A multimedia information recommendation method and device, and a storage medium

Abstract

The application discloses a multimedia information recommendation method and device and a storage medium. The method comprises the following steps: acquiring position signaling data of each user in a recommendation area; the user is a user on a terminal covered in the recommendation area; using a preset permanent area model, performing permanent area prediction on each user according to the position signaling data to obtain a permanent area prediction result corresponding to each user; based on the permanent area prediction result, determining a permanent user from each user, and pushing to-be-recommended multimedia information to a terminal corresponding to the permanent user. Through the technical solution, the accuracy and diversity of information recommendation are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a multimedia information recommendation method and device and storage medium. BACKGROUND

[0002] In recent years, with the development of communication technology, the public management department carries out the publicity work of public information, which is not limited to the way of leaflets, announcements and the like.

[0003] At present, the publicity way of public information generally adopts the text language transmission mode such as short message and WeChat group message. The regional information publicity and reach mode based on WeChat can only be realized through artificial WeChat forwarding, and cannot accurately lock the information reach crowd, so that the accuracy of information publicity is poor. Or, the regional information publicity and reach mode based on short message or WeChat takes text as the carrier, so that it is difficult for people of different age levels and different cultural levels to receive and understand information, and the information publicity mode is single. SUMMARY

[0004] To solve the above technical problems, the embodiments of the present application expect to provide a multimedia information recommendation method, device and storage medium, which improve the accuracy and diversity of information recommendation.

[0005] The technical scheme of the present application is realized as follows:

[0006] The present application provides a multimedia information recommendation method, which comprises:

[0007] An acquisition module is configured to acquire position signaling data of each user in a recommended area; the user is a user on a terminal covered in the recommended area;

[0008] A prediction module is configured to use a preset permanent area model to respectively predict permanent areas of the users according to the position signaling data, and obtain permanent area prediction results corresponding to the users.

[0009] Based on the permanent area prediction results, permanent users are determined from the users, and multimedia information to be recommended is pushed to terminals corresponding to the permanent users.

[0010] The present application provides a multimedia information recommendation device, which comprises:

[0011] An acquisition module is configured to acquire position signaling data of each user in a recommended area; the user is a user on a terminal covered in the recommended area;

[0012] A prediction module is configured to use a preset permanent area model to respectively predict permanent areas of the users according to the position signaling data, and obtain permanent area prediction results corresponding to the users.

[0013] A pushing module is configured to determine a resident user from the users based on the resident area prediction result, and push the multimedia information to be recommended to a terminal corresponding to the resident user.

[0014] The application provides a multimedia information recommendation device, comprising a processor, a memory and a communication bus.

[0015] The communication bus is configured to realize the communication connection between the processor and the memory.

[0016] The processor is configured to execute the computer program stored in the memory to realize the multimedia information recommendation method.

[0017] The application provides a computer readable storage medium, which stores one or more computer programs, and the one or more computer programs can be executed by one or more processors to realize the multimedia information recommendation method.

[0018] The application provides a multimedia information recommendation method, device and storage medium, and the method comprises the following steps: acquiring position signaling data of each user in a recommendation area; the user is a user on a terminal covered in the recommendation area; using a preset resident area model to predict a resident area of each user according to the position signaling data, and obtaining a resident area prediction result corresponding to each user; determining a resident user from the users based on the resident area prediction result, and pushing multimedia information to be recommended to a terminal corresponding to the resident user. According to the technical scheme provided by the application, the resident user in the recommendation area is predicted based on the position signaling data of each user, the resident user is effectively inferred, the resident user in the recommendation area can be accurately locked for multimedia information recommendation, the accuracy of information recommendation is improved, and the information pushing is carried out by taking multimedia as a carrier, so that the expression effect tends to be diversified, different age groups and different cultural groups can receive and understand the information, and the diversity of information recommendation is improved. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 A flowchart of a multimedia information recommendation method provided by an embodiment of the application is shown.

[0020] Figure 2 An exemplary flowchart of constructing a preset resident area model provided by an embodiment of the application is shown.

[0021] Figure 3 An exemplary flowchart of model iteration training provided by an embodiment of the application is shown.

[0022] Figure 4A schematic diagram illustrating an exemplary process for predicting resident areas, provided as an embodiment of the present invention;

[0023] Figure 5 A schematic diagram of the structure of a multimedia information recommendation device provided in an embodiment of the present invention. Figure 1 ;

[0024] Figure 6 A schematic diagram of the structure of a multimedia information recommendation device provided in an embodiment of the present invention. Figure 2 . Detailed Implementation

[0025] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings of the embodiments of the present invention. It is to be understood that the specific embodiments described herein are merely for explaining the relevant application and are not intended to limit the application. Furthermore, it should be noted that, for ease of description, only the parts relevant to the relevant application are shown in the accompanying drawings.

[0026] This invention provides a multimedia information recommendation method, implemented by a multimedia information recommendation device. Figure 1 This is a flowchart illustrating a multimedia information recommendation method provided in an embodiment of the present invention. Figure 1 As shown, the main steps include:

[0027] S101. Obtain the location signaling data of each user within the recommended area; the users are those on terminals covered within the recommended area.

[0028] It should be noted that, in the embodiments of the present invention, the recommended area can be an administrative region of a province or city, an area covered by a tourist attraction, an area where a commercial district is located, a densely populated urban area, or any other area. The specific recommended area can be defined according to actual needs and application scenarios. For example, the defined recommended area can be a circular area, a polygonal area, or an area of ​​other shapes. If the defined recommended area is circular, the multimedia information recommendation device can obtain the latitude and longitude [lngc, latc] of the center of the circular area and its radius Rc; if the defined recommended area is polygonal, the multimedia information recommendation device can obtain the vertex set information {[lng1, lat1], [lng2, lat2], [lng3, lat3], [lng4, lat4]...} of the polygonal area.

[0029] Specifically, in the embodiment of the present application, the video acquisition recommendation area of each user position signaling data, comprising: from the network server to obtain the position signaling data set; from the position signaling data set, the slice position signaling data of the target time period is cut out; according to the base station position included in the preset base station list, the target base station in the recommended area is determined; from the preset base station and user corresponding relationship, the user corresponding to the target base station is found, so as to obtain each user; from the slice position signaling data, the position signaling data of each user is screened out.

[0030] It should be noted that, in the embodiment of the present application, the position signaling data set obtained by the multimedia information recommendation device is the set of position signaling data on all base stations obtained from the network server, i.e. the position signaling data set includes the position signaling data of all users at different times of all base stations. Therefore, the multimedia information recommendation device needs to cut out the slice position signaling data of the target time period from the position signaling data set. Wherein, the target time period can be set according to actual demand and application scene.

[0031] Exemplarily, if the multimedia information recommendation device needs to predict whether each user in the recommended area is a resident user in the recommended area at 10 o'clock in the morning of a certain day, the target time period can be set to 9 o'clock to 10 o'clock in the morning of the day, or 8:30 to 9:30 in the morning of the day. In this way, the multimedia information recommendation device uses the data in the nearest time period to determine the time to predict the resident of each user in the recommended area, which can make the accuracy of the prediction result higher.

[0032] It should be noted that, in the embodiment of the present application, the slice position signaling data cut out by the multimedia information recommendation device includes the position signaling data obtained by all base stations in the target time period, therefore, the multimedia information recommendation device also needs to screen out the position signaling data of each user in the recommended area from the slice position signaling data. Exemplarily, the implementation process can be: the multimedia information recommendation device determines the base station in the recommended area according to the base station position included in the preset base station list, and then finds the user corresponding to the determined base station from the preset base station and user corresponding relationship, so as to obtain each user, and screen out the position signaling data corresponding to each user from the slice position signaling data.

[0033] It should be noted that in the embodiments of the present application, the preset base station list is the base station identity document (id) and geohash position information of all base stations pre-stored by the network server. If the recommended area is a circular area, when the distance H between the latitude and longitude position of the base station and the center of the circle is less than or equal to the radius Rc, it means that the corresponding base station is located within the circular area, and if the distance H between the latitude and longitude position of the base station and the center of the circle is greater than the radius Rc, it means that the corresponding base station is located outside the circular area. If the recommended area is a polygonal area, the ray method is used to identify the base stations within the polygonal area. In this way, the multimedia information recommendation device can determine all the base stations within the recommended area according to the base station positions included in the preset base station list.

[0034] It should be noted that in the embodiments of the present application, the preset base station and user correspondence includes the correspondence between the base station and the user. One base station can be used for communication by multiple users, so one base station can correspond to multiple users in the preset base station and user correspondence. After determining the base stations within the recommended area, the multimedia information recommendation device will find the users corresponding to the determined base stations according to the preset base station and user correspondence to obtain each user within the recommended area, and then filter out the position signaling data of each user from the slice position signaling data.

[0035] S102, using a preset permanent area model, respectively predicting the permanent area of each user according to the position signaling data to obtain the permanent area prediction result corresponding to each user.

[0036] It should be noted that in the embodiments of the present application, the multimedia information recommendation device uses each user corresponding to one position signaling data to obtain one permanent area prediction result; wherein the permanent prediction result is permanent or non-permanent.

[0037] Specifically, in the embodiments of the present application, the position signaling data includes at least one network communication quality data, and the multimedia information recommendation device uses the preset permanent area model to respectively predict the permanent area of each user according to the position signaling data to obtain the permanent area prediction result corresponding to each user, including: using the preset permanent area model to respectively predict the permanent area of each user according to at least one network communication quality data to obtain at least one prediction result corresponding to each user; determining the permanent area prediction result corresponding to each user according to at least one prediction result corresponding to each user.

[0038] It should be noted that in the embodiments of the present application, the location signaling data of each user includes at least one network communication quality data, and each network communication quality data includes network signal strength information of a serving cell (main cell) and a neighboring cell in which the user resides. The network signal strength information includes Reference Signal Receiving Power (RSRP), Reference Signal Receiving Quality (RSRQ), and Timing Advance (TADV).

[0039] For example, each network communication quality data obtained by the multimedia information recommendation device can be RSRP, RSRQ, and TADV of the main cell, RSRP and RSRQ of the neighboring cell 1, RSRP and RSRQ of the neighboring cell 2, RSRP and RSRQ of the neighboring cell 3, RSRP and RSRQ of the neighboring cell 4, and RSRP and RSRQ of the neighboring cell 5. The network signal strength information of the neighboring cell is obtained according to actual conditions, and the RSRP and RSRQ of 6 neighboring cells can be obtained, or the RSRP and RSRQ of 3 neighboring cells can be obtained.

[0040] It should be noted that in the embodiments of the present application, after the multimedia information recommendation device obtains at least one network communication quality data corresponding to each user, the multimedia information recommendation device inputs the at least one network communication quality data corresponding to each user into a preset camping area model to obtain at least one prediction result corresponding to each user. That is, each user corresponds to at least one network communication quality data, and each network communication quality data corresponds to one prediction result. In this way, at least one prediction result can be obtained for each user.

[0041] It should be noted that in the embodiments of the present application, an example implementation process of determining the camping area prediction result corresponding to each user according to the at least one prediction result corresponding to each user by the multimedia information recommendation device can be as follows: if one of the at least one prediction result corresponding to a user is camping, it is considered that the camping area prediction result of the user is camping; and if each of the at least one prediction result corresponding to a user is non-camping, it is considered that the camping area prediction result of the user is non-camping.

[0042] Specifically, in the embodiments of the present application, before the multimedia information recommendation device predicts the camping area of each user according to the location signaling data by using the preset camping area model to obtain the camping area prediction result corresponding to each user, the following steps can be further performed: obtaining training sample data; and constructing the preset camping area model by using a logistic regression algorithm according to the training sample data.

[0043] It should be noted that in the embodiments of the present application, the multimedia information recommendation device constructs the preset resident area model by using the preset resident area model. The preset area resident model is a model based on a logistic regression algorithm.

[0044] For example, the multimedia information recommendation device constructs the preset resident area model by the following process: obtaining training sample data; constructing the preset resident area model by using the logistic regression algorithm according to the training sample data.

[0045] It should be noted that in the embodiments of the present application, the multimedia information recommendation device needs to obtain the training sample data before constructing the preset resident area model.

[0046] Specifically, in the embodiments of the present application, the multimedia information recommendation device obtains the training sample data by: obtaining historical location signaling data corresponding to the region and historical regional user data corresponding to the region from the network service side for any region; using the historical regional user data to perform resident labeling on the historical location signaling data to obtain label data; and determining the historical location signaling data and the label data as the training sample data.

[0047] It should be noted that in the embodiments of the present application, the multimedia information recommendation device can construct a corresponding resident area model for any region to perform resident user prediction for the region. It can be understood that the present application constructs a regional crowd resident model (preset resident area model) based on a logistic regression algorithm, and uses the model to predict the real-time location signaling of the operator to obtain a regional resident target crowd data set, thereby providing a real-time regional resident crowd prediction method.

[0048] For example, the process of obtaining the training sample data is as follows: for any region, the multimedia information recommendation device obtains historical location signaling data and historical regional user data in the region from the network service side and the multimedia service side respectively, and then uses the historical regional user data to perform resident labeling on the historical location signaling data to obtain label data, and then determines the historical location signaling data and the label data as the training sample data. In this way, based on the operator location signaling data (historical location signaling data) and the regional user data (historical regional user data) of the short video service provider, the data of two independent industries is successfully integrated, the data fusion of the operator data and the Internet data is realized, and the data value is improved.

[0049] Specifically, in the embodiments of the present application, the historical position signaling data includes first communication identification information corresponding to each of the first users, and at least one historical network index data, each of which includes a first historical position code and first historical network communication quality data; the historical area user data includes second historical communication identification information corresponding to each of the second users, and at least one second historical position code; the multimedia information recommendation device uses the historical area user data to perform persistent labeling on the historical position signaling data to obtain label data, including: for each of the first users, using the corresponding first communication identification information to find target users with the same communication identification information from the second users; for each of the first users, comparing the first historical position code included in each of the historical network index data with at least one second historical position code of the corresponding target user to obtain a historical weight value of each of the historical network index data; selecting a target network index data with the largest historical weight value from the at least one historical network index data according to the respective historical weight values of the at least one historical network index data, and determining label data of the corresponding target network index data based on the largest historical weight value and a preset weight threshold.

[0050] It should be noted that, in the embodiments of the present application, the first communication identification information and the second communication identification information can be International Mobile Subscriber Identity (IMSI).

[0051] For example, one data record in the position signaling data obtained by the multimedia information recommendation device is shown in Table 1:

[0052] Table 1

[0053]

[0054] As shown in Table 1, for a certain area, the multimedia information recommendation device can obtain the area (Identity Document, ID) and the first communication identification information (IMSI) of each first user in the area from the network server, as well as at least one historical network index data, each of which includes a first historical position code (geohash) and first historical network communication quality data (main area RSRP, main area RSRQ, TADV, adjacent area RSRP1, adjacent area RSRQ1, adjacent area RSRP2, adjacent area RSRQ2, adjacent area RSRP3, adjacent area RSRQ3, adjacent area RSRP4, adjacent area RSRQ4, adjacent area RSRP5, adjacent area RSRQ5).

[0055] The multimedia information recommendation device obtains a data record in the regional user data from the multimedia service end, as shown in Table 2:

[0056] Table 2

[0057] Key Valve Explanation {Area ID}_{yyyyMMdd} {IMSI set} Terminal identification information set IMSI value {geohash} Position code corresponding to each terminal identification information

[0058] As shown in Table 2, for a certain region, the multimedia information recommendation device can obtain the regional ID, the second communication identifier information (IMSI) of all the second users in the region, and at least one second historical location encoding (geohash) from the network service end.

[0059] It should be noted that in the embodiments of the present application, after obtaining the historical regional user data and the historical location signaling data corresponding to a certain region, the multimedia information recommendation device will use the regional user data to permanently annotate the location signaling data. In the process of permanently annotating the location signaling data, the historical weight value of the historical location signaling data will be determined first. An exemplary implementation is that the multimedia information recommendation device will establish a corresponding relationship between the users with the same communication identifier information in the plurality of first users and the plurality of second users, obtain the target user corresponding to each first user, and then compare the first historical location encoding included in each historical network index data corresponding to each first user with at least one second historical location encoding of the corresponding target user to obtain the historical weight value of each historical network index data.

[0060] Specifically, in the embodiments of the present application, the multimedia information recommendation device compares the first historical location encoding included in each historical network index data corresponding to each first user with at least one second historical location encoding of the corresponding target user to obtain the historical weight value of each historical network index data, including: comparing the first historical location encoding included in each historical network index data corresponding to each first user with at least one second historical location encoding of the corresponding target user respectively to obtain at least one similar bit number comparison result corresponding to each historical network index data; determining a sub-weight value according to each similar bit number comparison result to obtain at least one sub-weight value corresponding to each historical network index data; and determining the sum of the at least one sub-weight value corresponding to each historical network index data as the corresponding historical weight value.

[0061] It should be noted that in the embodiments of the present application, the bit number of each location encoding can be set according to actual conditions and application scenarios. Table 3 lists the error range of different encoding bit numbers.

[0062] Table 3

[0063] geohash length Error (km) 4 ±20 5 ±2.4 6 ±0.61 7 ±0.076 8 ±0.019

[0064] In the comparison process of two position encodings, assuming that two position encodings are 8-bit, if the number of similar bits of the two position encodings is 8, the sub-weight value is determined as 2; if the number of similar bits of the two position encodings is 7, the sub-weight value is determined as 1; if the number of similar bits of the two position encodings is 6, the sub-weight value is determined as 0.5; if the number of similar bits of the two position encodings is less than 6, the sub-weight value is determined as 0. Wherein, the sub-weight value set for different similar bit comparison results, i.e. the number of different similar bits of the two position encodings, can be set according to actual conditions and application requirements, and the present application is not limited thereto.

[0065] An exemplary comparison process is as follows: assuming that the first user M corresponds to two historical network index data, the first historical position encodings included in the two historical network index data are geohashM1 and geohashM2 respectively; the target user N with the same identification information as the first user M corresponds to four second historical position encodings, which are geohashN1, geohashN2, geohashN3 and geohashN4. For the first user M, the multimedia information recommendation device will compare geohashM1 with geohashN1, geohashN2, geohashN3 and geohashN4 respectively in terms of the number of similar bits, and then determine a sub-weight value according to each similar bit comparison result. For example, geohashM1 and geohashN1 are compared in terms of the number of similar bits, and assuming that the similar bit comparison result is 8 bits, the sub-weight value is determined as 2; geohashM1 and geohashN2 are compared in terms of the number of similar bits, and assuming that the similar bit comparison result is 7 bits, the sub-weight value is determined as 1; geohashM1 and geohashN3 are compared in terms of the number of similar bits, and assuming that the similar bit comparison result is 6 bits, the sub-weight value is determined as 0.5; geohashM1 and geohashN4 are compared in terms of the number of similar bits, and assuming that the similar bit comparison result is less than 6 bits, the sub-weight value is determined as 0. In this way, the four sub-weight values corresponding to the historical network index data where geohashM1 is located are 2, 1, 0.5 and 0 respectively.

[0066] It should be noted that in the embodiments of the present application, for the historical network index data in which the geohashM1 is located, the multimedia information recommendation device determines the sum of the corresponding at least one sub-weight value as the corresponding historical weight value, that is, 2+1+0.5+0. In the same way, the historical weight value corresponding to the historical network index data in which the geohashM2 is located can be determined, which is assumed to be 4. In this way, the historical weight values corresponding to the two historical network index data in the first user M, that is, 3.5 and 4, can be obtained. Then, for the first user M, according to the historical weight values corresponding to the two historical network index data, the historical network index data with the largest historical weight value is selected as the target network index data from the two historical network index data with the largest value, that is, the historical network index data in which the geohashM2 is located is the target network index data. Further, the label data corresponding to the target network index data is determined according to the maximum historical weight value 4 and the preset weight threshold. The label data can be a resident label value, including: a first value and a second value, the first value is 1, representing resident; and the second value is 0, representing non-resident.

[0067] The exemplary way of determining the label data can be: if the weight value corresponding to the target network index data, that is, the maximum weight value, is greater than or equal to the preset threshold, the label data is determined to be the first value, that is, resident; if the weight value corresponding to the target network index data, that is, the maximum weight value, is less than the preset threshold, the label data is determined to be the second value, that is, non-resident. The preset threshold can be set according to actual needs and application scenarios, and the present application does not limit the value of the specific preset threshold.

[0068] It should be noted that in the embodiments of the present application, the historical network communication quality data and the label data included in the target network index data are actually determined as training sample data. Because, in the process of model training, the input is the historical network communication quality data included in the target network index data, and then the label data of the target network index data is expected to be obtained. An exemplary sample data record in the training sample data is shown in Table 4:

[0069] Table 4

[0070]

[0071]

[0072] As shown in Table 3, for a certain area, the training sample data determined by the multimedia information recommendation device includes: the first communication identification information (IMSI) of each first user in the plurality of first users within the area ID, and the historical network communication quality data (main area RSRP, main area RSRQ, TADV, adjacent area RSRP1, adjacent area RSRQ1, adjacent area RSRP2, adjacent area RSRQ2, adjacent area RSRP3, adjacent area RSRQ3, adjacent area RSRP4, adjacent area RSRQ4, adjacent area RSRP5, adjacent area RSRQ5) in the target network index data corresponding to each first user, and the label data (resident label value F) corresponding to the target network index data.

[0073] Figure 2 An exemplary flowchart for constructing a preset resident area model is provided for the embodiments of the present application. As shown in Figure 2 The multimedia information recommendation device first selects the map range (recommended area) according to the demand. Then, for the selected map range (recommended area), the real-time location signaling initial data set A (historical location signaling data) is obtained from the operator (network service end), and the corresponding area user data set B (historical area user data) is obtained from the short video service provider (multimedia service end). Then, the users with the same identification information IMSI in the data set A and the data set B are compared for geohash matching (resident annotation is performed on the data set A using the data set B, that is, the data in the data set A is annotated, the resident annotation is 1, and other annotations are non-resident 0), to obtain the training data set C (training sample data). According to the training sample data, an initial area resident logistic regression model (preset resident area model) is constructed using a logistic regression algorithm. One data record in the data set A is shown in Table 1, one data record in the data set B is shown in Table 2, and one data record in the training sample data C is shown in Table 4.

[0074] S103, based on the resident area prediction result, determining resident users from the users, and pushing the multimedia information to be recommended to the terminal corresponding to the resident users.

[0075] It should be noted that in the embodiments of the present application, the multimedia information to be recommended can be a variety of media forms such as promotional videos, audios, and images made for public information promotion, for example: tourism promotional videos / audios / images, anti-fraud videos / audios / images, health promotion videos / audios / images, or other any videos / audios / images for information promotion. The specific multimedia information to be recommended can be set according to actual needs and application scenarios, and the present application does not limit this.

[0076] Specifically, in the embodiment of the present application, the multimedia information recommendation device determines the resident users from the users based on the resident area prediction result, comprising: obtaining the area user data in the recommended area from the multimedia service end, and cutting out the slice user data of the target time period from the area user data; determining the resident users from the users by using the slice user data and the resident area prediction result.

[0077] It should be noted that, in the embodiment of the present application, the multimedia information recommendation device can obtain the area user data in the recommended area from the multimedia service end, which includes the user data of all time, and the area user data is used to combine the resident area prediction result based on the location signaling data in the target time period, so the slice user data of the same time period is cut out from the area user data to combine the slice user data and the resident area prediction result to determine the resident users from the users.

[0078] Specifically, in the embodiment of the present application, the multimedia information recommendation device determines the resident users from the users by using the slice user data and the resident area prediction result, comprising: determining the users whose resident area prediction result represents the first value as the first resident users, and determining the users whose resident area prediction result represents the second value as the first non-resident users from the users; calibrating the first non-resident users to determine the second resident users and the second non-resident users by using the slice user data; and determining the first resident users and the second resident users as the resident users.

[0079] For example, the slice user data includes the communication identification information of a plurality of users, if the slice user data includes the communication identification information of a user, but the resident area prediction result of the user is non-resident, the user is calibrated as a resident user, i.e. the second resident user is obtained; if the slice user data does not include the communication identification information of a user, and the resident area prediction result of the user is also non-resident, the user is the second non-resident user. In this way, the first resident users and the second resident users are determined as the resident users, which can further calibrate the resident users predicted based on the resident area prediction model, and determine the resident users from the predicted first non-resident users, thereby reducing the error of the prediction model and improving the accuracy of the determination of the resident users in the area.

[0080] Specifically, in the embodiment of the present application, after the multimedia information recommendation device calibrates the first non-resident users to determine the second resident users and the second non-resident users by using the slice user data, the following steps can also be performed: correcting the resident area prediction result corresponding to the second non-resident users to the first value.

[0081] It should be noted that in the embodiment of the present application, after the multimedia information recommendation device calibrates the first non-resident user with the slice user data, the obtained resident area prediction result corresponding to the second non-resident user is corrected to the first value, that is, 1.

[0082] Specifically, in the embodiment of the present application, after the multimedia information recommendation device corrects the resident area prediction result corresponding to the second non-resident user to the first value, the following steps can also be performed: according to the resident area prediction result of each user, the location signaling data is labeled as resident to obtain labeled location signaling data; using the labeled location signaling data, the preset resident area model is iteratively updated to determine the latest preset resident area model for use in the next resident prediction.

[0083] An exemplary resident labeling implementation can be: if the resident area prediction result of a user is resident, the location signaling data corresponding to the user is labeled as resident, that is, at least one network index data corresponding to the user is labeled as resident, that is, the resident label value is 1. The labeled location signaling data is shown in Table 3. Then, using the labeled location signaling data, the preset resident area model is iteratively updated to determine the latest preset resident area model for use in the next resident prediction. In this way, the preset resident area model can be iteratively updated in real time, ensuring the accuracy of the preset resident area prediction model.

[0084] It can be understood that using short videos instead of traditional short message messages and WeChat messages, the video itself has better expression effect than text, can express more information, and the demand unit can convey more useful information through the official short video, and it is more convenient for people of different age levels and different cultural levels to receive and understand information. The information touch ability and information publicity level of the demand unit are improved.

[0085] Figure 3 An exemplary model iteration training process diagram is provided for the embodiment of the present application. As shown in FIG. 4, the model iteration training process diagram includes the following steps: Figure 3As shown, the multimedia information recommendation device obtains the real-time location signaling test dataset A1 (location signaling data) of the recommendation area from the operator (network server). Next, the dataset A1 of the recommendation area is input into the regional resident logistic regression model (preset resident area model) for prediction, resulting in the prediction dataset C1 (resident area prediction result). Then, the regional user dataset B1 (regional user data) obtained from the short video service provider (multimedia server) is used to correct and verify the prediction dataset C1. Specifically, if the resident label value of the terminal's IMSI in dataset C1 is 0, and this IMSI exists in the regional user dataset B1, then the resident label value of this IMSI in C1 is corrected to 1, resulting in the final target dataset C2 for sending the short video (multimedia information to be recommended). Finally, the target dataset C2 is input into the regional resident logistic regression model for iterative model update training. A data record from the hourly slice test data A1 is shown in Table 1; a data record from the regional user dataset B1 is shown in Table 2; and a data record from C1 and C2 is shown in Table 4.

[0086] Figure 4 This is a schematic diagram illustrating an exemplary process for predicting resident areas, provided as an embodiment of the present invention. Figure 4 As shown, the flowchart for predicting the resident area consists of a schematic diagram of the process for constructing a pre-defined resident area model (see [link]). Figure 2 ) and a flowchart of the model iterative training process (see Figure 3 )composition.

[0087] For example, the main functions of the resident area prediction include two parts: the initial construction of a regional resident population model based on logistic regression, model prediction, model training, and short video recommendation. The logistic regression regional resident model construction and updating uses operator data sources from: base station information dimension tables (fields include: base station id, geohash) and real-time user trajectory signaling data (fields include: IMSI, base station id).

[0088] The construction of the logistic regression-based regional population resident model (preset resident area model) involves identifying the base station list within the selected area (recommended area) based on the map area (recommended area) of the demand unit and the base station data (preset base station list) of the operator. This is combined with real-time location signaling data (historical location signaling data) and user location data (historical regional user data) from short video service providers (multimedia servers). Through calculation, the initial training dataset and the initial logistic regression regional resident model are obtained. The detailed steps are as follows:

[0089] Step 1: Obtain the user's map circle selection area input information, such as the circle center's longitude and latitude [lngc, latc] and the radius Rc if the circle selection area is a circle; or obtain the polygon vertex set information {[lng1, lat1][lng2, lat2][lng3, lat3][lng4, lat4]...} if the circle selection area is a polygon.

[0090] Step 2: Area construction, scan the base station table, and determine whether the base station's longitude lngs and latitude lats are within the area selected in step 1-1: (1) circle, calculate the distance H between the base station's longitude and latitude and the circle center, if H < Rc, then the base station is located within the circular area, (2) polygon, use the ray method to identify all base stations within the polygon. After scanning is completed, the base station list station-list within the selected area is obtained.

[0091] Step 3: Initial logistic regression area resident model (preset resident area model) construction, using data includes: operator area real-time location signaling initial data set A (historical location signaling data), see Table 1 for specific data structure, short video service provider area user data set B (historical area user data), see Table 2 for specific data structure.

[0092] Step 3-1: The operator area real-time location signaling initial data set A is the initial 1-hour time slice operator area real-time location signaling data (location signaling data), including all user IMSI values, geohash values (8 bits), and weight values Q within the last hour in the target area (recommended area). The short video service provider area user data set B is the IMSI value (first communication identifier information) and geohash value (first historical location encoding) (8 bits) of all users in the target area provided by the short video service provider (multimedia information service end) within the last hour. According to the geohash length and distance error table in Table 3, the operator area real-time location signaling initial data set A and the short video service provider area user data set B are matched according to IMSI. (1) geohash eight bits are equal, the weight value Q is increased by 2; (2) geohash seven bits are equal, the weight value Q is increased by 1; (3) geohash six bits are equal, the weight value Q is increased by 0.5; after all data processing is completed, the data set A is labeled, Q value is greater than or equal to 1, labeled as resident user 1, and others are labeled as non-resident user 0. The initial training data set C (see Table 3 for specific data structure) is obtained. If there are multiple records for a certain IMSI in the operator area real-time location signaling initial data set A, and there are also multiple corresponding IMSI records in the short video service provider area user data set B, then the maximum weight value in the associated record is taken as the weight value of the IMSI.

[0093] Step 3-2: According to the labeled initial training data set C obtained in step 3-1, using the main zone RSRP, the main zone RSRQ, TADV, the adjacent zone RSRP1, the adjacent zone RSRQ1, the adjacent zone RSRP2, the adjacent zone RSRQ2, the adjacent zone RSRP3, the adjacent zone RSRQ3, the adjacent zone RSRP4, the adjacent zone RSRQ4, the adjacent zone RSRP5, and the adjacent zone RSRQ5 as the feature field, the initial training of the model is carried out using the logistic regression algorithm to obtain the initial logistic regression area resident model.

[0094] Step 4: Logistic regression area resident model prediction and training: the operator real-time location signaling data (location signaling data set) is divided by hour to obtain the hourly sliced location signaling data (sliced location signaling data), and the hourly sliced test data set A1 of the target area (the location signaling data of each user) is obtained from the hourly sliced location signaling data through the area base station list (target base station) in step 2 above.

[0095] Step 4-2: The hourly sliced test data set A1 obtained in step 4-1 (see Table 1) is used as the feature field, and the model prediction is carried out according to the logistic regression area resident model to obtain the prediction data set C1 (see Table 4 for the specific data structure).

[0096] Step 4-3: The prediction data set C1 obtained in step 4-2 is used to correct and verify the data with the area user data set B1 (area user data) of the short video service provider: if the resident label value of the user IMSI in the prediction data set C1 is 0 and the IMSI exists in the data set B1 (see Table 2 for the specific data structure), the resident label value of the IMSI in the prediction data set C1 is corrected to 1 to obtain the final short video sending target data set C2 (see Table 4 for the specific data structure).

[0097] Step 4-4: The target data set C2 in step 4-3 is used for model iterative update training.

[0098] Step 5: Short video recommendation process, including the demand unit registering an official account with TikTok, Kuaishou, etc. short video service providers, purchasing short video advertising positions, and obtaining the required IMSI data set from the target data set C2 in step 4, combining the user-related information (user ID, mobile phone number, etc.) of the short video service provider to perform regional short video recommendation and achieve the publicity and promotion effect of the demand unit.

[0099] The application provides a multimedia information recommendation method, position signaling data of each user in a recommendation area is acquired; the user is a user on a terminal covered in the recommendation area; a preset resident area model is used to predict a resident area of each user according to the position signaling data, and a resident area prediction result corresponding to each user is obtained; based on the resident area prediction result, a resident user is determined from each user, and multimedia information to be recommended is pushed to a terminal corresponding to the resident user. The multimedia information recommendation method provided by the application can accurately lock the resident user in the area for multimedia information recommendation, improve the accuracy of information recommendation, and push information to the resident user in the area by taking multimedia as a carrier, so that the expression effect tends to diversification, different age levels and different cultural levels of people can receive and understand information, and the diversity of information recommendation is improved.

[0100] The application provides a multimedia information recommendation device, Figure 5 The application provides a multimedia information recommendation device, Figure 1 As shown in Figure 5 The application provides a multimedia information recommendation device,

[0101] The application provides a multimedia information recommendation device,

[0102] The application provides a multimedia information recommendation device,

[0103] The application provides a multimedia information recommendation device,

[0104] In an embodiment of the application, the push module 503 is further configured to acquire area user data in the recommendation area from a multimedia service end, and cut out slice user data of a target time period from the area user data; and the resident user is determined from each user by using the slice user data and the resident area prediction result.

[0105] In an embodiment of the present application, the pushing module 503 is further configured to determine, from the users, a user whose resident area prediction result represents a first value as a first resident user, and a user whose resident area prediction result represents a second value as a first non-resident user; perform resident calibration on the first non-resident user by using the slice user data to determine a second resident user and a second non-resident user; and determine the first resident user and the second resident user as the resident users.

[0106] In an embodiment of the present application, the position signaling data includes at least one network communication quality data; and the prediction module 502 is further configured to perform resident area prediction on the users respectively by using the preset resident area model according to the at least one network communication quality data to obtain at least one prediction result corresponding to each user; and determine the resident area prediction result corresponding to each user according to the at least one prediction result corresponding to each user respectively.

[0107] In an embodiment of the present application, the obtaining module 501 is further configured to obtain a position signaling data set from a network service end; split slice position signaling data of a target time period from the position signaling data set; determine target base stations in the recommended area according to base station positions included in a preset base station list; find users corresponding to the target base stations from a preset base station and user corresponding relationship to obtain the users; and filter the position signaling data of the users from the slice position signaling data.

[0108] In an embodiment of the present application, the pushing module 503 is further configured to correct the resident area prediction result corresponding to the second non-resident user to the first value.

[0109] In an embodiment of the present application, the multimedia information recommendation device further includes a training module (not shown in the figure) configured to perform resident labeling on the position signaling data according to the resident area prediction results of the users to obtain labeled position signaling data; and perform iterative update on the preset resident area model by using the labeled position signaling data to determine a latest preset resident area model for use in the next resident prediction.

[0110] In an embodiment of the present application, the training module (not shown in the figure) is further configured to obtain training sample data; and construct the preset resident area model by using a logistic regression algorithm according to the training sample data.

[0111] In one embodiment of the present invention, the training module (not shown in the figure) is further configured to obtain historical location signaling data corresponding to any region from a network server and historical regional user data corresponding to the region from a multimedia server; use the historical regional user data to perform permanent annotation on the historical location signaling data to obtain tag data; and determine the historical location signaling data and the tag data as the training sample data.

[0112] In one embodiment of the present invention, the historical location signaling data includes first communication identification information corresponding to each of a plurality of first users, and at least one historical network indicator data, each historical network indicator data including a first historical location code and first historical network communication quality data; the historical area user data includes second historical communication identification information corresponding to each of a plurality of second users, and at least one second historical location code; the training module (not shown in the figure) is further configured to, for each of the plurality of first users, use the corresponding first communication identification information to find a target user with the same communication identification information from the plurality of second users; for each first user, compare the first historical location code included in each historical network indicator data with at least one second historical location code of the corresponding target user to obtain a historical weight value for each historical network indicator data; select the target network indicator data with the largest historical weight value from the at least one historical network indicator data according to the historical weight value of each of the at least one historical network indicator data, and determine the label data of the corresponding target network indicator data based on the largest historical weight value and a preset weight threshold.

[0113] In one embodiment of the present invention, the training module (not shown in the figure) is further configured to, for each first user, compare the first historical position code included in each historical network indicator data with at least one second historical position code of the corresponding target user for similarity bit comparison, to obtain at least one similarity bit comparison result corresponding to each historical network indicator data; determine a sub-weight value based on each similarity bit comparison result, to obtain at least one sub-weight value corresponding to each historical network indicator data; and determine the sum of the at least one sub-weight value corresponding to each historical network indicator data as the corresponding historical weight value.

[0114] This invention provides a multimedia information recommendation device. Figure 6 A schematic diagram of the structure of a multimedia information recommendation device provided in an embodiment of the present invention. Figure 2 .like Figure 6 As shown, the multimedia information recommendation device includes: a processor 601, a memory 602, and a communication bus 603;

[0115] The communication bus 603 is configured to realize the communication connection between the processor 601 and the memory 602.

[0116] The processor 601 is configured to execute the computer program stored in the memory 602 to realize the multimedia information recommendation method.

[0117] The present application provides a multimedia information recommendation device, acquires the position signaling data of each user in a recommendation area; the user is a user on a terminal covered in the recommendation area; uses a preset permanent area model, and carries out permanent area prediction on each user according to the position signaling data, obtains the permanent area prediction result corresponding to each user; based on the permanent area prediction result, determines the permanent user from each user, and pushes the multimedia information to be recommended to the terminal corresponding to the permanent user.The multimedia information recommendation device provided by the present application is based on the position signaling data of each user in the recommendation area, and the permanent user in the area is predicted, the permanent user is effectively inferred, the permanent user in the area can be accurately locked to recommend multimedia information, the accuracy of information recommendation is improved;And the permanent user in the area is pushed with multimedia as a carrier, so that the expression effect tends to diversification, can meet the information receiving and understanding of people of different age levels and different cultural levels, and the diversity of information recommendation is improved.

[0118] The present application provides a computer readable storage medium, the computer readable storage medium stores one or more computer programs, the one or more computer programs can be executed by one or more processors to realize the multimedia information recommendation method described above.The computer readable storage medium can be volatile memory (volatile memory), such as random access memory (Random-Access Memory, RAM);Or non-volatile memory (non-volatile memory), such as read-only memory (Read-Only Memory, ROM), flash memory (flash memory), hard disk (Hard Disk Drive, HDD) or solid state disk (Solid-State Drive, SSD);It can also be a device including one or any combination of the above-mentioned memories, such as mobile phone, computer, tablet device, personal digital assistant, etc.

[0119] Those skilled in the art will appreciate that embodiments of the application can be devised for a variety of applications. It is therefore intended that the disclosure of embodiments of the application be considered in a descriptive sense only and not limiting as such while

[0120] The present application is described in reference to the drawings using a flowchart and / or a block diagram of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagram, and combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing system or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the functionality

[0121] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the functionality

[0122] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 means for carrying out each of the functionality

[0123] The above description is only specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any changes or substitutions easily conceived by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A multimedia information recommendation method, characterized in that, The method includes: Obtain the location signaling data of each user within the recommended area; the users are those on terminals covered within the recommended area. Using a preset permanent area model, the permanent area of ​​each user is predicted based on the location signaling data, and the prediction result of the permanent area of ​​each user is obtained. The system obtains regional user data within the recommended area from the multimedia server and extracts slice user data for the target time period from the regional user data. From the users, the user whose permanent area prediction result is represented by a first value is determined as the first permanent user, and the user whose permanent area prediction result is represented by a second value is determined as the first non-permanent user. Using the sliced ​​user data, the first non-resident user is calibrated to be a permanent user, and the second permanent user and the second non-resident user are determined. The first resident user and the second resident user are identified as the resident user; The multimedia information to be recommended will be pushed to the terminal corresponding to the resident user.

2. The method according to claim 1, characterized in that, The location signaling data includes at least one network communication quality data point; the step of using a preset permanent area model to predict the permanent area of ​​each user based on the location signaling data, and obtaining the permanent area prediction result for each user, includes: Using the preset permanent area model, the permanent area of ​​each user is predicted based on the at least one network communication quality data, and at least one prediction result is obtained for each user. Based on the at least one prediction result corresponding to each user, the prediction result of the permanent residence area corresponding to each user is determined.

3. The method according to claim 1, characterized in that, The acquisition of location signaling data for each user within the recommended area includes: Obtain the location signaling dataset from the network server; Slice location signaling data for the target time period from the location signaling dataset; Based on the locations of base stations included in the preset base station list, the target base stations within the recommended area are determined; From the preset base station-user correspondence, find the user corresponding to the target base station, thereby obtaining each user; The location signaling data of each user is filtered out from the slice location signaling data.

4. The method according to claim 1, characterized in that, After using the sliced ​​user data to perform resident calibration on the first non-resident user and determine the second resident user and the second non-resident user, the method further includes: The predicted permanent location for the second non-permanent user is corrected to the first value.

5. The method according to claim 4, characterized in that, After correcting the predicted permanent area of ​​the second non-resident user to the first value, the method further includes: Based on the prediction results of the permanent residence areas of each user, the location signaling data is labeled with permanent residence information to obtain labeled location signaling data. Using the labeled location signaling data, the preset permanent area model is iteratively updated to determine the latest preset permanent area model for use in the next permanent area prediction.

6. The method according to any one of claims 1 to 5, characterized in that, Before using a preset permanent location model to predict the permanent location of each user based on the location signaling data, and obtaining the predicted permanent location result for each user, the method further includes: Obtain training sample data; Based on the training sample data, the preset permanent area model is constructed using a logistic regression algorithm.

7. The method according to claim 6, characterized in that, The acquisition of training sample data includes: For any given area, obtain historical location signaling data corresponding to that area from the network server, and obtain historical area user data corresponding to that area from the multimedia server; Using the historical area user data, the historical location signaling data is permanently labeled to obtain tag data; The historical location signaling data and the tag data are determined as the training sample data.

8. The method according to claim 7, characterized in that, The historical location signaling data includes first communication identification information corresponding to each of the multiple first users, and at least one historical network indicator data, each of which includes a first historical location code and first historical network communication quality data; the historical area user data includes second historical communication identification information corresponding to each of the multiple second users, and at least one second historical location code. The step of using the historical area user data to perform permanent labeling on the historical location signaling data to obtain tag data includes: For each of the plurality of first users, using the corresponding first communication identification information, a target user with the same communication identification information is found from the plurality of second users; For each first user, the first historical location code included in each historical network indicator data is compared with at least one second historical location code of the corresponding target user to obtain the historical weight value of each historical network indicator data. Based on the historical weight values ​​of at least one historical network indicator data, the target network indicator data with the largest historical weight value is selected from the at least one historical network indicator data, and the label data of the corresponding target network indicator data is determined based on the largest historical weight value and a preset weight threshold.

9. The method according to claim 8, characterized in that, For each first user, the first historical location code included in each historical network indicator data is compared with at least one second historical location code of the corresponding target user to obtain the historical weight value of each historical network indicator data, including: For each first user, the first historical location code included in each historical network indicator data is compared with at least one second historical location code of the corresponding target user by the number of similarity bits, so as to obtain at least one similarity bit comparison result for each historical network indicator data. Based on the comparison results of each similarity bit, a sub-weight value is determined, and at least one sub-weight value corresponding to each historical network indicator data is obtained; For each historical network indicator data, the sum of the corresponding at least one sub-weight value is determined as the corresponding historical weight value.

10. A multimedia information recommendation device, characterized in that, include: Processor, memory, and communication bus; The communication bus is used to realize the communication connection between the processor and the memory; The processor is configured to execute a computer program stored in the memory to implement the multimedia information recommendation method according to any one of claims 1-9.

Images