines in the connection process, and form a preliminary evolution trend curve.

[0078] In this embodiment, the corrected coordinate points are connected by using a smooth curve, such as a Bezier curve or a spline curve, to ensure smooth transition of the curve and no sharp broken lines. For example, connect the coordinate points (8:05, v1), (8:15, (v1+v3) / 2), (8:25, v3), etc. to form a preliminary evolution trend curve, which can reflect the change trend of the basic feature in the stage of the time sequence correlation chain in different evolution stages.

[0079] Step S1346: locally optimizing the preliminary evolution trend curve, for the paragraph whose fluctuation amplitude exceeds the preset fluctuation threshold, adjusting the paragraph by using the curve trend of the adjacent paragraph, and obtaining the final evolution trend curve after the local optimization is completed, the evolution trend curve is used for reflecting the stage internal basic feature change of the time sequence correlation chain in different evolution stages.

[0080] In this embodiment, each paragraph of the preliminary evolution trend curve is analyzed, and the fluctuation amplitude (such as the difference between the maximum and minimum vertical axis values of the curve in the paragraph) of each paragraph is calculated. If the fluctuation amplitude of a paragraph exceeds the preset fluctuation threshold (such as Δv_threshold), the curve trend of its adjacent paragraph is used to adjust the paragraph. For example, the fluctuation amplitude of a paragraph is large, the curve trend of the previous paragraph is steadily rising, and the curve trend of the next paragraph is steadily falling, so the curve of the paragraph is adjusted to make the transition smoother and the fluctuation amplitude lower than the threshold. After all local optimizations are completed, the final evolution trend curve is obtained.

[0081] Step S135: extracting the slope change feature and the inflection point distribution feature of the evolution trend curve, the slope change feature is obtained by calculating the slope change amplitude of the curve segments corresponding to adjacent stages, and the inflection point distribution feature is obtained by identifying the position points at which the curve slope changes from positive to negative or from negative to positive and the distribution of the position points, and the slope change feature and the inflection point distribution feature are combined as the evolution trend feature.

[0082] In this embodiment, the evolution trend curve is divided according to the curve segments corresponding to the evolution stages, such as one curve segment corresponding to each evolution stage. The slope change amplitudes of adjacent curve segments are calculated, for example, the slope of the first curve segment is k1, the slope of the second curve segment is k2, and the slope change amplitude is |k2-k1|. The slope change amplitudes of all adjacent curve segments are combined to obtain the slope change feature. Then the inflection points of the curve, that is, the position points at which the slope changes from positive to negative or from negative to positive, are identified, and the number, distribution time interval and other information of these position points are counted to obtain the inflection point distribution feature. The slope change feature and the inflection point distribution feature are combined to form the evolution trend feature.

[0083] Step S136: calculating the change amplitude of the time sequence correlation strength between all adjacent traffic flow time sequence segment nodes in each time sequence correlation chain, obtaining a correlation strength change sequence, and the change amplitude is obtained by the absolute value of the difference between the previous correlation strength and the next correlation strength.

[0084] In this embodiment, for a pair of adjacent nodes in a time sequence correlation chain, such as the correlation strength between node 1 and node 2 is s1, the correlation strength between node 2 and node 3 is s2, and the correlation strength between node 3 and node 4 is s3, the change amplitudes are |s2-s1| and |s3-s2| respectively. The change amplitudes are arranged in order to obtain a correlation strength change sequence, for example, [|s2-s1|, |s3-s2|, …].

[0085] Step S137: Analyze the fluctuation period feature and the peak value occurrence frequency feature of the correlation strength change sequence. The fluctuation period feature is obtained by identifying the time interval of the repeatedly appearing fluctuation pattern in the sequence. The peak value occurrence frequency feature is obtained by calculating the ratio of the number of values exceeding the preset peak value threshold to the total time step in the sequence. The fluctuation period feature and the peak value occurrence frequency feature are combined as the correlation strength change feature.

[0086] In this embodiment, the correlation strength change sequence is analyzed to identify the repeatedly appearing fluctuation pattern therein, such as the pattern [0.2, 0.3, 0.2] appearing in the sequence. The time interval of the repeatedly appearing fluctuation pattern is calculated to obtain the fluctuation period feature. The number of values exceeding the preset peak value threshold (such as 0.4) in the sequence is counted and divided by the total time step (i.e. the length of the sequence) to obtain the peak value occurrence frequency feature. The two features are combined to form the correlation strength change feature.

[0087] Step S138: Perform interactive fusion processing on the intra-stage basic feature, the independent components of the inter-stage transition feature, the evolution trend feature, and the correlation strength change feature. Calculate the correlation between different features. Adjust the fusion weight of each feature according to the correlation to generate the fused comprehensive feature.

[0088] In this embodiment, the correlation between the intra-stage basic feature, the independent components of the inter-stage transition feature (such as the difference and the fluctuation frequency), the evolution trend feature (the slope change feature and the inflection point distribution feature), and the correlation strength change feature (the fluctuation period feature and the peak value occurrence frequency feature) is calculated, such as the correlation coefficient between the intra-stage basic feature and the slope change feature, the correlation coefficient between the difference of the inter-stage transition feature and the peak value occurrence frequency feature, etc. The fusion weight of each feature is adjusted according to the correlation. The weight of the feature with high correlation can be appropriately increased, and the weight of the feature with low correlation can be appropriately decreased. Then the above features are fused by weighted splicing to generate the fused comprehensive feature, which contains the information of each original feature and has undergone weight adjustment.

[0089] Step S139: Integrate the comprehensive feature and the original features to form a time sequence evolution feature set corresponding to each time sequence correlation chain.

[0090] In this embodiment, the integrated feature after fusion is integrated with the original features such as the intra-stage basic feature, the independent components of the inter-stage transition feature, the evolution trend feature, and the correlation strength change feature, to form a time series evolution feature set. For example, the time series evolution feature set includes the integrated feature, the difference and fluctuation frequency of the intra-stage basic feature and the inter-stage transition feature, the slope change feature and inflection point distribution feature of the evolution trend feature, and the fluctuation period feature and peak value frequency feature of the correlation strength change feature. Each feature is a multi-dimensional numerical set, which is used for subsequent abnormal pattern matching.

[0091] Step S140: calling a pre-trained time series pattern mining model to perform abnormal evolution pattern matching processing on each time series evolution feature set to generate an abnormal pattern mining result, which includes the matched abnormal evolution pattern type and the identification information of the corresponding time series correlation chain.

[0092] Step S141: inputting each time series evolution feature set to the feature input layer of the time series pattern mining model, performing dimension normalization processing on each feature in the time series evolution feature set, and converting it into a standard feature vector conforming to the model processing format.

[0093] In this embodiment, the feature input layer of the time series pattern mining model receives each time series evolution feature set, and performs dimension normalization processing on each feature in the set. For example, the intra-stage basic feature (average change rate) may have a numerical range of 0 to 10, while the correlation strength change feature may have a numerical range of 0 to 1. Through normalization processing (such as min-max normalization, which converts the feature value to 0-1), the dimension of each feature is made to conform to the processing format of the model, and a standard feature vector is generated. Each element of the standard feature vector corresponds to a normalized feature value, and the dimension is consistent with the input dimension required by the model.

[0094] Step S142: inputting the standard feature vector into the feature attention layer of the time series pattern mining model, the feature attention layer calculating the importance weight of each feature in abnormal pattern matching, the importance weight being obtained through correlation analysis of the feature and the historical abnormal pattern feature, the higher the correlation, the greater the importance weight.

[0095] In this embodiment, the feature attention layer of the time series pattern mining model receives the standard feature vector, and calculates the correlation of each feature with the historical abnormal pattern feature. The historical abnormal pattern feature is extracted from the historical abnormal time series evolution feature set and stored in the database of the model. For example, if a certain feature (such as the slope change feature of the evolution trend feature) has a high correlation with the historical abnormal pattern feature, it means that the feature is important in abnormal pattern matching, and its importance weight is large; on the contrary, the weight of the feature with low correlation is small. Through the above method, the importance weight of each feature is calculated.

[0096] Step S143: According to the calculated importance weight, each feature in the standard feature vector is weighted to obtain a weighted feature vector.

[0097] In this embodiment, according to the importance weight calculated in step S142, each element (i.e. the normalized value of each feature) in the standard feature vector is weighted. For example, the standard feature vector is [v1, v2, v3, …], and the corresponding importance weight is [w1, w2, w3, …], and the weighted feature vector is [v1×w1, v2×w2, v3×w3, …], thereby obtaining the weighted feature vector, highlighting the influence of important features.

[0098] Step S144: The weighted feature vector is input into the feature mapping layer of the time series pattern mining model, and the feature mapping layer maps the weighted feature vector to a preset pattern feature space through a preset mapping rule. The pattern feature space is a high-dimensional feature space constructed based on historical abnormal pattern features, and a pattern space feature vector is obtained.

[0099] In this embodiment, the feature mapping layer of the time series pattern mining model receives the weighted feature vector, and maps the weighted feature vector to the pattern feature space according to the preset mapping rule (such as linear mapping, nonlinear mapping, etc., the specific mapping rule is determined during model training). The pattern feature space is a high-dimensional space, and each point in the space corresponds to a feature vector. The pattern feature space is constructed based on a large number of historical abnormal pattern features, so that similar abnormal patterns are close in space. Through mapping, a pattern space feature vector is obtained, which has a specific coordinate position in the pattern feature space.

[0100] Step S145: calling a pattern library of the time-series pattern mining model, calculating the similarity between the pattern space feature vector and each standard pattern feature vector in the pattern library, obtaining the similarity by comparing the spatial distance between the pattern space feature vector and each standard pattern feature vector in the pattern feature space, the closer the spatial distance, the higher the similarity, the pattern library stores standard pattern feature vectors corresponding to a plurality of preset abnormal evolution patterns, and each abnormal evolution pattern corresponds to a unique pattern identifier.

[0101] Step S1451: reading all standard pattern feature vectors corresponding to preset abnormal evolution patterns and pattern identifiers corresponding to each abnormal evolution pattern from the pattern library of the time-series pattern mining model, forming a corresponding set of standard pattern feature vectors and pattern identifiers.

[0102] In this embodiment, the pattern library of the time-series pattern mining model stores a plurality of abnormal evolution patterns, such as "traffic surge type", "traffic reduction type", "traffic fluctuation type", etc., each pattern corresponding to a standard pattern feature vector and a pattern identifier (such as ID1, ID2, ID3, etc.). Reading these standard pattern feature vectors and corresponding pattern identifiers from the pattern library forms a corresponding set, which facilitates subsequent similarity calculation.

[0103] Step S1452: for each standard pattern feature vector in the corresponding set, determining the coordinate position of the standard pattern feature vector in the pattern feature space, and simultaneously determining the coordinate position of the pattern space feature vector in the pattern feature space.

[0104] In this embodiment, the pattern feature space is a high-dimensional space, and each standard pattern feature vector and pattern space feature vector has a corresponding coordinate position. For example, the coordinates of the standard pattern feature vector V1 in the pattern feature space are (x1, y1, z1, …), and the coordinates of the pattern space feature vector V are (x, y, z, …).

[0105] Step S1453: calculating the spatial distance between the two coordinate positions, by traversing all dimensions of the two vectors, counting the difference in coordinate values in each dimension, and then obtaining the spatial distance based on the difference.

[0106] In this embodiment, the spatial distance between V and V1 is calculated, each dimension is traversed, the difference values of x-x1, y-y1, z-z1, etc. are calculated, and then the square root of the sum of the above difference values (Euclidean distance) or other distance measurement methods (such as Manhattan distance, cosine distance, etc.) are used to obtain the spatial distance.

[0107] Step S1454: determining the initial value of the similarity according to the calculated spatial distance, the smaller the spatial distance, the higher the initial value of the similarity.

[0108] In this embodiment, a mapping relationship between a spatial distance and an initial value of similarity degree is preset, such as when the spatial distance is 0, the initial value of similarity degree is 1; when the spatial distance is d1, the initial value of similarity degree is 0.8; when the spatial distance is d2, the initial value of similarity degree is 0.5, and so on (d1 and d2 are preset distance thresholds). According to the calculated spatial distance, the corresponding initial value of similarity degree is found. For example, the spatial distance is d0, if d0 < d1, the initial value of similarity degree is 1; if d1 ≤ d0 < d2, the initial value of similarity degree is 0.8, and so on.

[0109] Step S1455: The initial value of similarity degree is adjusted in combination with the time span of the time sequence correlation chain corresponding to the current time sequence evolution feature set. If the difference between the time span of the time sequence correlation chain and the time span corresponding to the standard mode feature vector is less than a preset difference threshold, the initial value of similarity degree is increased; if the difference between the time span of the time sequence correlation chain and the time span corresponding to the standard mode feature vector is greater than or equal to the preset difference threshold, the initial value of similarity degree is decreased.

[0110] In this embodiment, the time span corresponding to the standard mode feature vector is determined during model training, such as the standard time span of the "traffic surge type" mode is 15 minutes. The time span of the current time sequence correlation chain is 20 minutes, and the difference is 5 minutes. The preset difference threshold is 10 minutes, and since 5 minutes is less than 10 minutes, the initial value of similarity degree is increased, such as by 0.1. If the difference is greater than or equal to 10 minutes, the initial value of similarity degree is decreased, such as by 0.1.

[0111] Step S1456: The adjusted similarity degree is taken as the final similarity degree between the mode space feature vector and the corresponding standard mode feature vector.

[0112] In this embodiment, after the initial value of similarity degree obtained in step S1454 is adjusted in step S1455, the obtained value is the final similarity degree between the mode space feature vector and the corresponding standard mode feature vector. For example, the initial value of similarity degree is 0.8, the time span difference is less than the threshold, and after adjustment, the final similarity degree is increased by 0.1, which is 0.9.

[0113] Step S1457: The final similarity degrees corresponding to all standard mode feature vectors are recorded, and the corresponding mode identifiers are associated to form an association list of similarity degrees and mode identifiers.

[0114] In this embodiment, the final similarity degree of each standard mode feature vector is associated with its corresponding mode identifier to form a list. For example, the final similarity degree corresponding to mode identifier ID1 is 0.9, the final similarity degree corresponding to mode identifier ID2 is 0.7, the final similarity degree corresponding to mode identifier ID3 is 0.8, and so on. This list is the association list of similarity degrees and mode identifiers.

[0115] Step S146: Sort all the calculated similarity degrees, and filter out the abnormal evolution mode type and mode identifier corresponding to the standard mode feature vector with the highest similarity degree.

[0116] In this embodiment, the similarity degrees in the association list of similarity degrees and mode identifiers are sorted from high to low, and the standard mode feature vector with the highest similarity degree is found. The abnormal evolution mode type and mode identifier corresponding to the standard mode feature vector are the abnormal mode that the current time series evolution feature set matches best. For example, after sorting, the highest similarity degree is 0.9, the corresponding mode identifier is ID1, and the abnormal evolution mode type is "traffic surge type".

[0117] Step S147: Check whether the similarity degree corresponding to the abnormal evolution mode type with the highest similarity degree exceeds the preset matching threshold. If the similarity degree exceeds the preset matching threshold, determine that the abnormal evolution mode type is the abnormal evolution mode type that the current time series evolution feature set matches.

[0118] In this embodiment, the preset matching threshold is 0.7. Since the highest similarity degree is 0.9, which exceeds 0.7, it is determined that the abnormal evolution mode type ("traffic surge type") is the abnormal evolution mode type that the current time series evolution feature set matches.

[0119] Step S148: If the similarity degree does not exceed the preset matching threshold, the abnormal evolution mode type with the second highest similarity degree is taken as the candidate mode type, and the difference features between the current time series evolution feature set and the candidate mode type are recorded.

[0120] In this embodiment, it is assumed that the highest similarity degree is 0.65, which does not exceed the preset matching threshold 0.7. Therefore, the abnormal evolution mode type corresponding to the second highest similarity degree (for example, 0.6) is taken as the candidate mode type. Meanwhile, the differences between the current time series evolution feature set and the standard mode feature vector of the candidate mode type are analyzed, and the difference features are extracted, such as different values of a certain feature or different distributions of features. The above difference features are recorded.

[0121] Step S149: Record the matched abnormal evolution mode type or candidate mode type, the identifier information of the time series association chain corresponding to the abnormal evolution mode type or candidate mode type, and the difference features, and arrange them according to the preset format to generate the abnormal mode mining result.

[0122] In this embodiment, the matched abnormal evolution pattern type (or candidate pattern type), the identification information of the corresponding time sequence association chain (such as the number of the chain), and the difference feature (if any) are arranged in a preset format (such as JSON format or XML format) to generate an abnormal pattern mining result. For example, the abnormal pattern mining result is: {“abnormal evolution pattern type”:“traffic surge type”,“time sequence association chain identification”:“chain 1”,“difference feature”:null} (if there is no difference feature).

[0123] Step S150: generating a traffic flow regulation suggestion set based on the abnormal pattern mining result, adjusting the regulation content in combination with the influence range of the abnormal pattern, each suggestion in the traffic flow regulation suggestion set corresponding to an abnormal evolution pattern type, and each suggestion containing regulation direction information and regulation implementation step information for the corresponding abnormal evolution pattern.

[0124] Step S151: analyzing the abnormal pattern mining result to extract all abnormal evolution pattern types, identification information of the corresponding time sequence association chain, and difference features contained therein.

[0125] In this embodiment, the abnormal pattern mining result is analyzed to extract the abnormal evolution pattern type (such as“traffic surge type”), the corresponding time sequence association chain identification (such as“chain 1”), and the difference feature (if any). For example, from the abnormal pattern mining result in JSON format, these information is obtained by analyzing the fields.

[0126] Step S152: for each abnormal evolution pattern type, retrieving a basic regulation strategy associated with the abnormal evolution pattern type from a preset regulation strategy library, the regulation strategy library storing a mapping relationship between each abnormal evolution pattern type and a corresponding basic regulation strategy, and the basic regulation strategy containing a regulation target and a basic regulation measure.

[0127] In this embodiment, the regulation strategy library stores a basic regulation strategy corresponding to each abnormal evolution pattern type. For example, the basic regulation strategy of“traffic surge type” has a regulation target of reducing the traffic flow in the region, and the basic regulation measures include issuing traffic warning information, guiding vehicles to detour, adjusting the signal light duration of the tunnel entrance, etc. For the extracted“traffic surge type” abnormal evolution pattern type, the corresponding basic regulation strategy is retrieved from the regulation strategy library.

[0128] Step S153: querying the tunnel monitoring region information involved in the traffic flow time sequence segment corresponding to the time sequence association chain according to the identification information of the time sequence association chain corresponding to each abnormal evolution pattern type, the tunnel monitoring region information including the location of the monitoring region, the number of lanes in the region, and the traffic flow carrying upper limit corresponding to the region.

[0129] In this embodiment, according to the identification information of the time sequence correlation chain (such as "chain 1"), the tunnel monitoring areas involved in the traffic flow time sequence fragment corresponding to the time sequence correlation chain are queried. For example, the traffic flow time sequence fragment corresponding to chain 1 involves monitoring area A (tunnel entrance section) and monitoring area B (middle driving section), and the information of the two monitoring areas is queried: the position of monitoring area A is 500 meters before the tunnel entrance, the number of lanes is 3, and the upper limit of traffic flow carrying capacity is 2000 vehicles per hour; the position of monitoring area B is the middle section of the tunnel, the number of lanes is 4, and the upper limit of traffic flow carrying capacity is 2500 vehicles per hour.

[0130] Step S154: Based on the tunnel monitoring area information, analyze the surrounding monitoring areas that may be affected by the abnormal evolution mode type, and determine the range of the affected surrounding monitoring areas by calculating the traffic flow interaction frequency between the target monitoring area and the surrounding monitoring areas.

[0131] Step S1541: From the tunnel traffic management system, retrieve the traffic flow interaction data between the target monitoring area and all surrounding monitoring areas in the historical time period, which includes the traffic flow value from the target monitoring area to the surrounding monitoring area and the traffic flow value from the surrounding monitoring area to the target monitoring area per unit time.

[0132] In this embodiment, the target monitoring area is monitoring area A (tunnel entrance section), and the surrounding monitoring areas include monitoring area B (middle driving section), monitoring area C (exit section), monitoring area D (adjacent tunnel entrance section), etc. From the database of the tunnel traffic management system, the traffic flow interaction data between monitoring area A and these surrounding monitoring areas in the historical time period (such as the morning peak period in the past month) is retrieved. For example, the traffic flow value from monitoring area A to monitoring area B per unit time (1 hour) is 1500 vehicles, the traffic flow value from monitoring area B to monitoring area A is 200 vehicles, the traffic flow value from monitoring area A to monitoring area D is 500 vehicles, and the traffic flow value from monitoring area D to monitoring area A is 100 vehicles, etc.

[0133] Step S1542: For each surrounding monitoring area, calculate the traffic flow interaction frequency between the target monitoring area and the surrounding monitoring area, which is obtained by dividing the sum of the bidirectional traffic flow values per unit time by the unit time length.

[0134] In this embodiment, for the monitoring area B, the sum of the bidirectional traffic flow values in a unit time is 1500+200=1700 vehicles, and the unit time length is 1 hour, so the traffic flow interaction frequency is 1700 / 1=1700 vehicles / hour. For the monitoring area D, the sum of the bidirectional traffic flow values is 500+100=600 vehicles, and the traffic flow interaction frequency is 600 / 1=600 vehicles / hour.

[0135] Step S1543: comparing the traffic flow interaction frequency of each peripheral monitoring area with a preset traffic flow interaction frequency threshold.

[0136] In this embodiment, the preset traffic flow interaction frequency threshold is 1000 vehicles / hour. The traffic flow interaction frequency of the monitoring area B is 1700 vehicles / hour, which exceeds the threshold; the traffic flow interaction frequency of the monitoring area D is 600 vehicles / hour, which does not exceed the threshold.

[0137] Step S1544: if the traffic flow interaction frequency of any one peripheral monitoring area exceeds the traffic flow interaction frequency threshold, determining that the peripheral monitoring area is a high-correlation peripheral area possibly affected by the abnormal evolution mode.

[0138] In this embodiment, the traffic flow interaction frequency of the monitoring area B is 1700 vehicles / hour, which exceeds the threshold 1000 vehicles / hour, so the monitoring area B is determined to be a high-correlation peripheral area.

[0139] Step S1545: if the traffic flow interaction frequency of any one peripheral monitoring area does not exceed the traffic flow interaction frequency threshold, but the peripheral monitoring area is directly adjacent to the target monitoring area and no other monitoring area is interposed, determining that the peripheral monitoring area is a low-correlation peripheral area possibly affected by the abnormal evolution mode.

[0140] In this embodiment, the traffic flow interaction frequency of the monitoring area D is 600 vehicles / hour, which does not exceed the threshold, but the monitoring area D is not directly adjacent to the target monitoring area A (there are the monitoring area B and the tunnel body interposed), so the monitoring area D is not determined to be a low-correlation peripheral area. Assuming that there is another monitoring area E directly adjacent to the monitoring area A, and the traffic flow interaction frequency is 800 vehicles / hour, which does not exceed the threshold, then the monitoring area E is determined to be a low-correlation peripheral area.

[0141] Step S1546: if the traffic flow interaction frequency of any one peripheral monitoring area does not exceed the traffic flow interaction frequency threshold, and the peripheral monitoring area is not directly adjacent to the target monitoring area, determining that the peripheral monitoring area is an unaffected peripheral area.

[0142] In this embodiment, the monitoring area D is not directly adjacent to the target monitoring area A, and the traffic flow interaction frequency does not exceed the threshold, so the monitoring area D is determined to be an unaffected peripheral area.

[0143] Step S1547: integrate the high-correlation peripheral region and the low-correlation peripheral region to form an affected peripheral monitoring region range, and mark the correlation type of each affected peripheral monitoring region.

[0144] In this embodiment, the high-correlation peripheral region is monitoring region B, and the low-correlation peripheral region is monitoring region E. They are integrated to form an affected peripheral monitoring region range, and the correlation type of monitoring region B is marked as “high correlation”, and the correlation type of monitoring region E is marked as “low correlation”.

[0145] Step S155: integrate the tunnel monitoring region information and the affected peripheral monitoring region range into the corresponding basic control strategy, adjust the control object and the control range in the basic control strategy, extend the control object to the affected peripheral monitoring region, and adjust the control range according to the number of lanes in the region, to obtain a preliminary individualized control strategy for the specific monitoring region and the peripheral region.

[0146] In this embodiment, the control object of the basic control strategy is originally monitoring region A, and now the control object is extended to monitoring region B (high correlation) and monitoring region E (low correlation). In terms of control range, the number of lanes in monitoring region A is 3, the number of lanes in monitoring region B is 4, and the number of lanes in monitoring region E is 2. The signal light duration adjustment in the basic control measure is originally for 3 lanes in monitoring region A, and now it is for 4 lanes in monitoring region B, and the adjustment amplitude is adjusted according to the lane number ratio (for example, the more the number of lanes, the signal light release duration is appropriately increased); for 2 lanes in monitoring region E, the adjustment amplitude is correspondingly reduced. In this way, a preliminary individualized control strategy is obtained, and the control object includes monitoring regions A, B and E, and the control range is adjusted according to the number of lanes in each region.

[0147] Step S156: analyze the possible impact of each preliminary individualized control strategy on the traffic flow of the target monitoring region and the affected peripheral monitoring region after the implementation of the control strategy, and determine whether the control measure will cause new traffic flow abnormalities by simulating the traffic flow changes after the implementation of the control measure.

[0148] In this embodiment, a traffic flow simulation software is used to input the content of the preliminary individualized control strategy to simulate the traffic flow changes after the implementation of the control measure. For example, the traffic flow changes of monitoring regions A, B and E after the implementation of the control measure such as publishing traffic warning information, guiding vehicles to detour, and adjusting signal light duration are simulated. It is checked whether new traffic flow abnormalities will occur, such as whether the traffic flow of monitoring region E will suddenly increase after the control, or whether the traffic flow of monitoring region B will suddenly decrease to cause new congestion, etc.

[0149] Step S157: If the simulation result shows that it will cause new traffic flow abnormality, adjust the intensity of the regulation measures in the preliminary personalized regulation strategy, reduce the intensity of the regulation measures that may cause new abnormality, and re-perform the influence simulation until the simulation result shows that it will not cause new traffic flow abnormality.

[0150] In this embodiment, if the simulation result shows that the traffic flow of the monitoring area E increases sharply after regulation, exceeds the upper limit of its carrying capacity, and a new abnormality occurs. Then, the intensity of the regulation measures in the preliminary personalized regulation strategy is adjusted, such as reducing the number of guided vehicles that bypass to the monitoring area E, or reducing the signal light release time adjustment range of the monitoring area E. Then, the simulation is re-performed until the simulation result shows that the traffic flow of each monitoring area is within the normal range and no new abnormality occurs.

[0151] Step S158: If the simulation result shows that it will not cause new traffic flow abnormality, the current preliminary personalized regulation strategy is retained as the optimized personalized regulation strategy.

[0152] In this embodiment, after adjustment and re-simulation, the simulation result shows that the traffic flow of the monitoring areas A, B and E is within the normal range and no new abnormality occurs, so the current preliminary personalized regulation strategy is retained as the optimized personalized regulation strategy.

[0153] Step S159: All optimized personalized regulation strategies are integrated, and the corresponding abnormal evolution mode type identifier, monitoring area identifier and implementation priority are added to each strategy to form the traffic flow regulation suggestion set. Each suggestion in the traffic flow regulation suggestion set includes abnormal evolution mode type, corresponding monitoring area information and optimized regulation strategy content.

[0154] In this embodiment, all optimized personalized regulation strategies are collected, and the abnormal evolution mode type identifier (such as "flow surge type"), monitoring area identifier (such as the identifiers of the monitoring areas A, B and E) and implementation priority (determined according to the severity and influence range of the abnormality, such as the regulation priority of the high correlation area is higher than that of the low correlation area) are added to each strategy. Then, the above strategies are integrated to form the traffic flow regulation suggestion set. For example, the content of a suggestion is: the abnormal evolution mode type is "flow surge type", the corresponding monitoring areas are A, B and E, the optimized regulation strategy is to publish traffic warning information, guide part of the vehicles to bypass to the non-affected area, adjust the signal light time of the monitoring area A to increase by 5 seconds per cycle, the signal light time of the monitoring area B to increase by 8 seconds per cycle, and the signal light time of the monitoring area E to increase by 3 seconds per cycle, and the implementation priority is high. All the above suggestions are integrated to form the traffic flow regulation suggestion set.

[0155] Figure 2An exemplary hardware and software components of the tunnel traffic flow big data mining system 100 that can implement the idea of the present application in connection with time series are shown in the schematic diagram. For example, the processor 120 can be used in the tunnel traffic flow big data mining system 100 in connection with time series and for performing the functions in the present application.

[0156] For example, the tunnel traffic flow big data mining system 100 in connection with time series can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, a ROM, or a RAM, or any combination thereof. Exemplarily, the tunnel traffic flow big data mining system 100 in connection with time series can also include program instructions stored in a ROM, a RAM, or other types of non-transitory storage media, or any combination thereof. The methods of the present application can be implemented according to these program instructions. The tunnel traffic flow big data mining system 100 in connection with time series also includes an I / O interface 150 between the computer and other input / output devices.

[0157] It should be noted that, in order to simplify the expression of the present disclosure and to help the understanding of one or more embodiments of the present application, in the foregoing description of the embodiments of the present application, various features are sometimes incorporated into one embodiment, drawing or description thereof.

Claims

1. A tunnel traffic flow big data mining method combined with time series, characterized in that, The method comprises: acquiring a time series data set corresponding to tunnel traffic flow big data, the time series data set containing a plurality of traffic flow time series segments collected in continuous time dimensions, each traffic flow time series segment recording traffic flow change information of a specific monitoring area in a tunnel in a corresponding time interval; constructing a time series correlation chain based on the time series adjacent relationship between the traffic flow time series segments, performing feature filling processing on the time gap of adjacent segments during the construction process to obtain a time series correlation chain set, the time series correlation chain being used to represent the correlation relationship of traffic flow change information in different time intervals, and each time series correlation chain containing a node sequence and an association strength parameter between nodes; performing time series evolution feature extraction processing on each time series correlation chain, respectively extracting features from the time dimension and the correlation dimension and performing interactive fusion to obtain a time series evolution feature set corresponding to each time series correlation chain, the time series evolution feature set containing change trend features and correlation strength change features of the time series correlation chain in the time dimension; calling a pre-trained time series pattern mining model to perform abnormal evolution pattern matching processing on each time series evolution feature set to generate an abnormal pattern mining result, the abnormal pattern mining result containing an abnormal evolution pattern type matched and identification information of the time series correlation chain corresponding to the abnormal evolution pattern; generating a traffic flow regulation suggestion set based on the abnormal pattern mining result, adjusting the regulation content in combination with the influence range of the abnormal pattern, each suggestion in the traffic flow regulation suggestion set corresponding to an abnormal evolution pattern type, and each suggestion containing regulation direction information and regulation implementation step information for the corresponding abnormal evolution pattern. 2.The method of claim 1, wherein, The method comprises: extracting time interval information of all traffic flow time series segments from the time series data set, the time interval information containing the start time and end time of each traffic flow time series segment; determining the time series adjacent relationship between any two traffic flow time series segments according to the time interval information, the time series adjacent relationship being used to indicate whether the time intervals of two traffic flow time series segments have a continuous connection relationship, if there is an interval between the end time of a previous traffic flow time series segment and the start time of a subsequent traffic flow time series segment, it is determined that there is a time gap between the two traffic flow time series segments; for two traffic flow time series segments with a time gap, performing feature interpolation processing on the traffic flow change information of adjacent segments to generate a transition feature sequence that fills the time gap, the time length of the transition feature sequence being consistent with the time length of the time gap; for two traffic flow time series segments with the time series adjacent relationship, calculating the similarity degree of traffic flow change information in the two traffic flow time series segments, and taking the similarity degree as the time series correlation strength between the two traffic flow time series segments; The time-series correlation strength is dynamically calibrated based on the correlation strength distribution of the same type of segment in historical traffic flow data, so as to eliminate the influence of abnormal values on the correlation strength. According to a preset correlation strength threshold, traffic flow time-series segment pairs whose time-series correlation strength meets the threshold requirement are screened out to obtain an effective correlation pair set, each correlation pair in the effective correlation pair set including two traffic flow time-series segments and the calibrated time-series correlation strength. One traffic flow time-series segment in each effective correlation pair is taken as a starting node, and the other traffic flow time-series segment is taken as a subsequent node, a plurality of effective correlation pairs are sequentially connected in the order of time intervals, if a node is repeated in the connection process, the connection path with higher correlation strength is reserved, and an initial time-series correlation chain is constructed. Redundant nodes in the initial time-series correlation chain are removed, nodes in the initial time-series correlation chain whose traffic flow change information has a repetition degree exceeding a preset repetition degree threshold with adjacent nodes are deleted, and the time-series correlation strength between nodes before and after the removed node is updated. All optimized time-series correlation chains are integrated, the number of nodes, total time span and average correlation strength of each time-series correlation chain are recorded, and a time-series correlation chain set is formed. 3.The method of claim 2, wherein, The similarity degree of traffic flow change information in the two traffic flow time-series segments is calculated, and the similarity degree is taken as the time-series correlation strength between the two traffic flow time-series segments, including: Traffic flow change information in the two traffic flow time-series segments with the time-series adjacent relationship is extracted, if there is a transition feature sequence, the transition feature sequence and the traffic flow change information of the subsequent traffic flow time-series segment are combined to obtain a first traffic flow change sequence and a subsequent traffic flow change sequence, respectively. The first traffic flow change sequence and the subsequent traffic flow change sequence are subjected to length unification processing in the time dimension, and for the first traffic flow change sequence and the subsequent traffic flow change sequence after length unification, the traffic flow value difference at the corresponding position is calculated step by step, and a difference sequence composed of a plurality of time step difference values is obtained. The overall fluctuation degree of the difference sequence is calculated according to the traffic flow value difference of all time steps, and the overall fluctuation degree is represented by the distribution concentration of all difference values in the difference sequence. A basic value of the similarity degree is determined based on the overall fluctuation degree, and the lower the overall fluctuation degree, the higher the basic value of the similarity degree. The basic value of the similarity degree is corrected in combination with the position correlation degree of the monitoring areas corresponding to the two traffic flow time-series segments, if the monitoring areas corresponding to the two traffic flow time-series segments are adjacent, the basic value of the similarity degree is increased, and if the monitoring areas corresponding to the two traffic flow time-series segments are not adjacent, the basic value of the similarity degree is kept unchanged. The corrected similarity degree is taken as the time-series correlation strength between the two traffic flow time-series segments. 4.The method of claim 1, wherein, The time sequence evolution feature extraction processing is performed on each time sequence correlation chain, features are extracted from time dimension and correlation dimension respectively, and are interactively fused to obtain a time sequence evolution feature set corresponding to each time sequence correlation chain, including: Each time sequence correlation chain is divided into a plurality of continuous evolution stages according to the chronological order of time intervals, each evolution stage contains at least two adjacent traffic flow time sequence segment nodes, and the time span of each evolution stage remains consistent; Traffic flow change information of all traffic flow time sequence segment nodes in each evolution stage is extracted, and the average change rate of the traffic flow change information in each evolution stage is calculated, the average change rate is obtained by dividing the total change amount of the traffic flow value in the stage by the stage time span, and the average change rate is taken as the intra-stage basic feature; The transition of traffic flow change information between adjacent two evolution stages is analyzed, the difference between the average change rate of the previous evolution stage and the average change rate of the next evolution stage is calculated, and the fluctuation frequency of the traffic flow value in the transition process is counted, the difference and the fluctuation frequency are taken as independent components of the inter-stage transition feature respectively; Based on the intra-stage basic feature and the inter-stage transition feature of all evolution stages in each time sequence correlation chain, an evolution trend curve of the time sequence correlation chain is constructed, the middle time point of each evolution stage is taken as the horizontal axis coordinate, and the intra-stage basic feature is taken as the vertical axis coordinate, and the evolution trend curve is drawn; The slope change feature and the inflection point distribution feature of the evolution trend curve are extracted, the slope change feature is obtained by calculating the slope change amplitude of the curve segment of adjacent stages, and the inflection point distribution feature is obtained by identifying the position points and the distribution of the position points where the curve slope changes from positive to negative or from negative to positive, and the slope change feature and the inflection point distribution feature are combined as the evolution trend feature; The change amplitude of the time sequence correlation strength between all adjacent traffic flow time sequence segment nodes in each time sequence correlation chain is calculated to obtain a correlation strength change sequence, the change amplitude is obtained by the absolute value of the difference between the next correlation strength and the previous correlation strength; The fluctuation period feature and the peak value appearance frequency feature of the correlation strength change sequence are analyzed, the fluctuation period feature is obtained by identifying the time interval of the repeatedly appearing fluctuation mode in the sequence, and the peak value appearance frequency feature is obtained by calculating the ratio of the number of values exceeding the preset peak value threshold to the total time step in the sequence, and the fluctuation period feature and the peak value appearance frequency feature are combined as the correlation strength change feature; The intra-stage basic feature, the independent components of the inter-stage transition feature, the evolution trend feature and the correlation strength change feature are interactively fused, the correlation between different features is calculated, the fusion weight of each feature is adjusted according to the correlation, and a fused comprehensive feature is generated; The comprehensive feature and the original features are integrated to form a time sequence evolution feature set corresponding to each time sequence correlation chain. 5.The tunnel traffic flow big data mining method of claim 4, wherein, The evolution trend curve of the time sequence correlation chain is constructed based on the intra-stage basic features and the inter-stage transition features of all evolution stages in each time sequence correlation chain, with the intermediate time point of each evolution stage as the horizontal axis coordinate, and the intra-stage basic features as the vertical axis coordinate, and the evolution trend curve is drawn, including: The start time and the end time are extracted from the time interval information of each evolution stage, and the intermediate value of the start time and the end time is calculated as the intermediate time point of the evolution stage; Each evolution stage is marked with a coordinate point in a two-dimensional coordinate system, with the intermediate time point of each evolution stage as the horizontal axis coordinate value, and the intra-stage basic features of each evolution stage as the vertical axis coordinate value. Each coordinate point contains a horizontal axis coordinate value and a vertical axis coordinate value; Check if the marked coordinate points deviate abnormally. If the difference between the vertical axis coordinate value of any coordinate point and the vertical axis coordinate values of the adjacent two coordinate points exceeds the preset deviation threshold, the coordinate point is determined to be an abnormal coordinate point; The abnormal coordinate point is corrected by replacing the vertical axis coordinate value of the abnormal coordinate point with the average of the vertical axis coordinate values of the adjacent two normal coordinate points, and the horizontal axis coordinate value of the abnormal coordinate point is kept unchanged to obtain a corrected coordinate point set; All coordinate points in the corrected coordinate point set are connected in sequence using a smooth curve connection method to avoid sharp broken lines in the connection process, forming a preliminary evolution trend curve; The preliminary evolution trend curve is locally optimized. For paragraphs with fluctuation amplitude exceeding the preset fluctuation threshold, the curve trend of adjacent paragraphs is used to adjust the paragraph. After local optimization, the final evolution trend curve is obtained, which reflects the intra-stage basic feature changes of the time sequence correlation chain in different evolution stages. 6.The method of claim 1, wherein, The pre-trained time sequence pattern mining model is called to perform abnormal evolution pattern matching processing on each time sequence evolution feature set to generate an abnormal pattern mining result, including: Each time sequence evolution feature set is input into the feature input layer of the time sequence pattern mining model, and each feature in the time sequence evolution feature set is subjected to dimension normalization processing to convert it into a standard feature vector conforming to the model processing format; The standard feature vector is input into the feature attention layer of the time sequence pattern mining model. The feature attention layer calculates the importance weight of each feature in abnormal pattern matching. The importance weight is obtained through correlation analysis of the feature and the historical abnormal pattern feature. The higher the correlation, the greater the importance weight; According to the calculated importance weight, each feature in the standard feature vector is weighted to obtain a weighted feature vector; The weighted feature vector is input into the feature mapping layer of the time sequence pattern mining model. The feature mapping layer maps the weighted feature vector to a preset pattern feature space through a preset mapping rule. The pattern feature space is a high-dimensional feature space constructed based on historical abnormal pattern features, and a pattern space feature vector is obtained; The pattern library of the time sequence pattern mining model is called, similarity degrees between the pattern space feature vector and each standard pattern feature vector in the pattern library are calculated, the similarity degrees are obtained by comparing the spatial distances of the pattern space feature vector and each standard pattern feature vector in the pattern feature space, the closer the spatial distance, the higher the similarity degree, the pattern library stores standard pattern feature vectors corresponding to multiple preset abnormal evolution patterns, and each abnormal evolution pattern corresponds to a unique pattern identifier; All the calculated similarity degrees are sorted, and the abnormal evolution pattern type and the pattern identifier corresponding to the standard pattern feature vector with the highest similarity degree are screened out; It is checked whether the similarity degree corresponding to the abnormal evolution pattern type with the highest similarity degree exceeds a preset matching threshold, if the preset matching threshold is exceeded, the abnormal evolution pattern type is determined as the abnormal evolution pattern type matched by the current time sequence evolution feature set; If the preset matching threshold is not exceeded, the abnormal evolution pattern type with the second highest similarity degree is taken as a candidate pattern type, and the difference features of the current time sequence evolution feature set and the candidate pattern type are recorded; The matched abnormal evolution pattern type or the candidate pattern type, the identifier information of the time sequence correlation chain, and the difference features are recorded, and are arranged in a preset format to generate the abnormal pattern mining result. 7.The tunnel traffic flow big data mining method of combining time series according to claim 6, characterized in that, The calling of the pattern library of the time sequence pattern mining model, and the calculation of the similarity degrees between the pattern space feature vector and each standard pattern feature vector in the pattern library, include: All standard pattern feature vectors corresponding to preset abnormal evolution patterns and pattern identifiers corresponding to each abnormal evolution pattern are read from the pattern library of the time sequence pattern mining model to form a corresponding set of standard pattern feature vectors and pattern identifiers; For each standard pattern feature vector in the corresponding set, the coordinate position of the standard pattern feature vector in the pattern feature space is determined, and the coordinate position of the pattern space feature vector in the pattern feature space is determined; The spatial distance between the two coordinate positions is calculated by traversing all dimensions of the two vectors, counting the difference of the coordinate values in each dimension, and then obtaining the spatial distance based on the difference; The initial value of the similarity degree is determined according to the calculated spatial distance, the smaller the spatial distance, the higher the initial value of the similarity degree; The initial value of the similarity degree is adjusted in combination with the time span of the time sequence correlation chain corresponding to the current time sequence evolution feature set, if the difference between the time span of the time sequence correlation chain and the time span corresponding to the standard pattern feature vector is less than a preset difference threshold, the initial value of the similarity degree is increased, if the difference between the time span of the time sequence correlation chain and the time span corresponding to the standard pattern feature vector is greater than or equal to the preset difference threshold, the initial value of the similarity degree is reduced; The adjusted similarity degree is taken as the final similarity degree between the pattern space feature vector and the corresponding standard pattern feature vector; The final similarity degrees corresponding to all standard pattern feature vectors are recorded, and the corresponding pattern identifiers are associated to form an associated list of similarity degrees and pattern identifiers. 8.The method of claim 1, wherein, The traffic flow regulation suggestion set is generated based on the abnormal pattern mining result, and the regulation content is analyzed by combining the influence range corresponding to the abnormal pattern, including: The abnormal pattern mining result is analyzed, and all abnormal evolution pattern types, identification information of the time sequence correlation chain, and difference characteristics contained therein are extracted; For each abnormal evolution pattern type, the basic regulation strategy associated with the abnormal evolution pattern type in the preset regulation strategy library is retrieved, the mapping relationship between each abnormal evolution pattern type and the corresponding basic regulation strategy is stored in the regulation strategy library, and the basic regulation strategy includes a regulation target and a basic regulation measure; According to the identification information of the time sequence correlation chain corresponding to each abnormal evolution pattern type, the tunnel monitoring area information involved in the traffic flow time sequence segment corresponding to the time sequence correlation chain is queried, and the tunnel monitoring area information includes the location of the monitoring area, the number of lanes in the area, and the traffic flow carrying upper limit corresponding to the area; Based on the tunnel monitoring area information, the surrounding monitoring areas that may be affected by the abnormal evolution pattern type are analyzed, the range of the affected surrounding monitoring areas is determined by calculating the traffic flow interaction frequency between the target monitoring area and the surrounding monitoring areas, and the range of the affected surrounding monitoring areas is determined. The tunnel monitoring area information and the range of the affected surrounding monitoring areas are integrated into the corresponding basic regulation strategy, the regulation object and the regulation range in the basic regulation strategy are adjusted, the regulation object is extended to the affected surrounding monitoring areas, and the regulation range is adjusted according to the number of lanes in the area, to obtain a preliminary individualized regulation strategy for a specific monitoring area and surrounding area; The influence of each preliminary individualized regulation strategy on the traffic flow of the target monitoring area and the affected surrounding monitoring areas after implementation is analyzed, whether the regulation measure will lead to new traffic flow abnormalities is judged by simulating the traffic flow change after the implementation of the regulation measure; If the simulation result shows that it will lead to new traffic flow abnormalities, the regulation measure intensity in the preliminary individualized regulation strategy is adjusted, the regulation measure intensity that may trigger new abnormalities is reduced, and the influence simulation is performed again until the simulation result shows that it will not lead to new traffic flow abnormalities; If the simulation result shows that it will not lead to new traffic flow abnormalities, the current preliminary individualized regulation strategy is retained as the optimized individualized regulation strategy; All optimized individualized regulation strategies are integrated, each strategy is added with corresponding abnormal evolution pattern type identification, monitoring area identification, and implementation priority, and the traffic flow regulation suggestion set is formed, each suggestion in the traffic flow regulation suggestion set includes abnormal evolution pattern type, corresponding monitoring area information, and optimized regulation strategy content. 9.The tunnel traffic flow big data mining method of claim 8, wherein, The tunnel monitoring area information is analyzed based on the tunnel monitoring area information, the range of the affected surrounding monitoring areas is determined by calculating the traffic flow interaction frequency between the target monitoring area and the surrounding monitoring areas, and the range of the affected surrounding monitoring areas is determined. retrieve traffic flow interaction data of the target monitoring area and all surrounding monitoring areas in a historical time period from a tunnel traffic management system, the traffic flow interaction data including a traffic flow value from the target monitoring area to a surrounding monitoring area and a traffic flow value from the surrounding monitoring area to the target monitoring area in a unit time; for each surrounding monitoring area, calculate a traffic flow interaction frequency of the target monitoring area and the surrounding monitoring area, the traffic flow interaction frequency being obtained by dividing a sum of the bidirectional traffic flow values in a unit time by a unit time length; compare the traffic flow interaction frequency of each surrounding monitoring area with a preset traffic flow interaction frequency threshold value; if the traffic flow interaction frequency of any surrounding monitoring area exceeds the traffic flow interaction frequency threshold value, determine that the surrounding monitoring area is a high correlation surrounding area possibly affected by an abnormal evolution mode; if the traffic flow interaction frequency of any surrounding monitoring area does not exceed the traffic flow interaction frequency threshold value, but the surrounding monitoring area is directly adjacent to the target monitoring area and no other monitoring area is interposed, determine that the surrounding monitoring area is a low correlation surrounding area possibly affected by an abnormal evolution mode; if the traffic flow interaction frequency of any surrounding monitoring area does not exceed the traffic flow interaction frequency threshold value, and the surrounding monitoring area is not directly adjacent to the target monitoring area, determine that the surrounding monitoring area is an unaffected surrounding area; integrate the high correlation surrounding area and the low correlation surrounding area to form a range of affected surrounding monitoring areas, and mark the correlation type of each affected surrounding monitoring area. 10.A tunnel traffic flow big data mining system combined with time series, characterized in that, The tunnel traffic flow big data mining system combined with time series includes a processor and a memory, the memory and the processor are connected, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to realize the tunnel traffic flow big data mining method combined with time series in any one of claims 1-9.

Images