Group screening method, system, terminal device, and medium

Abstract

The application discloses a population screening method and system, a terminal device and a computer readable storage medium. The population screening method comprises the following steps: determining a resident population whose staying time in a to-be-screened area exceeds a preset time threshold, and obtaining a feature vector of a travel trajectory of the resident population; clustering the travel trajectory of the resident population based on the feature vector to obtain a student population in the resident population, so as to determine a graduate group from the student population. The application can realize accurate screening of the graduate group.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a group screening method, system, terminal device, and computer-readable storage medium. Background Technology

[0002] In order to conduct research and analysis on the employment status of graduates, the existing methods for screening and identifying graduates are generally based on mobile signaling big data, combined with the location of the university, age parameters in the user profile, and device identifiers of the user's devices (such as mobile phones), to screen out the student group, and then determine the graduate group from the student group.

[0003] However, the screening results of this group screening method are very rough.

[0004] For example, the user profile table mentioned above may have some missing data, making it impossible to associate with the device identification code in the mobile signaling big data. It may also have some missing age data, leading to a large error in screening student users. Among college students, there are some younger students in the gifted youth program and some older students. Screening by age alone will result in errors. Some students use family members' supplementary SIM cards, and the identity information registered on the supplementary SIM cards is not that of the students. Screening by age alone will ignore these students, causing errors. Summary of the Invention

[0005] The main objective of this invention is to provide a group screening method, system, terminal device, and computer-readable storage medium, which aims to achieve accurate screening of student groups and thereby improve the screening accuracy of graduate groups.

[0006] To achieve the above objectives, the present invention provides a group screening method, wherein the group screening includes:

[0007] Identify the resident population whose stay in the area to be screened exceeds a preset time threshold, and obtain the representation vector of the travel trajectory of the resident population;

[0008] Based on the representation vector, the travel trajectories of the resident population are clustered to obtain the student population within the resident population, so as to identify the graduate group from the student population.

[0009] Optionally, the step of obtaining the representation vector of the travel trajectory of the resident population includes:

[0010] Obtain the user travel zipper table of each user in the permanent population of the area to be filtered. The user travel zipper table includes multiple user travel trajectories and corresponding user travel times.

[0011] According to the user's travel time, sort the multiple user travel trajectories to obtain the user travel trajectory sequence for each user;

[0012] The user's travel trajectory sequence is used as input to a preset neural network to obtain the representation vector of the user's travel trajectory sequence output by the hidden layer of the preset neural network.

[0013] Optionally, before the step of using the user's travel trajectory sequence as input to a preset neural network and obtaining the representation vector of the user's travel trajectory sequence output by the hidden layer of the preset neural network, the method further includes:

[0014] Extract each trajectory point from the user's travel trajectory sequence and construct a trajectory point dictionary;

[0015] Any trajectory point in the trajectory point dictionary is taken as the first trajectory point, and other trajectory points other than the first trajectory point are taken as the second trajectory point.

[0016] Construct a trajectory point pair containing the first trajectory point and the second trajectory point;

[0017] The trajectory point pairs are used as training data and input into the preset neural network to obtain the trajectory prediction results output by the output layer of the preset neural network. Based on the trajectory prediction results, the weight matrix of the hidden layer is trained. The trajectory prediction results include the probability that the output trajectory point is the second trajectory point when the input trajectory point is the first trajectory point.

[0018] Optionally, the step of clustering the travel trajectories of the resident population based on the representation vector to obtain the student population within the resident population includes:

[0019] Based on the representation vector, multiple initial centroids are determined, and according to the multiple initial centroids, the representation vector is divided into corresponding initial trajectory point clusters;

[0020] Obtain the distance between the representation vector of each trajectory point in the initial trajectory point cluster and each initial centroid, and update the trajectory points in each initial trajectory point cluster based on the multiple distances, so as to iterate the multiple initial trajectory point clusters to obtain multiple converged trajectory point clusters;

[0021] Obtain the number of representation vectors in each converged trajectory point cluster, and take the trajectory point cluster with the most vectors as the target trajectory point cluster;

[0022] The resident population corresponding to the target trajectory point cluster is considered as the student population.

[0023] Optionally, after the step of clustering the travel trajectories of the resident population based on the representation vector to obtain the student population within the resident population, the method further includes:

[0024] Determine the time periods during which students stay on campus and the time periods during which they leave campus within the university area, wherein the time periods during which students stay on campus include the first and second time periods of different semesters of the academic year;

[0025] The total student population is obtained by statistically analyzing the student population within the filter area during the first time period.

[0026] Remove distracting individuals from the entire student population, wherein the distracting individuals are located in the screening area during the school departure period;

[0027] The graduate group is obtained from the entire student population after removing the interference group, and the students who are not in the screening area during the second time period.

[0028] Optionally, the region to be screened includes: a university area, and before the step of obtaining the representation vector of the resident population within the region to be screened, the method further includes:

[0029] Determine a detailed list of base stations in university areas nationwide and a travel zipper list for university students nationwide;

[0030] The invalid base station data in the user travel zipper table is removed to obtain the processed user travel zipper table;

[0031] Based on the processed user travel zipper table and the base station identifier in the national university area base station details table, the processed user travel zipper table is associated with the national university area base station details table;

[0032] Based on the latitude and longitude of the regional boundaries of the university areas, filter out the university activity population from the associated user travel zipper table and the national university area base station details table;

[0033] Based on the length of time the university students spend in the preset screening area, the resident population in the screening area and the corresponding travel trajectories of the resident population are determined.

[0034] Optionally, the step of removing invalid base station data from the user travel zipper table to obtain the processed user travel zipper table includes:

[0035] The system obtains the switching frequency of the user's device switching back and forth between two communication base stations, and when the switching frequency is greater than a preset frequency threshold, it obtains the first base station data corresponding to the two communication base stations.

[0036] Obtain data from a second base station that is not located within the area of ​​the university;

[0037] The first base station data and the second base station data are removed from the user travel zipper table to obtain the processed user travel zipper table.

[0038] To achieve the above objectives, the present invention also provides a group screening system, the group screening system comprising:

[0039] The acquisition module is used to identify resident populations whose stay duration exceeds a preset time threshold within the area to be screened, and to acquire the representation vector of the travel trajectory of the resident populations.

[0040] The filtering module is used to cluster the travel trajectories of the resident population based on the representation vector to obtain the student population among the resident population, so as to determine the graduate group from the student population.

[0041] In this invention, each functional module of the population screening system implements the steps of the population screening method described above during operation.

[0042] To achieve the above objectives, the present invention also provides a terminal device, the terminal device comprising: a memory, a processor, and a group screening program stored in the memory and executable on the processor, wherein the group screening program, when executed by the processor, implements the steps of the group screening method as described above.

[0043] Furthermore, to achieve the above objectives, the present invention also proposes a computer-readable storage medium storing a crowd screening program, which, when executed by a processor, implements the steps of the crowd screening method as described above.

[0044] In addition, to achieve the above objectives, the present invention also provides a computer program product comprising a computer program that, when executed by a processor, implements the steps of the population screening method as described above.

[0045] This invention provides a group screening system, terminal device, computer-readable storage medium, and computer program product. It identifies resident populations whose stay in a screening area exceeds a preset time threshold and obtains a representation vector of the travel trajectory of the resident population. Based on the representation vector, the travel trajectory of the resident population is clustered to obtain a student population among the resident population, so as to identify a graduate population from the student population.

[0046] Compared to existing graduate screening methods, this invention first obtains the representation vectors of the travel trajectories of resident populations within the target area. Then, based on these representation vectors, the travel trajectories of resident populations are clustered to filter out student groups, and finally, graduate groups are determined from these student groups. Therefore, this invention filters student groups and corresponding graduate groups based on travel trajectories, avoiding student identification errors caused by age-based screening and achieving accurate identification of student groups.

[0047] Based on this, the present invention can solve the problem of inaccurate identification of graduate groups, ensure accurate identification of graduates, and provide reliable data support for subsequent statistical analysis related to the employment of graduate groups. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the hardware operating environment involved in the embodiments of the present invention;

[0049] Figure 2 This is a schematic diagram of the first process of an embodiment of the population screening method of the present invention;

[0050] Figure 3 This is a schematic diagram of the second process of an embodiment of the population screening method of the present invention;

[0051] Figure 4 This is a schematic diagram of the functional modules of an embodiment of the population screening system of the present invention.

[0052] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0053] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0054] like Figure 1 As shown, Figure 1 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of the present invention.

[0055] It should be noted that the terminal device in the embodiments of the present invention can be a terminal device used to achieve accurate screening of graduate students. Specifically, the terminal device can be a smartphone, personal computer, server or other network device.

[0056] like Figure 1As shown, the device may include: a processor 1001, such as a CPU; a network interface 1004; a user interface 1003; a memory 1005; and a communication bus 1002. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0057] Those skilled in the art will understand that Figure 1 The device structure shown does not constitute a limitation on the device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0058] like Figure 1 As shown, the memory 1005, as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a group filtering program. The operating system is a program that manages and controls the device's hardware and software resources, supporting the operation of the group filtering program and other software or programs. Figure 1 In the device shown, the user interface 1003 is mainly used for data communication with the client; the network interface 1004 is mainly used for establishing a communication connection with the server; and the processor 1001 can be used to call the group screening program stored in the memory 1005 and perform the following operations:

[0059] Identify the resident population whose stay in the area to be screened exceeds a preset time threshold, and obtain the representation vector of the travel trajectory of the resident population;

[0060] Based on the representation vector, the travel trajectories of the resident population are clustered to obtain the student population within the resident population, so as to identify the graduate group from the student population.

[0061] Furthermore, the processor 1001 can also be used to call the crowd filtering program stored in the memory 1005, and also perform the following operations:

[0062] Obtain the user travel zipper table of each user in the permanent population of the area to be filtered. The user travel zipper table includes multiple user travel trajectories and corresponding user travel times.

[0063] According to the user's travel time, sort the multiple user travel trajectories to obtain the user travel trajectory sequence for each user;

[0064] The user's travel trajectory sequence is used as input to a preset neural network to obtain the representation vector of the user's travel trajectory sequence output by the hidden layer of the preset neural network.

[0065] Furthermore, the processor 1001 can also be used to call the crowd filtering program stored in the memory 1005, and also perform the following operations:

[0066] Before the step of using the user's travel trajectory sequence as input to a preset neural network and obtaining the representation vector of the user's travel trajectory sequence output by the hidden layer of the preset neural network, the method further includes:

[0067] Extract each trajectory point from the user's travel trajectory sequence and construct a trajectory point dictionary;

[0068] Any trajectory point in the trajectory point dictionary is taken as the first trajectory point, and other trajectory points other than the first trajectory point are taken as the second trajectory point.

[0069] Construct a trajectory point pair containing the first trajectory point and the second trajectory point;

[0070] The trajectory point pairs are used as training data and input into the preset neural network to obtain the trajectory prediction results output by the output layer of the preset neural network. Based on the trajectory prediction results, the weight matrix of the hidden layer is trained. The trajectory prediction results include the probability that the output trajectory point is the second trajectory point when the input trajectory point is the first trajectory point.

[0071] Furthermore, the processor 1001 can also be used to call the crowd filtering program stored in the memory 1005, and also perform the following operations:

[0072] Based on the representation vector, multiple initial centroids are determined, and according to the multiple initial centroids, the representation vector is divided into corresponding initial trajectory point clusters;

[0073] Obtain the distance between the representation vector of each trajectory point in the initial trajectory point cluster and each initial centroid, and update the trajectory points in each initial trajectory point cluster based on the multiple distances, so as to iterate the multiple initial trajectory point clusters to obtain multiple converged trajectory point clusters;

[0074] Obtain the number of representation vectors in each converged trajectory point cluster, and take the trajectory point cluster with the most vectors as the target trajectory point cluster;

[0075] The resident population corresponding to the target trajectory point cluster is considered as the student population.

[0076] Further, the region to be screened includes: university areas. After the step of clustering the travel trajectories of the resident population based on the representation vector to obtain the student population among the resident population, the processor 1001 can also be used to call the group screening program stored in the memory 1005 and perform the following operations:

[0077] Determine the time periods during which students stay on campus and the time periods during which they leave campus within the university area, wherein the time periods during which students stay on campus include the first and second time periods of different semesters of the academic year;

[0078] The total student population is obtained by statistically analyzing the student population within the filter area during the first time period.

[0079] Remove distracting individuals from the entire student population, wherein the distracting individuals are located in the screening area during the school departure period;

[0080] The graduate group is obtained from the entire student population after removing the interference group, and the students who are not in the screening area during the second time period.

[0081] Furthermore, prior to the step of obtaining the representation vectors of the resident population within the region to be screened, the processor 1001 may also invoke the population screening program stored in the memory 1005 and perform the following operations:

[0082] Determine a detailed list of base stations in university areas nationwide and a travel zipper list for university students nationwide;

[0083] The invalid base station data in the user travel zipper table is removed to obtain the processed user travel zipper table;

[0084] Based on the processed user travel zipper table and the base station identifier in the national university area base station details table, the processed user travel zipper table is associated with the national university area base station details table;

[0085] Based on the latitude and longitude of the regional boundaries of the university areas, filter out the university activity population from the associated user travel zipper table and the national university area base station details table;

[0086] Based on the length of time the university students spend in the preset screening area, the resident population in the screening area and the corresponding travel trajectories of the resident population are determined.

[0087] Furthermore, the processor 1001 can also be used to call the crowd filtering program stored in the memory 1005, and also perform the following operations:

[0088] The system obtains the switching frequency of the user's device switching back and forth between two communication base stations, and when the switching frequency is greater than a preset frequency threshold, it obtains the first base station data corresponding to the two communication base stations.

[0089] Obtain data from a second base station that is not located within the area of ​​the university;

[0090] The first base station data and the second base station data are removed from the user travel zipper table to obtain the processed user travel zipper table.

[0091] With the development of geographic information technology and 5G technology, mobile signaling data exhibits characteristics such as diversity, complexity, and massive volume, making the exploration and mining of signaling data an important topic. However, the method of identifying current college students / recent graduates solely using signaling data without introducing external data or sensitive user information suffers from inaccurate identification, leading to an inability to accurately analyze the employment status of graduates.

[0092] Currently, the identification of college students / recent graduates mainly relies on a rough screening based on the geographical location of the university and age in the user profile. However, age screening based on user profiles inevitably presents the following problems:

[0093] (1) The user profile table has some missing data, which cannot be linked through IMSI. There is also some missing age data, which leads to a large error in screening student users.

[0094] (2) There are some younger students in the gifted youth program and some older students among the college students. Screening by age alone may lead to errors.

[0095] (3) Some students use their family members’ supplementary cards. The identity information registered on the supplementary cards is not that of the students. If the screening is done by age alone, these students will be overlooked, causing errors.

[0096] In addition, due to the high degree of similarity in the travel trajectories of college students on campus, current research on population identification lacks the representation of user travel trajectory data.

[0097] The above-mentioned problems can all lead to inaccurate identification of graduates.

[0098] Therefore, in order to solve the above problems and improve the accuracy of identifying student groups and graduate groups within student groups, this invention proposes a method for identifying college graduates based on mobile signaling data. This invention is data-driven and helps to objectively understand the current situation of college students and recent graduates, providing data and technical support for subsequent analysis on graduate employment trends and employment areas.

[0099] Reference Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the population screening method of the present invention.

[0100] In this embodiment, an example of a group screening method is provided. It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0101] Step S10: Determine the resident population whose stay time in the area to be screened exceeds a preset time threshold, and obtain the representation vector of the travel trajectory of the resident population.

[0102] It should be noted that, in this embodiment, the area to be screened can specifically be a university area. Correspondingly, the permanent residents can include people who stay in the university area for a longer period of time than a preset time threshold. For example, people who stay in the university area for more than 80 hours per month (i.e., stay in the university area for 16 days per month and stay for 5 hours per day) can be considered as permanent residents.

[0103] It is worth noting that, considering that university students generally carry mobile devices such as smartphones, and as described above, with the development of geographic information technology and 5G technology, mobile device signaling data is characterized by diversity, complexity, and massive volume. Therefore, in this embodiment, the IMSI (International Mobile Subscriber Identity) in the mobile device information data can be used to identify the base station data of each student's device (such as a mobile phone) accessing the base station. Based on the base station data, and combined with mobile device data such as IMSI, important information such as each student's travel trajectory can be identified to determine the resident population within the university area.

[0104] Based on this, the terminal device can use the skip-gram trajectory prediction algorithm to perform representation learning on the aforementioned resident population and obtain the representation vector of the aforementioned resident population.

[0105] Therefore, in this embodiment, instead of using user profiles to screen students in university areas, the representation vector of the travel trajectory of the resident population can be directly obtained. The travel trajectory of the resident population can then be used to screen the graduate population, achieving a more rigorous and accurate group screening and avoiding incorrect screening of the graduate population due to missing or inaccurate user profile data.

[0106] Step S20: Based on the representation vector, cluster the travel trajectories of the resident population to obtain the student population in the resident population, so as to determine the graduate group from the student population.

[0107] In this embodiment, considering that the travel trajectories of university students have a high degree of overlap, after obtaining the representation vector of the travel trajectory of the aforementioned permanent residents, the terminal device can perform cluster analysis on the permanent residents in the university area based on the representation vector using a clustering algorithm (such as K-means) to select the group with the most significant overlap from the permanent residents as the student population.

[0108] Based on this, the terminal device can determine the graduate group from the student population according to the time period in which the student population is in the screening area (such as the university area).

[0109] For example, students who are not located in university areas between September and November can be identified as graduates.

[0110] In this embodiment, the terminal device can perform representation learning on the resident population within the university area to obtain representation vectors for these resident populations. After obtaining the representation vectors of the resident population's travel trajectories, the terminal device can use these representation vectors to select the group with the most significant overlap characteristics from the resident population as the student population. Furthermore, based on the time periods during which these student populations are located within the target area (university area), the graduating group is determined from the student population.

[0111] Therefore, compared to existing graduate screening methods, this invention first obtains the representation vector of the travel trajectories of the resident population within the screening area. Then, based on this representation vector, students can be screened from the resident population. Finally, based on the time periods during which the students are located within the screening area, graduates can be screened from the student population. Thus, this invention screens students based on travel trajectories and then, combined with corresponding time periods, screens graduates, avoiding the inaccurate graduate identification problems caused by age-based screening. This achieves accurate identification of graduates and provides reliable data support for subsequent statistical analysis related to graduate employment.

[0112] Based on the first embodiment of the population screening method of the present invention, a second embodiment of the population screening method of the present invention is proposed.

[0113] In this embodiment, step S10 above, "obtaining the representation vector of the travel trajectory of the resident population," may include:

[0114] Step S101: Obtain the user travel zipper table of each user in the resident population of the area to be filtered, and obtain the travel trajectory of multiple users in the user travel zipper table and the corresponding user travel time.

[0115] Step S102: Sort the multiple user travel trajectories according to the user's travel time to obtain the user travel trajectory sequence for each user;

[0116] Step S103: The user travel trajectory sequence is used as the input of a preset neural network to obtain the representation vector of the user travel trajectory sequence output by the hidden layer of the preset neural network.

[0117] It should be noted that, in this embodiment, signaling data from telecommunications operators can be used to collect information such as the location of the base station sector where each IMSI identifier is located, and the entry and exit time information. The data is then cleaned to form a user travel zipper table. This table contains information such as IMSI, MSISDN (Mobile Subscriber International ISDN / PSTN number, a number that uniquely identifies a mobile user in the Public Telephone Network Switching Network Numbering Plan), LACCELL (base station identifier), base station latitude and longitude, and base station entry and exit time, as well as other user travel trajectory information.

[0118] Terminal devices can use the skip-gram algorithm to obtain the representation vector of the travel trajectory of the resident population in the university area.

[0119] Specifically, for example, the terminal device can obtain each user's travel zipper table from the aforementioned user travel zipper table based on the device identifier of each user's device. Furthermore, it can obtain the user's travel trajectory and corresponding travel time from the user travel zipper table. For example, for a given user, their travel trajectory may include location 1, location 2, location 3, and location 4.

[0120] Furthermore, the base station to which the user's device is connected when the user arrives at the aforementioned locations can be obtained from the signaling data of the telecommunications operator, and the user's travel time at the location can be recorded when the device connects to the base station.

[0121] Then, according to the above-mentioned user travel time, the travel trajectories of the above-mentioned users are sorted in order to obtain the user travel trajectory sequence of each user.

[0122] Based on this, the terminal device can input the above-mentioned user travel trajectory sequence into a preset neural network and obtain the representation vector of the user travel trajectory sequence output by the hidden layer of the preset neural network.

[0123] It is worth noting that in this embodiment, the aforementioned preset neural network includes an input layer, a hidden layer, and an output layer. The representation vector output by the hidden layer is actually the representation vector corresponding to the user's travel trajectory sequence after sequence representation learning.

[0124] Therefore, in this embodiment, the user's travel trajectory sequence is obtained based on base station data, and a neural network is used to obtain the representation vector corresponding to the user's travel trajectory sequence. This representation vector is then used to screen the student group. Compared with the method of screening graduates based on age, this embodiment can eliminate the misidentification of the graduate group caused by age error, and achieve a more rigorous and accurate screening of the graduate group.

[0125] Furthermore, before step S103 above, "using the user travel trajectory sequence as input to a preset neural network and obtaining the representation vector of the user travel trajectory sequence output by the hidden layer of the preset neural network", the following may also be included:

[0126] Step S104: Extract each trajectory point from the user's travel trajectory sequence and construct a trajectory point dictionary;

[0127] Step S105: Take any trajectory point in the trajectory point dictionary as the first trajectory point, and take other trajectory points other than the first trajectory point as the second trajectory point.

[0128] Step S106: Construct a trajectory point pair containing the first trajectory point and the second trajectory point;

[0129] Step S107: The trajectory point pairs are used as training data and input into the preset neural network to obtain the trajectory prediction result output by the output layer of the preset neural network. Based on the trajectory prediction result, the weight matrix of the hidden layer is trained. The trajectory prediction result includes the probability that the output trajectory point is the second trajectory point when the input trajectory point is the first trajectory point.

[0130] It should be noted that, in this embodiment, as described above, the preset neural network includes an input layer, a hidden layer, and an output layer. Specifically, for the hidden layer in this embodiment, it is assumed that 300 features are used to represent a trajectory point (i.e., each word is a 256-dimensional vector), meaning the hidden layer has 300 neurons, and the weights of the hidden layer are a 10000*256 matrix, where 10000 represents the number of trajectory points in the trajectory point dictionary. The training objective in this embodiment is to learn the weight matrix of the hidden layer, and the output of the hidden layer is a 1*256 vector. As for the output layer in this embodiment, it is a softmax regression classifier, where each node can output a value (probability) between 0 and 1, and the sum of the probabilities of all nodes is 1.

[0131] In summary, this embodiment can input paired trajectory points into a preset neural network and obtain the probability distribution of the output layer to train the weight matrix of the hidden layer in the neural network.

[0132] Based on this, if there is a user travel trajectory sequence: [location 1, location 2, location 3, location 4, location 5, location 6], then the terminal device can obtain a trajectory point (i.e., the first trajectory point in this embodiment) from the above user travel trajectory sequence as input, such as location 2. Then, it can obtain other trajectory points other than the first trajectory point as the second trajectory point, such as location 3.

[0133] In one embodiment, the method for obtaining the first trajectory point and the second trajectory point may specifically include:

[0134] (1) Determine the skip_window parameter (this parameter is used to determine the sliding window for selecting trajectory points); if the first trajectory point is position 2 and skip_window = 2, then take the two nearest trajectory points in the left and right directions from position 2 respectively, that is (position 1, position 2, position 3, position 4), and the size of the sliding window at this time span = skip_window * 2 + 1 = 5.

[0135] (2) In addition, this embodiment also needs to determine num_skips (this parameter is used to determine the number of output trajectory points, i.e. the number of second trajectory points). This parameter is no greater than 2*skip_window, that is, at most all other trajectory points except the first trajectory point are used as output trajectory points (but they cannot be repeated); for example, if num_skips=2 is set, then two trajectory points can be selected from the upper and lower positions as output: (position 1, position 3), and the resulting trajectory point pairs include: (position 2, position 1) and (position 2, position 3).

[0136] Furthermore, the terminal device can use the aforementioned trajectory point pairs as training data, input them into a preset neural network, and obtain the trajectory prediction result output by the output layer of the preset neural network. The trajectory prediction result represents the probability that each trajectory point in the trajectory point dictionary will appear simultaneously with the input trajectory point. For example, when the input trajectory point is the aforementioned first trajectory point, the probability that the output trajectory point is the aforementioned second trajectory point.

[0137] In another embodiment, after obtaining the first trajectory point and the second trajectory point, one-hot encoding can be performed on each trajectory point. For example, the first trajectory point can be positionally encoded to obtain the encoded vector position 2 = [0,0,1,0,…0]. If the trajectory point dictionary size is 10000, the probability that the current trajectory point is the input is 1 / 10000. Furthermore, the encoded trajectory points can be input into the aforementioned preset neural network for training.

[0138] Therefore, this embodiment trains the travel trajectory sequence using the skip-gram algorithm. By using the current user's location node, it predicts the probability of the location node appearing in the past and future n time points, thereby training a representation vector. The representation vector contains high-order implicit information about the region corresponding to the node, which can be used as a region perception result. That is, since the user trajectory data is actually a sequence of location points, the feature vector obtained through the representation algorithm can also be considered as the feature vector of the region corresponding to the current trajectory point. Therefore, this embodiment can pre-train the neural network, enabling the pre-trained neural network to learn the representation of the input user travel trajectory sequence and obtain the corresponding representation vector. Based on this representation vector, student groups can be screened from the resident population. Compared to age-based graduate screening methods, this invention achieves a more rigorous and realistic graduate group screening, improving the accuracy of graduate group screening.

[0139] Furthermore, in step S20 above, "clustering the travel trajectories of the resident population based on the representation vector to obtain the student population within the resident population" may include:

[0140] Step S201: Divide the representation vector into multiple initial trajectory point clusters and obtain the initial centroid of each initial trajectory point cluster;

[0141] Step S202: Obtain the distance between the representation vector of each trajectory point in the initial trajectory point cluster and each initial centroid, and update the trajectory points in each initial trajectory point cluster based on the multiple distances, so as to iterate the multiple initial trajectory point clusters to obtain multiple converged trajectory point clusters;

[0142] Step S203: Obtain the number of representation vectors in each of the converged trajectory point clusters, and take the trajectory point cluster with the most representation vectors as the target trajectory point cluster.

[0143] Step S204: The resident population corresponding to the target trajectory point cluster is taken as the student population.

[0144] In this embodiment, considering the high degree of overlap in the travel trajectories of university students, the K-means clustering algorithm can be applied to segment the resident population within the university area into the student population. It is understood that in this embodiment, the resident population can be divided into two categories: students and non-students.

[0145] Based on this, after obtaining the representation vectors of the above-mentioned user travel trajectory sequence, the terminal device can randomly select two representation vectors as initial centroids through the K-means clustering algorithm. As explained above, the representation vector is the vector representation of the trajectory points.

[0146] Furthermore, the terminal device can determine the initial trajectory point clusters corresponding to each initial centroid based on the aforementioned initial centroids, wherein each initial trajectory point cluster contains multiple representation vectors. The distance between each representation vector in the initial trajectory point cluster and each of the aforementioned initial centroids is calculated, and the trajectory points in the initial trajectory point cluster are updated based on these distances.

[0147] For example, if this embodiment includes an initial trajectory point cluster A and an initial trajectory point cluster B, where the initial centroid of the initial trajectory point cluster A is a and the initial centroid of the initial trajectory point cluster B is b, then when the distance between the representation vector A1 in the initial trajectory point cluster A and the initial centroid a is greater than the distance between the representation vector A1 and the initial centroid b, the representation vector A1 can be assigned to the initial trajectory point cluster B, and the centroids of the initial trajectory point cluster A and the initial trajectory point cluster B can be recalculated to obtain new centroids. The new centroids are then used to update the initial centroids until the centroids of the initial trajectory point cluster A and the initial trajectory point cluster B no longer change, or the iteration ends. At this point, multiple convergent trajectory point clusters can be obtained, namely, convergent trajectory point cluster A and convergent trajectory point cluster B.

[0148] Furthermore, after obtaining multiple convergent trajectory point clusters, the trajectory point cluster with the largest number of representation vectors is taken as the target trajectory point cluster, and the resident population corresponding to the representation vectors in the target trajectory point cluster is taken as the student population.

[0149] Therefore, in this embodiment, the K-means clustering method is used to segment the user's trajectory representation vector to classify the population. Compared with the screening method based on the user's age and the device identifier of the user's device, the present invention can improve the accuracy of student identification and better reflect the high overlap of the actual student group's travel trajectories.

[0150] In this embodiment, user travel trajectory sequences can be obtained based on base station data, and a neural network can be used to obtain the corresponding representation vectors. These representation vectors can then be used to screen student groups. Compared to age-based graduate screening, this embodiment eliminates misidentification of graduates due to age errors, achieving a more rigorous and accurate graduate screening. Furthermore, the neural network can be pre-trained, enabling it to learn representations from the input user travel trajectory sequences and obtain corresponding representation vectors. These vectors can then be used to screen student groups from the resident population, improving the accuracy of graduate screening. Additionally, this embodiment can use clustering to identify student groups from the resident population, fully utilizing the high overlap in student travel trajectories to achieve accurate student identification.

[0151] Based on the first and second embodiments of the population screening method of the present invention, a third embodiment of the population screening method of the present invention is proposed.

[0152] In this embodiment, after step S20, "clustering the travel trajectories of the resident population based on the representation vector to obtain the student population among the resident population," the following may be included:

[0153] Step S30: Determine the time period for students to stay on campus and the time period for students to leave campus within the university area, wherein the time period for staying on campus includes the first time period and the second time period of different semesters of the academic year;

[0154] Step S40: Count the number of students in the area to be screened during the first time period to obtain the total student population;

[0155] Step S50: Remove the interfering group from the entire student population, wherein the interfering group is in the screening area during the time period of leaving school;

[0156] Step S60: From the entire student population after removing the interference group, select the student group that is not in the screening area during the second time period as the graduate group.

[0157] It should be noted that in this embodiment, the time period during which graduates stay in the university area has certain characteristics. For example, February to June and September to November are the first and second semesters of the academic year, which are the normal time periods for students to be on campus, while July to August is the time period for students to stay on campus, which is the summer vacation period.

[0158] In existing graduate screening methods, the first step is generally to screen IMSI sets that match the age range of graduates; the second step is to screen IMSI sets whose residence is within the designated school area in any month from February to June; and the third step is to screen IMSI sets whose residence is not within the school area in September to November from the results of the first two steps. The IMSI set obtained after these three steps is the set of graduates for the year under study.

[0159] However, the existing graduate screening methods do not take into account the impact of summer vacation. For example, there may be students staying on campus during the summer vacation in university areas. In order to avoid identifying these students as graduates, these students can be excluded.

[0160] Therefore, in order to avoid the negative impact of the summer vacation on the screening of graduates, this embodiment defines three time periods: the period of staying on campus and the period of leaving campus. The period of staying on campus includes the first and second periods of different semesters of the academic year. For example, the first period is from February to June, the second period is from September to November, and the period of leaving campus is from July to August.

[0161] Furthermore, the terminal device can collect statistics on the student population in the university area from February to June, and take the union of these statistics to obtain the total student population. Then, to eliminate the impact of the summer vacation, the students who were permanently residing in the university area from July to August (i.e., the interference group in this embodiment) can be removed from the total student population. Finally, students who were not in the university area from September to November can be identified as the graduating group from the remaining student population after removing the interference group.

[0162] Therefore, in this embodiment, the above-mentioned screening method for graduates eliminates the impact of holidays and improves the accuracy of graduate identification.

[0163] It is understandable that the method for determining whether a user is within a university area in this embodiment can be based on the user travel zipper table in the above embodiment. Since the user travel zipper table contains user travel trajectory information such as the base station identifier, base station latitude and longitude, device identifier of the user's device, and the time when the user's device enters and exits the base station, it is possible to determine whether the user is within a university area and the time they are within the university area through the above information.

[0164] Furthermore, before step S10, "obtaining the representation vector of the travel trajectory of the resident population in the area to be screened," the following may also be included:

[0165] Step S70: Determine the detailed table of base stations in university areas nationwide and the travel zipper table of university students nationwide;

[0166] Step S80: Remove invalid base station data from the user travel zipper table to obtain the processed user travel zipper table;

[0167] Step S90: Associate the processed user travel zipper table and the detailed base station table of the national university area based on the base station identifier in the processed user travel zipper table and the detailed base station table of the national university area.

[0168] Step S100: Filter out the university activity population from the associated user travel zipper table and the national university area base station details table;

[0169] Step S110: Based on the length of time the university students spend in the preset screening area, determine the resident population in the screening area and the travel trajectory of the resident population.

[0170] In this embodiment, the terminal device can collect base station information for university areas nationwide (i.e., the preset screening area in this embodiment) based on the latest data released by the Ministry of Education, forming a detailed base station table for university areas nationwide. This detailed base station table includes important AOI information for universities, such as the region ID, the province where the university is located, the university name, the names of each campus, and the LACCELL (base station identifier).

[0171] Simultaneously, terminal devices can utilize signaling data from telecom operators to collect information such as the location of the base station sector where the IMSI identifier of each user's device (e.g., a mobile phone) is located, as well as entry and exit times. This data is then cleaned to create a user travel zipper table. This table includes crucial information about the user's travel trajectory, such as IMSI, MSISDN, LACCELL, base station latitude and longitude, and base station entry and exit times.

[0172] Based on this, the terminal device can remove invalid base station data from the aforementioned user travel zipper list to obtain a processed user travel zipper list. This invalid base station data includes, but is not limited to, drift points and ping-pong points.

[0173] Drift point can be understood as the signal emitted by the base station drifting to places that the base station should not cover due to wind or reflection. Some boundary roaming is caused by signal drift. Ping-pong point can be understood as the phenomenon in a general mobile communication system where, if the signal strength of two base stations changes drastically in a certain area, the mobile phone will switch back and forth between the two base stations, producing the "ping-pong effect".

[0174] Then, the terminal device can use the processed user travel zipper table and the LACCELL in the national university area base station details table to associate the processed user travel zipper table and the national university area base station details table. Based on the AOI information of the university area (such as the latitude and longitude of the university area boundaries in this embodiment), it can filter out university-related users from the associated user travel zipper table and the national university area base station details table.

[0175] Finally, the above-mentioned university activity groups can be screened to obtain the length of time each user stays in the university area. Those whose stay exceeds a preset threshold can be identified as permanent residents of the university area. For example, those who stay in the university area for more than 80 hours per month (i.e., stay in the university area for 16 days per month and 5 hours per day) can be identified as permanent residents, and their travel trajectories can be determined.

[0176] Further, in step S80 above, "removing invalid base station data from the user travel zipper table to obtain a processed user travel zipper table" may include:

[0177] Step S801: Obtain the switching frequency of the user's device switching back and forth between two communication base stations, and when the switching frequency is greater than a preset frequency threshold, obtain the first base station data of the two communication base stations;

[0178] Step S802: Obtain data of a second base station that is not located in the area to be filtered;

[0179] Step S803: Remove the first base station data and the second base station data from the base station zipper data to obtain the processed user travel zipper table.

[0180] It should be noted that, in this embodiment, as described above, there may be ping-pong points and drift points in the user travel zipper table. Therefore, in order to form a smooth and continuous user travel trajectory, which is beneficial to the subsequent representation learning of the user trajectory representation vector, it is necessary to remove ping-pong points and drift points in advance.

[0181] Specifically, for example, the terminal device first needs to determine the switching frequency of the user's device switching between any two communication base stations. When this switching frequency is greater than a preset frequency threshold, it means that the signal strength of the two base stations changes drastically, causing the device to frequently switch between the two base stations. At this time, the first base station data of the two base stations can be obtained, such as the base station's latitude and longitude coordinates and base station identification. In addition, there may be base station signals from other base stations outside the area to be screened drifting into the university area, causing the user's device in the university area to connect to that base station, resulting in incorrect identification of the user's travel trajectory. The terminal device also needs to obtain the base station data of that base station (i.e., the second base station data in this embodiment), which also includes the base station's latitude and longitude coordinates and base station identification.

[0182] Then, the terminal device can remove the aforementioned first base station data and second base station data from the user travel zipper table to obtain the processed user travel zipper table.

[0183] In another embodiment, after removing the aforementioned ping-pong points and drift points, the terminal device can also calculate the dwell points and movement points in the user's travel zipper table according to the spatial dwelling model. The dwell point can be understood as the point where the user stays in a certain place for more than a certain time threshold, that is, the duration of the user's device accessing a certain base station exceeds a certain time threshold (e.g., 30 minutes). Conversely, the movement point can be understood as the duration of the user's device accessing a certain base station does not exceed a certain time threshold.

[0184] Based on this, multiple base stations can be aggregated. Among them, multiple base stations that can be aggregated must meet at least the following conditions: the base stations of each base station are very close to each other (for example, the distance between two base stations does not exceed a certain distance threshold), and multiple base stations can be aggregated into one point.

[0185] Therefore, in this embodiment, by removing base station data and aggregating dwell points in the user travel zipper table as described above, the user travel zipper table is preprocessed, resulting in a smooth and continuous user travel trajectory, which is beneficial for subsequent student screening.

[0186] In this embodiment, considering the characteristics of the time spent by graduating students in university areas—for example, February to June and September to November are the periods when students stay on campus, while July to August is the period when they leave—the graduating group can be screened from the student population in university areas based on the time periods during which the resident population is in the university area. Furthermore, this embodiment can generate a detailed table of base stations in university areas nationwide and a user travel zipper table, and remove ping-pong dots and drift points from the user travel zipper table to obtain a smooth and continuous user travel trajectory.

[0187] Therefore, in this invention, as Figure 3 As shown, after obtaining a detailed list of base stations in university areas nationwide and a user travel zipper table, and processing the user travel zipper table, the system can link these tables to filter out the resident population in university areas. Furthermore, it can perform representation learning on the travel trajectories of these resident populations to obtain corresponding representation vectors. Based on these representation vectors, student groups can be filtered out from the resident population. Finally, based on the time periods these student groups spend in university areas, the system can identify the graduating class. Therefore, this invention represents user travel trajectory data and uses these representation vectors to identify student groups. Compared to existing methods that directly link user profile tables to filter students by age, this invention uses a clustering method to identify the representation vectors of user travel trajectories, thereby filtering out university students and graduating students. This has the following advantages:

[0188] 1) The user profile table has some missing data, making it impossible to correlate with IMSI, and some age data is also missing, leading to significant errors in screening student users. In contrast, this invention uses K-means clustering to segment the user population based on the representation vector of user trajectories, which is more realistic.

[0189] 2) Among university students, there are some younger students in special programs for gifted children and some older students. Additionally, some students use supplementary cards registered under family members' names, but the identity information on these supplementary cards is not that of the students; screening is based solely on age, which introduces errors. This invention's K-means clustering, which segments the population based on the representation vector of user trajectories, can effectively identify this group of students.

[0190] 3) The majority of the identification process can be completed using only signaling data, saving manpower and resources for collecting user profiles and improving identification efficiency.

[0191] Furthermore, embodiments of the present invention also propose a population screening system, referring to... Figure 4 , Figure 4 This is a schematic diagram of the functional modules of an embodiment of the group screening method of the present invention. Figure 4 As shown, the population screening system of the present invention includes:

[0192] The acquisition module 10 is used to identify resident populations whose stay time in the area to be screened exceeds a preset time threshold, and to acquire the representation vector of the travel trajectory of the resident populations.

[0193] The filtering module 20 is used to cluster the travel trajectories of the resident population based on the representation vector to obtain the student population in the resident population, so as to determine the graduate group from the student population.

[0194] The specific implementation methods of each functional module of the population screening system of the present invention are basically the same as those of the above-described population screening methods, and will not be repeated here.

[0195] Furthermore, embodiments of the present invention also propose a computer-readable storage medium storing a crowd screening program, which, when executed by a processor, implements the steps of the crowd screening method as described above.

[0196] The various embodiments of the population screening system and computer-readable storage medium of the present invention can be referred to the various embodiments of the population screening method of the present invention, and will not be repeated here.

[0197] Furthermore, embodiments of the present invention also provide a computer program product, the computer program product comprising a computer program, which, when executed by a processor, implements the steps of the crowd screening method as described in any of the embodiments of the above-described crowd screening method.

[0198] The specific embodiments of the computer program product of the present invention are basically the same as the embodiments of the above-described group screening method, and will not be described in detail here.

[0199] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, product, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, product, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, product, or apparatus that includes that element.

[0200] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0201] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, 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 (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or other network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0202] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method of population screening, characterized by, The population screening method comprises: determining a constant population staying in a region to be screened for a duration longer than a preset duration threshold, and obtaining a feature vector of a travel trajectory of the constant population; based on the feature vector, clustering the travel trajectory of the constant population to obtain a student population in the constant population, to determine a graduate population from the student population; the step of clustering the travel trajectory of the constant population based on the feature vector to obtain the student population in the constant population comprises: based on the feature vector, determining a plurality of initial centroids, and dividing the feature vector into corresponding initial trajectory point clusters according to the plurality of initial centroids; obtaining the distance between the feature vector of each trajectory point in the initial trajectory point cluster and each initial centroid, and updating the trajectory points in each initial trajectory point cluster based on a plurality of distances to obtain a plurality of converged trajectory point clusters by iterating the plurality of initial trajectory point clusters; obtaining the number of feature vectors in each converged trajectory point cluster, and taking the trajectory point cluster with the largest number as a target trajectory point cluster; taking the constant population corresponding to the target trajectory point cluster as the student population.

2. The population screening method of claim 1, wherein, The step of obtaining the feature vector of the travel trajectory of the constant population comprises: obtaining a user travel link list of each user in the constant population in the region to be screened, the user travel link list including a plurality of user travel trajectories and corresponding user travel times; sorting a plurality of user travel trajectories according to the user travel times to obtain a user travel trajectory sequence of each user; taking the user travel trajectory sequence as the input of a preset neural network to obtain the feature vector of the user travel trajectory sequence output by the hidden layer of the preset neural network.

3. The population screening method of claim 2, wherein, Before the step of taking the user travel trajectory sequence as the input of a preset neural network to obtain the feature vector of the user travel trajectory sequence output by the hidden layer of the preset neural network, it further comprises: extracting each trajectory point in the user travel trajectory sequence to construct a trajectory point dictionary; taking any trajectory point in the trajectory point dictionary as a first trajectory point and taking other trajectory points except the first trajectory point as second trajectory points; constructing a trajectory point pair comprising the first trajectory point and the second trajectory point; taking the trajectory point pair as training data, inputting the preset neural network to obtain a trajectory prediction result output by the output layer of the preset neural network, and training the weight matrix of the hidden layer according to the trajectory prediction result, wherein the trajectory prediction result includes the probability that the output trajectory point is the second trajectory point when the input trajectory point is the first trajectory point.

4. The population screening method of claim 1, wherein, The region to be screened comprises a college area, and after the step of clustering the travel trajectory of the constant population based on the feature vector to obtain the student population in the constant population, it further comprises: determining the stay time period and the leave time period of the student population in the college area, wherein the stay time period includes a first time period and a second time period of different semesters in an academic year; counting a student population in the to-be-screened region in the first time period to obtain a whole student population; removing an interference population from the whole student population, wherein the interference population is in the to-be-screened region in the off-campus time period; obtaining a student population not in the to-be-screened region in the second time period from the whole student population after removing the interference population as a graduate group.

5. The population screening method of claim 4, wherein, Before the step of obtaining the representation vector of the resident population in the to-be-screened region, the method further comprises: determining a nationwide university regional base station detailed table and a nationwide university group user travel link list; removing invalid base station data in the user travel link list to obtain a processed user travel link list; associating the processed user travel link list with the nationwide university regional base station detailed table according to base station identifiers in the processed user travel link list and the nationwide university regional base station detailed table; screening a university activity population from the associated user travel link list and the nationwide university regional base station detailed table according to the regional boundary longitude and latitude of the university region; determining a resident population in the to-be-screened region and a travel trajectory corresponding to the resident population according to a stay duration of the university activity population in a preset to-be-screened region.

6. The population screening method of claim 5, wherein, The step of removing invalid base station data in the user travel link list to obtain a processed user travel link list comprises: obtaining a switching frequency of a user-held device switching back and forth between two communication base stations, and obtaining first base station data corresponding to the two communication base stations when the switching frequency is greater than a preset frequency threshold; obtaining second base station data of a communication base station not in the university region; removing the first base station data and the second base station data from the user travel link list to obtain a processed user travel link list.

7. A population screening system, characterized by, The population screening system comprises: an obtaining module configured to determine a resident population staying in a to-be-screened region for more than a preset duration threshold, and obtain a representation vector of a travel trajectory of the resident population; a screening module configured to cluster the travel trajectory of the resident population based on the representation vector, obtain a student population in the resident population, and determine a graduate group from the student population; the screening module is further configured to determine a plurality of initial centroids based on the representation vector, and divide the representation vector into corresponding initial trajectory point clusters according to the plurality of initial centroids; obtain distances between representation vectors of trajectory points in the initial trajectory point clusters and the initial centroids, and update the trajectory points in each of the initial trajectory point clusters based on a plurality of the distances to iterate a plurality of the initial trajectory point clusters to obtain a plurality of converged trajectory point clusters; obtain a number of representation vectors in each of the converged trajectory point clusters, and take a trajectory point cluster with the largest number as a target trajectory point cluster; take a resident population corresponding to the target trajectory point cluster as a student population.

8. A terminal device, comprising: The terminal device comprises a memory, a processor and a population screening program stored on the memory and executable on the processor, the population screening program, when executed by the processor, implements the steps of the population screening method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a population screening program, the population screening program, when executed by the processor, implements the steps of the population screening method according to any one of claims 1 to 6.

Images