Abnormal Location Identification Method, Device, Equipment and Storage Medium

Abstract

This application relates to artificial intelligence. This application provides an abnormal place recognition method, device, equipment and storage medium. The abnormal place recognition method combines a first input vector corresponding to the parameters of the place to be recognized itself (i.e., population information and status information) and a second input vector corresponding to the parameters between places (i.e., propagation probability) to generate an input vector representing the place to be recognized. Then, through a clustering analysis model, a place feature vector corresponding to the place to be recognized is obtained, and according to the comparison result between the place feature vector and the feature vector of a preset standard place, it is determined whether the place to be recognized is an abnormal place. Thus, by further adding parameters between places to construct a place feature vector representing the place to be recognized, not only the abnormal recognition accuracy of the place to be recognized is improved, but also an abnormal warning of the place is carried out based on the propagation rate between places, improving the abnormal recognition efficiency of the place to be recognized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular, to a method, device, equipment and computer-readable storage medium for identifying abnormal places. Background Art

[0002] The outbreak of infectious diseases not only affects the lives and health of the public, but also has a serious impact on national security. In order to prevent the outbreak of infectious diseases, it is necessary to identify whether abnormal situations such as infectious diseases will occur in a place. The current method for identifying abnormal places is to separately identify abnormalities in each place. Due to the characteristics of rapid outbreak and spread of infectious diseases, the current identification of abnormal places can only identify places where infectious diseases have occurred, and the identification efficiency is low. Therefore, how to improve the accuracy of identifying abnormal places has become an urgent problem to be solved at present. Summary of the Invention

[0003] The main object of the present invention is to provide a method, device, equipment and computer-readable storage medium for identifying abnormal places, aiming to improve the accuracy of identifying abnormal places.

[0004] To achieve the above object, the present invention provides a method for identifying abnormal places, the method for identifying abnormal places includes: obtaining population information and status information within a preset time period in a place to be identified, and generating a first input vector corresponding to the population information and the status information based on a preset conversion model; obtaining the transmission probability between the place to be identified and adjacent places, and generating a second input vector corresponding to the transmission probability based on the conversion model; generating a place feature vector corresponding to the first input vector and the second input vector based on a preset clustering analysis model, and determining the abnormal identification result of the place to be identified according to the feature vector of a preset standard place and the place feature vector.

[0005] In addition, to achieve the above object, the present invention also provides a device for identifying abnormal places, the device for identifying abnormal places includes: a place conversion module, configured to obtain population information and status information within a preset time period in a place to be identified, and generate a first input vector corresponding to the population information and the status information based on a preset conversion model; an inter-place conversion module, configured to obtain the transmission probability between the place to be identified and adjacent places, and generate a second input vector corresponding to the transmission probability based on the conversion model; an abnormal analysis module, configured to generate a place feature vector corresponding to the first input vector and the second input vector based on a preset clustering analysis model, and determine the abnormal identification result of the place to be identified according to the feature vector of a preset standard place and the place feature vector.

[0006] In addition, to achieve the above object, the present invention further provides an abnormal place recognition device, which includes a processor, a memory, and an abnormal place recognition program stored on the memory and executable by the processor. When the abnormal place recognition program is executed by the processor, the steps of the abnormal place recognition method as described above are implemented.

[0007] In addition, to achieve the above object, the present invention further provides a computer-readable storage medium, on which an abnormal place recognition program is stored. When the abnormal place recognition program is executed by a processor, the steps of the abnormal place recognition method as described above are implemented.

[0008] The present invention provides an abnormal place recognition method, which includes the following steps: obtaining population information and status information within a preset time period in a place to be recognized, and generating a first input vector corresponding to the population information and the status information based on a preset conversion model; obtaining the propagation probability between the place to be recognized and adjacent places, and generating a second input vector corresponding to the propagation probability based on the conversion model; generating a place feature vector corresponding to the first input vector and the second input vector based on a preset clustering analysis model, and determining the abnormal recognition result of the place to be recognized according to the feature vector of a preset standard place and the place feature vector. By the above method, the present invention combines the parameters of the place to be recognized itself (i.e., population information and status information) with the parameters between places (i.e., propagation probability) to generate an input vector representing the place to be recognized, then obtains the place feature vector corresponding to the place to be recognized through a clustering analysis model, and determines whether the place to be recognized is an abnormal place according to the comparison result between the place feature vector and the feature vector of the preset standard place. Thus, further adding the parameters between places to construct the place feature vector representing the place to be recognized not only improves the abnormal recognition accuracy of the place to be recognized, but also performs abnormal warning of the place based on the propagation rate between places, improving the abnormal recognition efficiency of the place to be recognized. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a schematic hardware structure diagram of the abnormal place recognition device involved in the embodiment solution of the present invention;

[0010] Figure 2 It is a schematic flowchart of the first embodiment of the abnormal place recognition method of the present invention;

[0011] Figure 3 It is a schematic flowchart of the second embodiment of the abnormal place recognition method of the present invention;

[0012] Figure 4 It is a schematic diagram of the functional modules of the first embodiment of the abnormal place recognition device of the present invention.

[0013] The implementation, functional features and advantages of the object of the present invention will be further described with reference to the embodiments and the accompanying drawings. Detailed implementation manners

[0014] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0015] The abnormal place recognition method involved in the embodiment of the present invention is mainly applied to an abnormal place recognition device, and the abnormal place recognition device can be a device with display and processing functions such as a PC, a portable computer, a mobile terminal, etc.

[0016] Referring to Figure 1 , Figure 1 is a schematic hardware structure diagram of the abnormal place recognition device involved in the solution of the embodiment of the present invention. In the embodiment of the present invention, the abnormal place recognition device may include a processor 1001 (such as a CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components; the user interface 1003 may include a display screen (Display) and an input unit such as a keyboard (Keyboard); the network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface); the memory 1005 may be a high-speed RAM memory or a stable memory (non-volatile memory), such as a disk memory, and the memory 1005 may optionally be a storage device independent of the aforementioned processor 1001.

[0017] Those skilled in the art can understand that Figure 1 the hardware structure shown in does not constitute a limitation on the abnormal place recognition device, and may include more or fewer components than shown in the figure, or combine some components, or different component arrangements.

[0018] Continuing to refer to Figure 1 , Figure 1 the memory 1005 as a computer-readable storage medium in may include an operating system, a network communication module, and an abnormal place recognition program.

[0019] In Figure 1 , the network communication module is mainly used to connect to the server and perform data communication with the server; and the processor 1001 can call the abnormal place recognition program stored in the memory 1005 and execute the abnormal place recognition method provided by the embodiment of the present invention.

[0020] The embodiment of the present invention provides an abnormal place recognition method.

[0021] Reference Figure 2 , Figure 2 is a schematic flowchart of the first embodiment of the abnormal place recognition method of the present invention.

[0022] In this embodiment, the abnormal place recognition method includes the following steps:

[0023] Step S10, obtain the population information and status information of the place to be recognized within a preset time period, and generate a first input vector corresponding to the population information and the status information based on a preset conversion model.

[0024] In this embodiment, the preset time period refers to the data statistics within the selected time of the place to be recognized, which can be one hour, one day, seven days, eight days, nine days, etc. When actually selecting, the appropriate number of days can be selected according to the monitoring needs.

[0025] Specifically, within each monitoring time window t (for example, t is a day, that is, monitoring the data of one day), the population information and the status information of the place to be recognized are obtained from two aspects. One aspect is the demographic statistics data of the floating population and resident population within the place to be recognized. That is, within the monitoring time window t, the gender structure, age structure, etc. of the floating population and resident population within the place to be recognized are counted. Among them, the place to be recognized can be a place with a large population gathering or a sealed place such as a school, an institution, an enterprise or a public institution, a library, a museum, a gym, a cinema, a construction site, a shopping mall, etc. For the place to be recognized, that is, a school, an office building, etc., demographic data can be obtained. The other aspect is the distribution of the symptoms that need to be focused on within the place to be recognized. The symptoms that need to be focused on (that is, the target symptoms of infectious disease monitoring, the existing symptom list sorted out by experts), that is, the distribution of symptoms such as fever, headache, projectile vomiting, shock, etc. Among them, the symptom data is obtained from the symptoms reported by the people within the place to be recognized. When the place to be recognized is a school, students asking for sick leave will describe the typical symptoms. When the place to be recognized is an office building, employees asking for sick leave will describe the typical symptoms.

[0026] The preset conversion model is an autoencoder neural network. The autoencoder neural network (autoencoder, AE) is a type of artificial neural network (Artificial Neural Networks, ANNs) used in semi-supervised learning and unsupervised learning. The function of the autoencoder neural network is to perform representation learning on the input information by taking the input information as the learning target. The autoencoder neural network consists of two parts: an encoder and a decoder. According to the learning paradigm, the autoencoder neural network can be divided into a contractive autoencoder, a regularized autoencoder, and a Variational AutoEncoder (VAE). Among them, the first two are discriminative models, and the latter is a generative model. According to the construction type, the autoencoder neural network can be a feedforward structure or a recursive structure neural network. The autoencoder neural network is applied to dimensionality reduction and anomaly detection. The autoencoder neural network containing convolutional layer construction can be applied to computer vision problems, including image denoising, neural style transfer, etc.

[0027] Specifically, the population information and the status information are combined and spliced into a first model input quantity, and the first model input quantity is input into the autoencoder neural network. The encoder network in the autoencoder neural network encodes the first model input quantity and maps the first model input quantity to a first intermediate vector with a lower dimension. The decoder network in the autoencoder neural network decodes the first vector and maps the first intermediate vector back to a first output vector with the same dimension as the first model input quantity. Among them, the first intermediate vector is used as the first input vector.

[0028] Step S20, obtain the propagation probability between the to-be-recognized venue and the adjacent venues, and generate a second input vector corresponding to the propagation probability based on the conversion model.

[0029] Specifically, for the probability of infectious disease transmission between the to-be-identified venue and any adjacent venue, the transmission probability is calculated based on the characteristic data in three aspects. The first aspect is the distance between the to-be-identified venue and the adjacent venue, which is represented by the straight-line distance between the to-be-identified venue and the adjacent venue. The closer the distance between the to-be-identified venue and the adjacent venue, the higher the transmission probability. The second aspect is the traffic condition between the to-be-identified venue and the adjacent venue, which is represented by the number of public transportation routes between the to-be-identified venue and the adjacent venue. The more public transportation routes, the higher the transmission probability. The third aspect is the population condition between the to-be-identified venue and the adjacent venue, which is represented by the sum of the populations of the to-be-identified venue and the adjacent venue. The larger the population, the higher the transmission probability. The probability of infectious disease transmission between the to-be-identified venue and the adjacent venue is the sum of the product of the distance value and the distance weight, the product of the number of connected routes and the public transportation weight, the product of the total population and the population weight, and the initial weight. Among them, the weight is the proportion probability represented by the autoencoder neural network after training for the three aspects of distance, routes, and total number of entrances.

[0030] It can be understood that when there is only one adjacent venue of the to-be-identified venue, the transmission probability between the to-be-identified venue and the adjacent venue is unique; when there are multiple adjacent venues of the to-be-identified venue, the transmission probability between the to-be-identified venue and the adjacent venues is a probability set, and the probability set includes multiple transmission probabilities.

[0031] In this embodiment, the transmission probability between the to-be-identified venue and the adjacent venue is converted into a second model input quantity, and the second model input quantity is input into the autoencoder neural network. The encoder network in the autoencoder neural network encodes the second model input quantity and maps the second model input quantity to a second intermediate vector with a lower dimension. The decoder network in the autoencoder neural network decodes the second intermediate vector and maps the second intermediate vector back to a second output vector with the same dimension as the second model input quantity. Among them, the second intermediate vector is used as the second input vector.

[0032] Step S30: Based on a preset clustering analysis model, generate a venue feature vector corresponding to the first input vector and the second input vector, and determine the abnormal recognition result of the to-be-identified venue according to the feature vector of the preset standard venue and the venue feature vector.

[0033] Specifically, based on the DBSCAN clustering algorithm, the first input vector and the second input vector are input into the DBSCAN clustering algorithm for calculation to obtain the venue feature vector. The venue feature vector is compared with the feature vector of a preset standard venue. If the comparison result is greater than a preset value, it is determined that the venue to be identified corresponding to the venue feature vector is an abnormal venue.

[0034] Among them, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm is a density-based spatial clustering algorithm. This algorithm divides regions with sufficient density into clusters and discovers clusters of arbitrary shapes in a spatial database with noise. It defines a cluster as the largest set of density-connected points.

[0035] The DBSCAN algorithm is based on the fact that a cluster can be uniquely determined by any of its core objects. Equivalent statement: For any data object p that satisfies the core object condition, the set of all data objects o that are density-reachable from p in the database D forms a complete cluster C, and p belongs to C.

[0036] Furthermore, the specific clustering process is as follows:

[0037] Scan the entire dataset to find any core point, and expand this core point. The expansion method is to find all density-connected data points starting from the core point, traverse all core points within the ε-neighborhood of this core point (because border points cannot be expanded), and find points that are density-connected to these data points until there are no more data points to expand. Finally, the boundary nodes of the clusters formed by clustering are all non-core data points. Then, rescan the dataset (excluding any data points in the clusters found previously), find core points that have not been clustered, and repeat the above steps to expand this core point until there are no new core points in the dataset. The data points in the dataset that are not included in any cluster form the abnormal points.

[0038] The preset standard venue features are obtained by inputting the population information, status information of a large number of normal venues, and the propagation probability between normal venues into the DBSCAN clustering algorithm during the training of the DBSCAN clustering algorithm. The venue features representing normal venues are the standard venue features.

[0039] This embodiment provides an abnormal place recognition method, and the abnormal place recognition method includes the following steps: obtaining population information and status information within a preset time period in the place to be recognized, and generating a first input vector corresponding to the population information and the status information based on a preset conversion model; obtaining the propagation probability between the place to be recognized and adjacent places, and generating a second input vector corresponding to the propagation probability based on the conversion model; generating a place feature vector corresponding to the first input vector and the second input vector based on a preset clustering analysis model, and determining the abnormal recognition result of the place to be recognized according to the feature vector of a preset standard place and the place feature vector. By the above method, the present invention combines the parameters of the place to be recognized itself (i.e., population information and status information) with the parameters between places (i.e., propagation probability) to generate an input vector representing the place to be recognized, and then obtains the place feature vector corresponding to the place to be recognized through a clustering analysis model. According to the comparison result between the place feature vector and the feature vector of the preset standard place, it is determined whether the place to be recognized is an abnormal place. Thus, further adding the parameters between places to construct the place feature vector representing the place to be recognized not only improves the abnormal recognition accuracy of the place to be recognized, but also performs abnormal warning of the place based on the propagation rate between places, improving the abnormal recognition efficiency of the place to be recognized.

[0040] Refer to Figure 3 , Figure 3 which is a schematic flowchart of the second embodiment of the abnormal place recognition method of the present invention.

[0041] Based on the above Figure 2 shown embodiment, in this embodiment, the step S20 further includes:

[0042] Step S21, obtaining the distance value, the number of connected routes, and the total population between the place to be recognized and adjacent places;

[0043] Specifically, the characteristic data between the to-be-identified venue and adjacent venues is obtained through three aspects. The characteristic data includes distance value, number of connected routes, and total population. The first aspect is the distance situation between the to-be-identified venue and adjacent venues, which is represented by the straight-line distance d between the to-be-identified venue and adjacent venues. The closer the distance between the to-be-identified venue and adjacent venues, the higher the transmission probability. The second aspect is the traffic situation between the to-be-identified venue and adjacent venues, which is represented by the number of public transportation routes l between the to-be-identified venue and adjacent venues. The more public transportation routes, the higher the transmission probability. The third aspect is the population situation between the to-be-identified venue and adjacent venues, which is represented by the sum of the populations o of the to-be-identified venue and adjacent venues. The larger the population, the higher the transmission probability. For example, the straight-line distance between the to-be-identified venue and adjacent venues is 1 kilometer, the number of public transportation routes between the to-be-identified venue and adjacent venues is 7, and the sum of the populations of the to-be-identified venue and adjacent venues is 3000 people.

[0044] Step S22, calculate according to a preset formula and the distance value, the number of connected routes, and the total population to obtain the transmission probability between the to-be-identified venue and adjacent venues. The preset formula is:

[0045]

[0046] is the transmission probability between the to-be-identified venue i and an adjacent venue j, is the preset initial weight, is the distance value between i and j, is the weight corresponding to the distance value, is the number of connected routes between i and j, is the weight corresponding to the number of connected routes, is the total population of i and j, is the weight corresponding to the total population.

[0047] Specifically, the infectious disease transmission probability p between the to-be-identified venue and adjacent venues (such as key venue i and key venue j) ij = w0 + w1d ij + w2l ij + w3o ij where w0 is the initial weight, w1 is the distance weight, w2 is the public transportation weight, and w3 is the population weight. When the straight-line distance between the to-be-identified venue and adjacent venues is 1 kilometer, the number of public transportation routes between the to-be-identified venue and adjacent venues is 7, and the sum of the populations of the to-be-identified venue and adjacent venues is 3000 people, the transmission probability p between the to-be-identified venue and adjacent venues ij= w0 + 1000w1 + 7w2 + 3000w3。

[0048] Step S23: Extract features from the propagation probability according to the pre-trained autoencoder neural network to obtain a second intermediate vector, and use the second intermediate vector as the second input vector.

[0049] In this embodiment, for a place i to be recognized, construct an infectious disease propagation probability vector pi = (pi1, pi2, pi3,..., pip) from p other adjacent places to this place to be recognized. Similarly, construct an autoencoder neural network, which consists of an encoder and a decoder network structure. First, map the second model input quantity to a second intermediate vector h = (h1, h2, h3,..., h q ) with a lower dimension (for example, q dimensions, where q is much smaller than p) through the encoder network, and then map the second intermediate vector back to the second output vector through the decoder network. That is, both the second model input quantity and the second output vector are pi = (pi1, pi2, pi3,..., pip). Through the autoencoder neural network structure, convert high-dimensional data into low-dimensional data, capture significant features from the data, and finally use the second intermediate vector h = (h1, h2, h3,..., h q ) as the second input vector.

[0050] Based on the above Figure 2 shown embodiment, in this embodiment, the step of determining the anomaly recognition result of the place to be recognized according to the place feature vector and the feature vector of the preset standard place in step S30 specifically includes:

[0051] Step S31, convert the place feature vector into a place coordinate point, and calculate the shortest distance between the place coordinate point and the preset standard coordinate point set, where the standard coordinate point set is the coordinate point set corresponding to the feature vector of the standard place;

[0052] Step S32, determine the anomaly recognition result of the place to be recognized according to the comparison result between the shortest distance and the preset value.

[0053] Specifically, convert the place feature vector into a place coordinate point so that the place feature vector can be displayed as a coordinate point in the coordinate system, where the place coordinate point corresponds to the place to be recognized;

[0054] Calculate the shortest distance between the coordinates of the venue and the set of standard coordinate points, compare the shortest distance with a preset value. If the shortest distance is greater than the preset value, it is determined that the coordinates of the venue do not belong to the set of standard coordinate points, and the coordinates of the venue belong to the outlier points, that is, the abnormal points, then it is determined that the venue to be identified belongs to the abnormal venue.

[0055] Further, step S32 further includes:

[0056] Calculate the Euclidean distance between the coordinates of the venue and each coordinate point in the set of standard coordinate points, and select the minimum Euclidean distance from the obtained Euclidean distances as the shortest distance.

[0057] Specifically, calculate the Euclidean distance between the coordinates of the venue and each coordinate point in the set of standard coordinate points, obtain the calculation results, and select the one with the smallest Euclidean distance value in the calculation results as the shortest distance between the coordinates of the venue and the set of preset standard coordinate points.

[0058] Based on the above Figure 2 In the shown embodiment, in this embodiment, generating the first input vector corresponding to the population information and the status information based on the preset conversion model specifically includes:

[0059] Step S11, perform feature extraction on the population information and the status information according to the pre-trained autoencoder neural network to obtain a first intermediate vector, and use the first intermediate vector as the first input vector.

[0060] Among them, the acquisition channels of the population information and the status information include: for each monitoring time window t (for example, t is in days, that is, monitoring the data of 1 day), obtain the characteristics of the venue to be identified from two aspects. One aspect is the demographic statistical data of the flowing and resident population in the venue to be identified, that is, in the monitoring time window t, count the gender structure, age structure, etc. of the flowing and resident population in the venue to be identified. Among them, the venue to be identified can be a school, government agencies, enterprises and institutions, library, museum, gym, cinema, construction site, shopping mall, etc. where there is a large population gathering or the venue is sealed. For the venue to be identified, that is, schools, office buildings, etc., demographic data can be obtained. Another aspect is the distribution of the symptoms that need to be focused on in the venue to be identified. The symptoms that need to be focused on (that is, the target symptoms for infectious disease monitoring, the existing symptom list sorted out by experts), that is, the distribution of symptoms such as fever, headache, projectile vomiting, shock, etc.

[0061] In this embodiment, construct m demographic statistical features d = (d1, d2, d3,..., d in the venue to be identified from each monitoring time window t m)( ) and n key symptom monitoring features s = (s1, s2, s3,..., s n ) as the first model input quantity, that is, x = (d1, d2, d3,..., d m , s1, s2, s3,..., s n ), construct an autoencoder neural network. First, map the first model input quantity to a first intermediate vector v = (v1, v2, v3, v k ) with a lower dimension (e.g., k dimensions, where k is much smaller than m + n) through the encoder network, and then map the first intermediate vector back to the first output vector through the decoder network. That is, both the first input vector and the first output vector are x = (d1, d2, d3,..., d m , s1, s2, s3,..., s n ). Through the autoencoder neural network structure, convert high-dimensional data into low-dimensional data, capture significant features from the data, and finally use the first intermediate vector v = (v1, v2, v3, v k ) as the vector representation of the place to be recognized itself.

[0062] Among them, converting high-dimensional data into low-dimensional data is carried out through the constructed autoencoder neural network based on deep learning. The autoencoder network consists of an encoder and a decoder. Specifically, first map the 1000-dimensional input vector x to a 100-dimensional vector y through the encoder network, and then map y back to a 1000-dimensional output vector x through the decoder network. That is, the input and output of the autoencoder network are the same, both being x. Learn the network parameters in the autoencoder neural network in an unsupervised manner, and use the output y of the encoder in the autoencoder neural network as the vector representation y = (y1, y2, y3, y k ) of the place to be recognized itself.

[0063] Furthermore, generating the first input vector corresponding to the population information and the status information based on the preset conversion model includes:

[0064] Based on the preset conversion model, construct the features corresponding to the population information and the features corresponding to the status information. Among them, the features corresponding to the population information include the gender distribution and age group distribution of the population in the place to be recognized, and the features corresponding to the status information include each symptom information in the preset symptom list;

[0065] Generate the first input vector according to the features corresponding to the population information and the features corresponding to the status information

[0066] In this embodiment, the age of a person is 30 and the gender is male (more demographic information is omitted here). The demographic information is the demographic statistical feature d = (d1, d2, d3, ..., d m ) specifically: (age range [0, 1), age range [1, 6), age range [6, 14), age range [14, 45), age range [45, 65), age range [65, +∞), male gender, female gender, student. For the information of the person in the example, the demographic feature corresponds to (0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 1, 0, …), where 0 indicates no corresponding demographic information and 1 indicates corresponding demographic information. The two 1s in the feature vector represent the age range [14, 45) and male gender in sequence.

[0067] A person's symptoms are fever and headache. The status information is the symptom monitoring feature s = (s1, s2, s3, ..., s n ) specifically (fever, headache, diarrhea, chest tightness, …). For the person in the example, the corresponding symptom monitoring feature is (1, 1, 0, 0, …). Among them, 0 indicates no corresponding symptom and 1 indicates corresponding symptom. The two 1s in the symptom monitoring feature represent the symptoms of fever and headache in sequence.

[0068] Construct m demographic statistical features d = (d1, d2, d3, ..., d m ) and n key symptom monitoring features s = (s1, s2, s3, ..., s n ) within each monitoring time window t in the place to be identified as the first model input quantity, that is, x = (d1, d2, d3, ..., d m , s1, s2, s3, ..., s n ). Construct an autoencoder neural network. First, map the first model input quantity to a first intermediate vector v = (v1, v2, v3, v k ) with a lower dimension (for example, k dimensions, where k is much smaller than m + n) through the encoder network, and then map the first intermediate vector back to the first output vector through the decoder network, that is, both the first input vector and the first output vector are x = (d1, d2, d3, ..., d m , s1, s2, s3, ..., s n ). Through the autoencoder neural network structure, convert high-dimensional data into low-dimensional data, capture significant features from the data, and finally use the first intermediate vector v = (v1, v2, v3, v k ) as the vector representation of the place to be identified itself, that is, the first input vector.

[0069] Further, generating the place feature vectors corresponding to the first input vector and the second input vector based on the preset clustering analysis model includes:

[0070] Reducing the dimensions of the first input vector and the second input vector based on the DBSCAN clustering algorithm to generate the place feature vectors.

[0071] Specifically, comprehensively using the vectors of the place to be identified itself and between adjacent places as the final vector representation of the place to be identified. That is, for the key place i, the vector representation is w i = (v i , h i ) = (v i1 , v i2 , v i3 ,..., v ik , h i1 , h i2 , h i3 ,..., h ip ). Apply the DBSCAN (Density-based spatial clustering of applications with noise) clustering algorithm to establish a clustering analysis model, and use the identified outliers as abnormal key places for early warning.

[0072] The applied DBSCAN clustering algorithm is an unsupervised learning algorithm. By this method, cluster the places to be identified represented by feature vectors. Generally, a large number of normal places have similar features and will be clustered together, while the features of abnormal places are inconsistent with those of normal places and will not be clustered into one category. Usually, they are clustered as outliers into one category, and early warning for infectious diseases is carried out for such abnormal places.

[0073] In addition, an abnormal place identification device is further provided in an embodiment of the present invention.

[0074] Refer to Figure 4 , Figure 4 which is a schematic diagram of the functional modules of the first embodiment of the abnormal place identification device of the present invention.

[0075] In this embodiment, the abnormal place identification device includes:

[0076] A place conversion module 10, configured to obtain the population information and status information within a preset time period in the place to be identified, and generate a first input vector corresponding to the population information and the status information based on a preset conversion model;

[0077] An inter-place conversion module 20, configured to obtain the transmission probability between the place to be identified and adjacent places, and generate a second input vector corresponding to the transmission probability based on the conversion model;

[0078] Anomaly analysis module 30 is configured to generate a place feature vector corresponding to the first input vector and the second input vector based on a preset clustering analysis model, and determine an anomaly recognition result of the place to be recognized according to the feature vector of the preset standard place and the place feature vector.

[0079] Further, the place conversion module 20 specifically includes:

[0080] A place information collection unit is configured to obtain the distance value, the number of connected routes, and the total population between the place to be recognized and adjacent places.

[0081] A place calculation unit is configured to calculate, according to a preset formula and the distance value, the number of connected routes, and the total population, a propagation probability between the place to be recognized and an adjacent place, where the preset formula is:

[0082]

[0083] is the propagation probability between the place to be recognized i and an adjacent place j, is a preset initial weight, is the distance value between i and j, is the weight corresponding to the distance value, is the number of connected routes between i and j, is the weight corresponding to the number of connected routes, is the total population of i and j, is the weight corresponding to the total population.

[0084] A second feature dimension reduction unit is configured to perform feature extraction on the propagation probability according to a pre-trained autoencoder neural network to obtain a second intermediate vector, and use the second intermediate vector as the second input vector.

[0085] Further, the anomaly analysis module 30 specifically includes:

[0086] A distance calculation unit is configured to convert the place feature vector into a place coordinate point and calculate the shortest distance between the place coordinate point and a preset standard coordinate point set, where the standard coordinate point set is a coordinate point set corresponding to the feature vector of the standard place;

[0087] An identification unit is configured to determine an anomaly recognition result of the place to be recognized according to a comparison result between the shortest distance and a preset value.

[0088] Further, the distance calculation unit includes:

[0089] A distance analysis unit, configured to calculate the Euclidean distance between the coordinates of the venue and each coordinate in the set of standard coordinates, and select the minimum Euclidean distance from the obtained Euclidean distances as the shortest distance.

[0090] Further, the venue conversion module 10 specifically includes:

[0091] A first feature dimensionality reduction unit, configured to perform feature extraction on the population information and the status information according to a pre-trained autoencoder neural network to obtain a first intermediate vector, and use the first intermediate vector as the first input vector.

[0092] Further, the venue conversion module 10 specifically includes:

[0093] An information feature acquisition unit, configured to construct the features corresponding to the population information and the features corresponding to the status information based on a preset conversion model, where the features corresponding to the population information include the gender distribution and age group distribution of the population in the venue to be recognized, and the features corresponding to the status information include each symptom information in a preset symptom list.

[0094] An information feature dimensionality reduction unit, configured to generate the first input vector according to the features corresponding to the population information and the features corresponding to the status information.

[0095] Further, the abnormal venue recognition device includes a clustering analysis module, and the clustering analysis module includes:

[0096] A feature extraction unit, configured to reduce the dimensionality of the first input vector and the second input vector based on the DBSCAN clustering algorithm to generate the venue feature vector.

[0097] Wherein, each module in the above abnormal venue recognition device corresponds to each step in the embodiment of the above abnormal venue recognition method, and its functions and implementation processes will not be elaborated here one by one.

[0098] In addition, an embodiment of the present invention further provides a computer-readable storage medium.

[0099] An abnormal venue recognition program is stored on the computer-readable storage medium of the present invention. When the abnormal venue recognition program is executed by a processor, the steps of the abnormal venue recognition method as described above are implemented.

[0100] Wherein, the method implemented when the abnormal venue recognition program is executed can refer to each embodiment of the abnormal venue recognition method of the present invention, and will not be elaborated here.

[0101] It should be noted that in this text, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, such that a process, method, article or system comprising a series of elements not only includes those elements but also other elements not expressly listed, or elements inherent to such process, method, article or system. Without further limitation, an element defined by the phrase "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or system comprising that element.

[0102] The serial numbers of the above embodiments of the present invention are for description only and do not represent the superiority or inferiority of the embodiments.

[0103] This application can be used in numerous general-purpose or special-purpose computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This application can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0104] Through the description of the above embodiments, those skilled in the art can clearly understand that the above-described embodiment methods can be implemented by means of software plus a necessary general-purpose hardware platform. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art can be embodied in the form of a software product. This computer software product is stored in a storage medium as described above (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to enable a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in various embodiments of the present invention.

[0105] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be equally included in the patent protection scope of the present invention.

Claims

1. An abnormal place recognition method, characterized in that, The abnormal place recognition method includes the following steps: Obtain the population information and status information within the place to be recognized during a preset time period, and generate a first input vector corresponding to the population information and the status information based on a preset conversion model; Obtain the propagation probability between the place to be recognized and adjacent places, and generate a second input vector corresponding to the propagation probability based on the conversion model; Generate a place feature vector corresponding to the first input vector and the second input vector based on a preset clustering analysis model, and determine the abnormal recognition result of the place to be recognized according to the feature vector of the preset standard place and the place feature vector; Among them, the obtaining of the propagation probability between the place to be recognized and adjacent places includes: Obtain the distance value, the number of connected routes, and the total population between the place to be recognized and adjacent places; Calculate according to a preset formula and the distance value, the number of connected routes, and the total population to obtain the propagation probability between the place to be recognized and adjacent places, where the preset formula is: is the propagation probability between the to-be-recognized venue i and an adjacent venue j, is the preset initial weight, is the distance value between i and j, is the weight corresponding to the distance value, is the number of connected routes between i and j, is the weight corresponding to the number of connected routes, is the total population of i and j, is the weight corresponding to the total population.

2. The abnormal place recognition method according to claim 1, wherein The determining of the abnormal recognition result of the place to be recognized according to the place feature vector and the feature vector of the preset standard place includes: Convert the place feature vector into a place coordinate point, and calculate the shortest distance between the place coordinate point and a preset standard coordinate point set, where the standard coordinate point set is a coordinate point set corresponding to the feature vector of the standard place; Determine the abnormal recognition result of the place to be recognized according to the comparison result between the shortest distance and a preset value.

3. The abnormal place recognition method according to claim 2, wherein, The calculating of the shortest distance between the place coordinate point and the preset standard coordinate point set further includes: Calculate the Euclidean distance between the place coordinate point and each coordinate point in the standard coordinate point set, and select the minimum Euclidean distance from the obtained Euclidean distances as the shortest distance.

4. The abnormal place recognition method according to claim 1, wherein The generating of the first input vector corresponding to the population information and the status information based on a preset conversion model includes: Extract features from the population information and the status information according to a pre-trained autoencoder neural network to obtain a first intermediate vector, and use the first intermediate vector as the first input vector.

5. The abnormal place recognition method according to claim 4, wherein The generating of the first input vector corresponding to the population information and the status information based on a preset conversion model includes: Based on a preset conversion model, construct the features corresponding to the population information and the features corresponding to the status information, where the features corresponding to the population information include the gender distribution and age group distribution of the population within the place to be recognized, and the features corresponding to the status information include each symptom information in a preset symptom list; Generate the first input vector according to the features corresponding to the population information and the features corresponding to the status information.

6. The abnormal place recognition method according to any one of claims 1-5, characterized in that The generating of the place feature vector corresponding to the first input vector and the second input vector based on a preset clustering analysis model includes: Reduce the dimensions of the first input vector and the second input vector based on the DBSCAN clustering algorithm to generate the place feature vector.

7. An abnormal place recognition device, characterized in that, The abnormal place recognition device includes: A venue conversion module, configured to obtain population information and status information within a preset time period in a venue to be recognized, and generate a first input vector corresponding to the population information and the status information based on a preset conversion model; An inter-venue conversion module, configured to obtain the propagation probability between the venue to be recognized and an adjacent venue, and generate a second input vector corresponding to the propagation probability based on the conversion model; An anomaly analysis module, configured to generate a venue feature vector corresponding to the first input vector and the second input vector based on a preset clustering analysis model, and determine an anomaly recognition result of the venue to be recognized according to the feature vector of a preset standard venue and the venue feature vector; Wherein, the inter-venue conversion module is further configured to: Obtain the distance value, the number of connected routes, and the total population between the venue to be recognized and the adjacent venue; Calculate according to a preset formula and the distance value, the number of connected routes, and the total population to obtain the propagation probability between the venue to be recognized and the adjacent venue, where the preset formula is: is the propagation probability between the to-be-recognized venue i and an adjacent venue j, is the preset initial weight, is the distance value between i and j, is the weight corresponding to the distance value, is the number of connected routes between i and j, is the weight corresponding to the number of connected routes, is the total population of i and j, is the weight corresponding to the total population.

8. An abnormal place recognition device, characterized in that, The anomaly venue recognition device includes a processor, a memory, and an anomaly venue recognition program stored on the memory and executable by the processor. When the anomaly venue recognition program is executed by the processor, the steps of the anomaly venue recognition method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium, characterized in that, An anomaly venue recognition program is stored on the computer-readable storage medium. When the anomaly venue recognition program is executed by a processor, the steps of the anomaly venue recognition method according to any one of claims 1 to 6 are implemented.

Images