Method and system for time-sharing recommendation of media information bits, medium, product

Abstract

The application provides a media information bit time-sharing recommendation method and system, a computer readable storage medium, and a computer program product. The media information bit time-sharing recommendation of the application first determines the target user group of each media information bit in the divided time-sharing carousel time period according to the user itinerary, estimates the click expectation value of the goods to the media information bit in the time-sharing carousel time period for the target user group, and determines the media information bit corresponding to the high click expectation value for the goods to be put in and the time-sharing carousel time period on the media information bit according to the numerical value of the click expectation value, so as to find the media information bit corresponding to the high click expectation value and the time-sharing carousel time period for the goods to be put in, accurately select the media information bit and the time-sharing carousel time period on the media information bit, and accurately put the goods into the target user group.

Description

Technical Field

[0001] This application relates to the field of computer application technology, specifically to a time-sharing recommendation method and system for media information bits, a computer-readable storage medium, and a computer program product. Background Technology

[0002] In public settings such as travel, there are numerous media information slots. Many products continuously play relevant product media content to passersby by placing ads in these media slots, hoping to stimulate the purchase of the advertised products and ultimately achieve value conversion for the target audience passing by the media slots.

[0003] Furthermore, to increase the capacity of product placement in media slots, product placement in these slots is often corresponding to a specific time-sharing rotation period. Multiple products are placed in the same media slot, each with its own corresponding time-sharing rotation period. In other words, different products are placed in different time-sharing rotation periods in the media slot, achieving the goal of multiple products being displayed in a cyclical manner on that media slot.

[0004] In this context, the key challenge for product placement is determining which media placement to choose and during which time slot of that placement to ensure the product is accurately targeted to the target audience at the right time and location. Summary of the Invention

[0005] One objective of this application is to solve the technical problem of how to accurately select media information slots and time-sharing periods on media information slots so that products can be accurately delivered to the target audience.

[0006] According to one aspect of the embodiments of this application, a time-sharing recommendation method for media information bits is disclosed, the method comprising:

[0007] For media information slots divided into time-sharing carousel periods, the target user group for placing products on the media information slots during the time-sharing carousel periods is obtained based on the user's schedule.

[0008] Estimate the expected clicks of products placed on the media information slots for the target user group, and obtain the expected click values ​​of the products on the media information slots during the time-sharing carousel period;

[0009] Based on the expected click value of the product in each media information position corresponding to each time-sharing carousel period, media information positions with high expected click value and time-sharing carousel periods in the media information positions are recommended for the product.

[0010] According to one aspect of the embodiments of this application, the method of dividing media information slots into time-sharing playback periods and obtaining the target user group for the products displayed on the media information slots during the time-sharing playback periods based on user schedules includes:

[0011] Corresponding to the split time-sharing carousel period, the user's travel data is obtained, and the travel data matches the time-sharing carousel period in time;

[0012] By using the user's travel data during the time-sharing period, the station to which the user's trip belongs at that time is determined;

[0013] Users belonging to the site during the time-sharing carousel period are aggregated to form the target user group for the media information positions mapped by the site to display products during the time-sharing carousel period.

[0014] According to one aspect of the embodiments of this application, the step of estimating the click expectation of placing products on the media information position for the target user group and obtaining the click expectation value of the products on the media information position during the time-sharing carousel period includes:

[0015] For the user's travel data during the time-sharing period, the number of times the user exits at each gate in the subway station gate group is obtained from the gate information based on the user's exit station and exit gate.

[0016] The exit gate to which a user belongs during a specific time slot is determined based on the distribution of the number of times a user exits the subway station at each gate. The exit gate is the exit that the user can pass through when exiting the subway station.

[0017] According to one aspect of the embodiments of this application, the step of estimating the click-through rate of each product for each user within the target user group during the time-sharing broadcast period in the media information position includes:

[0018] For the target user group of the media information position during the time-sharing rotation period, user-product pairs are constructed by using the user profile of each user and the product characteristics corresponding to the product;

[0019] By predicting the click-through rate of the user's product pair, the click-through rate of the media information position for each user in the target user group during the time-sharing carousel period is obtained.

[0020] According to one aspect of the embodiments of this application, the user profile corresponds to the user's own travel data and a combination of the own travel data and third-party supplementary data.

[0021] According to one aspect of the embodiments of this application, recommending media information positions with high click-expectation values ​​and time-sharing periods for the product based on the click-expectation values ​​of the product in each media information position corresponding to each time-sharing period includes:

[0022] By comparing the expected click value of the product in each time slot of each media information position, the media information position with the highest expected click value and the time slot of the product in the media information position are recommended to the product.

[0023] According to one aspect of the embodiments of this application, the media information slot of the time-sharing rotation period is competed for by two or more products;

[0024] The method further includes recommending media information positions with high click-expectation values ​​for the product based on the click-expectation values ​​of the product in each media information position corresponding to each time-sharing carousel period, and recommending media information positions with high click-expectation values ​​after the time-sharing carousel period of the media information positions.

[0025] For two or more products competing for the media information slot, the click value of each product is obtained;

[0026] The conversion value of each product is calculated using the click value and the expected click value of the product in the media information position during the competing time slot carousel period.

[0027] The time-sharing rotation period of the media information slot will be recommended for the placement of high-conversion-value products.

[0028] According to one aspect of the embodiments of this application, a media information slot time-sharing recommendation system based on user travel profile is disclosed. The system includes a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps of the method as described above.

[0029] According to one aspect of the embodiments of this application, a computer program product is disclosed, including a computer program that is executed by a processor to implement the steps of the method as described above.

[0030] This application embodiment addresses media information positions and their segmented time-sharing periods. First, based on user activity, it determines the target user group for each media information position within its segmented time-sharing period. Then, it estimates the expected click value for the target user group when placing products on the media information positions during the time-sharing period. Based on this expected click value, it identifies the media information positions with high expected click values ​​for the products to be advertised, along with the corresponding time-sharing periods. This process ensures accurate selection of media information positions and their time-sharing periods, guaranteeing precise targeting of products to the intended audience.

[0031] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0032] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0033] The above and other objectives, features and advantages of this application will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0034] Figure 1 A flowchart of a time-sharing recommendation method for media information bits according to an embodiment of this application is shown.

[0035] Figure 2 It is based on Figure 1 The corresponding embodiment shows a flowchart of a method for obtaining the target user group of products displayed in the media information slots during the time-sharing carousel period based on the user's schedule.

[0036] Figure 3 It is based on Figure 2 The corresponding embodiment shows a flowchart describing the steps of determining the station to which a user's trip belongs based on the user's travel data during the time-sharing period.

[0037] Figure 4 It is based on Figure 3 The corresponding embodiment shows a flowchart describing the steps of determining the exit gate to which a user belongs during a time-sharing period based on the distribution of the number of times a user exits the subway station at each gate group.

[0038] Figure 5 This is a flowchart illustrating a method for determining a user's assigned exit by exit bias statistics according to an exemplary embodiment.

[0039] Figure 6It is based on Figure 1 The flowchart described in the corresponding embodiment describes the steps of estimating the click expectation of products placed on media information positions for target user groups and obtaining the click expectation value of products on media information positions during time-sharing periods.

[0040] Figure 7 It is based on Figure 6 The corresponding embodiment shows a flowchart describing the steps of estimating the click-through rate of each product for each user group within the target user group during the time-sharing rotation period of the media information position. Detailed Implementation

[0041] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided to make the description of this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art. The drawings are merely illustrative of this application and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted.

[0042] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more exemplary embodiments. Numerous specific details are provided in the following description to give a full understanding of exemplary embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced with one or more of the specific details omitted, or other methods, components, steps, etc., can be employed. In other instances, well-known structures, methods, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0043] Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0044] To effectively promote products through media placement, an area often features numerous media display slots. Each slot rotates different products at different times for passersby to see. By observing the products displayed on these media slots, passersby become aware of the products and are subsequently drawn to them, leading to purchases. Therefore, the rotation of products across media slots is effective in converting product placement into value.

[0045] Therefore, for the placement of a product, how to ensure that the product is placed at the right time and place to the target user group is a problem that media information placement recommendation urgently needs to solve.

[0046] Therefore, this application provides a time-sharing recommendation method for media information slots, which recommends media information slots for products based on the time periods and locations traversed by the target user group, as well as the time-sharing rotation period of the media information slots.

[0047] See Figure 1 , Figure 1 A flowchart of a time-division recommendation method for media information bits according to an embodiment of this application is shown. This application provides a time-division recommendation method for media information bits, the method comprising:

[0048] Step S110: For the time-sharing carousel period of the media information slot, obtain the target user group for placing products in the media information slot during the time-sharing carousel period based on the user's schedule;

[0049] Step S120: Estimate the expected clicks of products placed on media information positions for the target user group, and obtain the expected click values ​​of products on media information positions during the time-sharing carousel period;

[0050] Step S130: Based on the expected click value of the product in each media information corresponding to each time-sharing carousel period, recommend media information positions with high expected click value and time-sharing carousel periods in the media information positions for the product.

[0051] These steps are explained in detail below.

[0052] In step S110, the media information position corresponds to multiple time-sharing rotation periods. The products placed to the media information position will play the content according to their corresponding time-sharing rotation periods, so that multiple products will be rotated in a loop at the media information position.

[0053] Time-sharing rotation is a time-based segmentation of media playback for media information positions, and the segmented time periods are the time-sharing rotation periods for the media information positions.

[0054] For each time-sharing period of a media information segment, step S110 will determine the target user group for that media information segment in each time-sharing period. The target user group for that media information segment in a time-sharing period refers to the people who pass by that media information segment during that time-sharing period.

[0055] To further explain, the target user group determined by the time-sharing period of the media information slot based on the user's schedule consists of all users who pass through the media information slot during this time-sharing period.

[0056] Therefore, in step 110, the media information positions that each user passes through and the time of passing through the media information positions are determined according to the user's itinerary. For example, the media information positions covered by the travel trajectory corresponding to the user's itinerary are the media information positions that the user passes through, and the time period that can be matched with the travel time is the time-sharing time period of the media information positions that the user passes through.

[0057] In one exemplary embodiment, all users corresponding to the media information position in each time-sharing period are obtained based on the user journeys that occur in the divided time-sharing period, so as to form the target user group of the media information position in each time-sharing period.

[0058] By analogy, the target user group for each media information position during each time slot will be obtained.

[0059] For example, please also see Figure 2 , Figure 2 It is based on Figure 1 The corresponding embodiment shows a flowchart of a method for obtaining the target user group of products displayed in the media information slots during the time-sharing carousel period based on the user's schedule.

[0060] The step S110 of the embodiment of this application, which involves splitting media information slots into time-sharing carousel periods and obtaining the target user group for the products displayed in the media information slots during the time-sharing carousel periods based on the user's itinerary, includes:

[0061] Step S111: Corresponding to the split time-sharing carousel period, obtain the user's travel data, which is consistent with the time-sharing carousel period in terms of time.

[0062] Step S112: Determine the station to which the user's trip belongs during the time-sharing period by using the user's travel data during the time-sharing period;

[0063] Step S113: Aggregate users belonging to the site during the time-sharing carousel period to form the target user group for the media information positions mapped by the site to deliver products during the time-sharing carousel period.

[0064] These steps are explained in detail below.

[0065] First, it should be noted that the determination of the target user group for a media information position in each time slot will be based on user travel profiles, i.e., user itineraries, in order to accurately locate users for the media information position in each time slot.

[0066] The user travel profile referred to includes, but is not limited to, user itineraries in travel scenarios such as subways and buses.

[0067] As users travel, such as taking the subway, their travel behaviors generate corresponding trip data. For example, entering and exiting a subway station generates a user trip, which exists in the form of trip data. Therefore, it should be understood that if a user's travel occurs within a time-sharing loop and is spatially associated with a media information segment, then that media information segment and its time-sharing loop period correspond to the current user. The current user becomes the target user for that media information segment during that time-sharing loop period, forming a target user group with other users.

[0068] Therefore, in the execution of step S111, the travel data corresponding to each time-sharing ...

[0069] OriginDestination (OD) data indicates a user's journey and includes the user identifier, the station identifier that triggered the trip, and the trigger time of the trip. For example, for subway travel data, the user's journey data indicates the user's journey when entering and exiting the station through the turnstiles, as well as the exit station and exit turnstile involved in the journey. The user journey includes fields such as journey identifier (journey ID), user identifier (user ID), city identifier (city ID), line number, entry time, entry station code, entry turnstile code, exit time, exit station code, and exit turnstile code.

[0070] Therefore, it is possible to obtain user travel data for each segment of the time-sharing carousel, thereby adapting the target user group for each media information position to each segment of the time-sharing carousel.

[0071] For each time-sharing carousel period, after obtaining the user's travel data for each time-sharing carousel period under the action of step S111, the execution of step S112 can determine which station in the travel for the user who generated the travel trip during each time-sharing carousel period.

[0072] As is understandable, the station to which a user belongs is the station that triggered their travel behavior during their journey. For example, in a subway trip, the entry and exit stations indicated by the user's journey are the stations to which the user belongs during the corresponding time-sharing period.

[0073] In other words, during this time-sharing playback period, when a user passes through the site they belong to, it becomes clear that the user will also pass through the media information segment mapped by that site.

[0074] Therefore, by taking user travel, i.e. user journey, as the entry point, we can determine the target user group for the time-sharing products of each media information location, and accurately locate all users passing through each media information location in time and space.

[0075] The acquisition of trip data based on the segmented time-sharing periods provides an orderly and reliable data source for determining the station to which a user's trip belongs at each time segment, thereby ensuring that the station affiliation determined in this way corresponds to the segmented time-sharing periods in time.

[0076] In one exemplary embodiment, user behavior includes subway travel; therefore, the station referred to includes an exit station, which includes several exits.

[0077] Correspondingly, the media information positions are the surrounding areas radiating from the exits. Therefore, each exit has its own mapped media information positions, and the media playback of each media information position is divided into several time-sharing rotation periods.

[0078] Please refer to the following: Figure 3 , Figure 3 It is based on Figure 2 The corresponding embodiment shows a flowchart describing the steps of determining the station to which a user's trip belongs based on the user's travel data during the time-sharing period.

[0079] The step S112 provided in this application embodiment, which determines the station to which a user's trip belongs during a specific time slot based on the user's travel data during the time-sharing period, includes:

[0080] Step S1121: For the user's travel data during the time-sharing period, obtain the number of times the user exits at each gate in the subway station gate group from the gate information based on the user's exit station and exit gate.

[0081] Step S1122: Determine the exit gate to which the user belongs during the time-sharing period based on the distribution of the number of times the user exits the turnstiles at each turnstile in the subway station turnstile group. The exit gate is the exit gate that the user can pass through when exiting the turnstile group.

[0082] The following is a detailed explanation of these two steps.

[0083] First, it should be noted that the travel data belongs to a time-sharing period. Therefore, the number of times a user exits the subway station gate group at each gate, as obtained from the travel data, and the exit gate to which the user belongs, also correspond to this time-sharing period.

[0084] User entry and exit actions triggered by the turnstiles generate corresponding user trips, which are stored in trip data. In an exemplary embodiment, the trip data exists in the form of a user trip table. Each user trip generated when a user enters or exits the turnstile is stored in the constructed user trip table.

[0085] The user's itinerary describes the user's exit gate. In terms of data, the exit gate described by the user's itinerary exists in the form of an exit gate code, and the location of the gate can be obtained from the corresponding gate information.

[0086] In step S1121, the distribution of the number of times a user exits the station gate group gates during the time-sharing period is obtained based on the user's travel data during the time-sharing period. This determines the user's preferred exit direction or even the exit, and then obtains the exit trajectory mapped to the preferred exit direction or even the exit. The media information bits covered by the exit trajectory are the media information bits mapped to the exit.

[0087] The number of times a user exits at each turnstile is calculated from their origin-destination (OD) data. For example, a user's OD data indicates the user's journey when entering or exiting a turnstile, as well as the exit points and turnstiles involved in that journey.

[0088] For each user trip, we can obtain the number of times the user exits from a certain gate at a certain station. Therefore, by counting all user trips of the user, we can determine the number of times the user exits from each gate at a certain station at different times.

[0089] Based on this, and combined with existing turnstile information, the turnstile group to which each turnstile belongs, as well as the line and city to which the turnstile group is located, are determined. Finally, the user is given complete information on the number of times they exit the subway station at each turnstile in the turnstile group at different times.

[0090] In one exemplary embodiment, the trip data exists in the form of a user trip table. Each user trip generated when a user enters or exits the station through the turnstile is stored in the constructed user trip table.

[0091] The fields in the user's itinerary include: date, itinerary identifier (itinerary id), user identifier (user id), city identifier (city id), route number, entry time, entry station code, entry gate code, exit time, exit station code, and exit gate code, etc., which will not be listed here.

[0092] In one exemplary embodiment, the turnstile information is collected based on the site design and is used to indicate the location of each turnstile. The turnstile information exists in the form of a turnstile information table, and the fields in the table include: turnstile code, city identifier, line number, site code, turnstile group number within the site, turnstile number within the site, and equipment type.

[0093] By statistically analyzing travel data and matching travel data with gate information, we can determine the number of times a user exits a station at a particular gate, as well as the location of that gate, especially the gate group to which that gate belongs. This allows us to ultimately determine the number of times a user exits a station at each gate in the subway station gate group at a particular time.

[0094] By statistically analyzing travel data and matching travel data with gate information, we can determine the number of times a user exits a station at a particular gate, as well as the location of that gate, especially the gate group to which that gate belongs. This allows us to ultimately determine the number of times a user exits a station at each gate in the subway station gate group at a particular time.

[0095] By executing step S1121, the turnstiles through which a user exits the subway at each station are recorded and counted, so as to match the exit to the user at that station.

[0096] In the execution of step S111, the travel data of each user during the time-sharing period is obtained. For example, the travel data is the aforementioned user travel table. The user's travel and the exit gate code in the user's travel are obtained from the user travel table.

[0097] To count the number of times a user exits at a turnstile, the system counts the number of times the user exits at the corresponding turnstile based on the turnstile code for each user trip, and determines the turnstile group to which the turnstile belongs based on the turnstile information, thereby ultimately obtaining the number of times the user exits at each turnstile in the subway station turnstile group at different times.

[0098] For example, the turnstile information is represented by a turnstile information table. Data is collected for each turnstile group within each subway station, and for each turnstile within that group, to construct a relational database and obtain the turnstile information table.

[0099] To further explain, the gate information table represents the location of the gate, and the gate code it contains indicates the unique code of the gate within the city. For example, the gate code may consist of a site code, a device type, and a device number, and the device number is the gate station serial number in the gate information table.

[0100] For example, if the site code is 0253, the device type is 01, and the device number is 07, then the gate number is 02530107.

[0101] As mentioned earlier, in addition to the gate code, the fields in the gate information table also include city identifier, line number, station code, gate group number within the station, gate number within the station, and equipment type.

[0102] The city identifier is used to uniquely identify the city where it is located; in other words, each city has a unique city identifier. The station code is used to uniquely identify the station in the city, so that each subway station has a unique code within the city. The gate group number is used to identify each group of gates in the station. For example, if there are 4 rows of gates in the station, they will correspond to the numbers 1-4 respectively. The gate station number is the equipment number. For example, if there are 10 gates in the station, the equipment number can be a number from 1 to 10.

[0103] Based on the gate information table and the user itinerary table, the system obtains the number of times a user exits at each gate in the subway station gate group at different times. This number of exits describes the user's exit behavior. For users who generate user itineraries by traveling by subway, the system obtains the number of times they exit at the exit gates of the subway station gate group corresponding to their destination station at different times. Based on this data, the system accurately maps the user's itinerary to the exit of the corresponding subway station, realizing the exit location of the user during the travel process.

[0104] In step S1121, the user's number of exits at each gate of a subway station is obtained. In step S1122, the distribution of the number of exits at each gate is used to determine the station to which the user belongs, thereby determining the most likely exit for the user's journey.

[0105] In step S1122, the distribution of the number of times a user exits the subway station turnstiles at different times describes the number of times a single user exits the turnstile 1, the number of times they exit the turnstile 2, ..., and the number of times they exit the turnstile n_zj (n_zj is the total number of turnstiles in the entire turnstile group) in the turnstile group.

[0106] For example, on the one hand, the distribution of the number of times a user exits at each gate of a subway station can be used to determine the exit to which the user belongs; on the other hand, the distribution of the number of times a user exits at each gate of a subway station and the user profile can be used as characteristics of the user to predict the exit to which the user belongs.

[0107] Please also see Figure 4 , Figure 4 It is based on Figure 3 The corresponding embodiment shows a flowchart describing the steps of determining the exit gate to which a user belongs during a time-sharing period based on the distribution of the number of times a user exits the subway station at each gate group.

[0108] The step S1122 provided in this application embodiment for determining the exit gate to which a user belongs during a time-sharing period based on the distribution of the number of times a user exits at each gate of a subway station at different times includes:

[0109] Step S301: Based on the number of times the user exits at each gate in the subway station and the user profile, predict the probability of the user's exit direction. The exit direction corresponds to at least one exit.

[0110] Step S302: Determine the exit gate to which the user belongs during the time-sharing rotation period based on the probability of the user's exit direction.

[0111] The following is a detailed explanation of these two steps.

[0112] User profiles are used to describe a user's state and can be composed of constructed user tags. For example, a user profile exists in the form of a user tag table, which includes a user identifier and several user tags corresponding to that user identifier.

[0113] User identification will uniquely identify subway users. User tags include, but are not limited to, gender, age, life stage, occupation, total number of rides in the past 7 / 14 / 30 / 90 days, total ride duration in the past 7 / 14 / 30 / 90 days, average entry time of the first user trip on weekdays in the past 7 / 14 / 30 / 90 days, average exit time of the most recent trip on weekdays in the past 7 / 14 / 30 / 90 days, average commuting distance per trip in the past 7 / 14 / 30 / 90 days, total number of rides, high-frequency stations, total number of times commercial benefits were claimed, total number of times commercial benefits were redeemed, total amount of commercial benefits redeemed, average transaction price per transaction for commercial benefits redeemed, total number of times travel benefits were claimed, number of times each card and coupon was used, and frequency of use of the single-trip card.

[0114] It should be understood that for a user, the number of times they exit the station at different times within a certain time range, such as a single month or a single group of turnstiles, may be relatively low and is affected by various factors such as actual conditions. Therefore, using the distribution of the number of times a user exits the station at different times at each turnstile of the subway station and the user profile as the user's characteristics provides a data foundation for predicting the exit to which the user belongs, and also improves the richness and scalability of the characteristics, which is conducive to achieving accurate prediction based on existing data.

[0115] The exit prediction performed for the user's assigned exit should be explained as follows: the exit predicted is an exit corresponding to the turnstile group, which is the exit that the user can pass through when exiting the turnstile group.

[0116] One turnstile group corresponds to at least one exit. The exit that a user is most likely to pass through after exiting through one of the turnstiles in this turnstile group is the predicted exit to which the user belongs.

[0117] The prediction of a user's assigned exit can be obtained by using a pre-trained classification model to obtain the probability of the user belonging to each exit, and then the probability is used to finally determine the user's assigned exit.

[0118] For example, the pre-trained model can be a binary classification model, which uses the distribution of the number of times a user exits the subway station gate group at different times and the user profile as features. Under this feature, the direction of the user after exiting the station is used as the corresponding direction value of the feature, and each direction value uniquely corresponds to an exit that can be accessed from the gate group.

[0119] For example, if a user with the feature mapping exits the gate group and moves to the left, the labeled direction value is 0; conversely, if a user with the feature mapping exits the gate group and moves to the right, the labeled direction value is 1.

[0120] Based on this, a training dataset is constructed.

[0121] Wherein, each training dataset D train In a single record This is a characteristic of the record, which describes the user status of the corresponding user.

[0122] Specifically, This includes the distribution of the number of times a user exits the subway station at each gate in a given time period and the user profile, and the user profile can be represented by multiple user tags. It is the characteristic of user i in gate group j, R m It is the m-dimensional real space of the m-dimensional feature mapping of user i.

[0123] For example, x ij ∈R m = [Date(YYYYMM), User i's User Tag 1, User i's User Tag 2, ..., User i's User Tag n] label The number of times user i exits through turnstile j in 1 minute, the number of times user i exits through turnstile j in 2 minutes, ..., the number of times user i exits through turnstile j in n minutes. zj [Time-based exit count], the date (YYYYMM) represents the month, such as 202312, which represents December 2023. The date feature indicates the time range corresponding to other features. And user tag n label This refers to the nth term of user i. label n user tags r Represents the number of users, n i n represents the number of turnstiles that user i faces when exiting the subway station. label y represents the number of user tags. ij∈{0,1} represents the direction value of the feature annotation in a single record of the corresponding training dataset. ij ∈{0,1} represents the final exit direction for user i under the corresponding feature (especially in terms of its actual number of exits).

[0124] That is, y in a single record of the training dataset ij This refers to the direction value of the label. For example, if user i exits through gate group j and turns left, then y... ij =0, if to the right, then y ij =1.

[0125] The model is trained using the constructed training dataset to obtain a model that can accurately predict the exit to which a user belongs, providing users with a precise judgment of their exit location without the need for location acquisition, i.e., no longer limited by GPS positioning and location permission acquisition.

[0126] Furthermore, the training for exit prediction will be conducted on turnstile groups that correspond only to a single exit. A training dataset will be constructed for each turnstile group corresponding to a single exit. Features will be constructed based on all records related to that turnstile group, and the constructed features will be labeled with directional values ​​according to the unique orientation of that turnstile group.

[0127] Specifically, the number of times a user exits at each gate in a subway station is obtained through travel data and gate information. This yields user gate group exit data, which can be in tabular form, maintained by fields such as user, subway station, gate group, gate, and number of times exiting at each gate. From this data, user gate group exit data corresponding only to a single exit is obtained. The number of times a user exits at each gate in a subway station is then calculated based on this data, and features are constructed accordingly.

[0128] In addition, corresponding user profiles will be extracted for feature construction, and the constructed features will be labeled with directional values ​​to finally form a training dataset.

[0129] The constructed training dataset shields the influence of other exits, considering only the prediction of the user's exit in the feature dimension. This ensures that subsequent predictions are not affected by other factors, greatly improving the accuracy and reliability of the predictions.

[0130] For example, the distance relationship between the entrance and exit of the turnstile group will also be constructed. The distance relationship between the entrance and exit of the turnstile group is used to indicate the distance of each turnstile in the turnstile group in a station relative to the exit in each direction. So that after predicting the most likely direction of the user after exiting the turnstile, the user can be directly mapped to the nearest exit according to the distance relationship between the entrance and exit of the turnstile group. This exit is the exit to which the user belongs as predicted.

[0131] As mentioned earlier, the model trained using the constructed training dataset can be a binary classification model. This binary classification model will then be used to predict the exit a user belongs to. Specifically, the binary classification model predicts the most likely direction a user will take after exiting the subway station's turnstiles based on the user's time-based exit frequency at each turnstile and their user profile. The exit mapped from this direction is then used as the user's assigned exit.

[0132] To further explain, a predictive dataset is constructed based on the number of times subway users exit at each turnstile in the subway station's turnstile group and their user profiles. Unlike the training dataset, the prediction dataset is not limited to the gate group corresponding to a unique exit. In other words, regardless of whether the gate group corresponds to a unique exit, all data of the subway users will be used to construct the prediction dataset, and then the exit to which the subway user belongs will be predicted by the features of the constructed prediction dataset.

[0133] A station has several turnstile groups for entry and / or exit, and for the turnstile groups for exit, the turnstile group includes a left half group of turnstiles and a right half group of turnstiles. It should be understood that users exiting from the left half group of turnstiles tend to exit in one direction, while users exiting from the right half group of turnstiles tend to exit in another direction.

[0134] Based on the distribution of the number of times a user exits each gate in the station's turnstile group, it can be determined whether the user prefers the left or right half of the turnstile group, and thus determine the preferred exit direction. The user's assigned exit is then determined based on the preferred exit direction. Since this exit is based on the distribution of the number of times a user exits each gate in the station's turnstile group, this exit is the user's assigned exit during the corresponding time-sharing rotation period.

[0135] To further explain, a predictive dataset is constructed based on the time-segmented exit frequency of users at each turnstile in the subway station and user profiles. Unlike the training dataset, the prediction dataset is not limited to the gate group corresponding to a unique exit. In other words, regardless of whether the gate group corresponds to a unique exit, all of the user's data will be used to construct the prediction dataset, and then the exit to which the user belongs will be predicted based on the features of the constructed prediction dataset.

[0136] Prediction dataset D predict In the middle, each record This represents the characteristics of user i in gate group j, whose exit point has not yet been determined. As mentioned earlier, these characteristics describe the number of times user i exits each gate in gate group j and the user's tag. The list of included characteristics is similar to... similar.

[0137] Specifically, to obtain the applicable binary classification model... exit First, the problem loss function for the t-th iteration is constructed as follows:

[0138]

[0139] in, f is the predicted value of the ij-th record in the t-th iteration. t For the t-th tree, f t (x)=w q (x), q represents tree f t The structure, w is a tree f t The weights of the middle leaves, and using w i Let Ω(f) represent the weight of the i-th leaf. t ) is about tree f t The regularization term with complexity L is then L (t) At point f t (x i By performing a second-order approximation, we obtain:

[0140]

[0141] in,

[0142]

[0143] Therefore, the optimal solution to the problem is:

[0144]

[0145] Where q is a tree with a fixed structure, and γ and λ are the coefficients of the first-order and second-order regularization terms, respectively.

[0146] Based on the iterative convergence of the optimal solution in each round on the training dataset, a binary classification model is obtained. exit .

[0147] Based on the obtained binary classification model exit The prediction dataset D will be used. predict Each record Use binary classification models one by one exitThe prediction is performed to obtain the characteristic bias of user i in gate group j, that is, the probability that user i belongs to the exit corresponding to gate group j.

[0148] For example, if the turnstile group j corresponds to two exits. In other words, after exiting the station through turnstile group j, user i has two directions, each corresponding to an exit. The probability that user i will favor the other direction, i.e., belong to the other exit, is...

[0149] At this time, if there are more than two exits in one direction of gate group j, that is, the exits in that direction are not unique, then the binary classification model predicts that user i will be more inclined to the direction with more than two exits after exiting gate group j. The exit that user i is most likely to pass through will be determined according to the flow of each exit. This exit is the exit to which user i belongs.

[0150] Specifically, each exit in one direction has its own traffic flow, which can be sensed by passenger flow counters installed at each exit. The traffic flow corresponding to each exit is used to determine the traffic proportion of each exit, and then the exit to which user i belongs is determined based on the traffic proportion.

[0151] To further explain, determining the exit that user i is most likely to pass through based on the traffic flow at each exit is an execution process that uses a random number to determine user i's affiliation based on a proportional distribution. The proportion referred to is the percentage of traffic flow at each exit.

[0152] After predicting the direction user i will take after exiting gate group j using a binary classification model, if there are multiple exits in the direction the user i is taking, the exit to which user i belongs will be determined based on the traffic proportion of each exit and the random number assigned to user i.

[0153] For multiple exits in the direction of departure, a random number range is constructed for each exit based on its traffic flow ratio. For example, if there are three exits in the same direction, and the traffic flow ratios of each exit are 0.3, 0.2, and 0.5 respectively, then the random number range mapped to each exit is constructed as follows: the random number range for the first exit with a traffic flow ratio of 0.3 is [0, 0.3); the random number range for the second exit with a traffic flow ratio of 0.2 is [0.3, 0.3 + 0.2); and the random number range for the third exit with a traffic flow ratio of 0.5 is [0.3 + 0.2, 0.3 + 0.2 + 0.5].

[0154] For user i, generate a random number between 0 and 1. Determine the exit corresponding to user i based on the range of the random number the generated number falls into. If the random number falls in [0, 0.3), then user i's journey corresponds to the first exit; if the random number falls in the range [0.3, 0.3+0.2), then user i belongs to the second exit; if the random number falls in the range [0.3+0.2, 0.3+0.2+0.5), then user i belongs to the third exit.

[0155] Thus, as mentioned above, in the execution of step S301, features are constructed based on the number of times the user exits at each gate of the subway station gate group at different times and the user tags contained in the user profile. The user is then predicted using a pre-trained model, such as the binary classification model mentioned above, to obtain the probability of each exit direction after the user exits at the gate group of the specified subway station.

[0156] A user can travel in more than one direction after exiting a turnstile. For example, in many cases, a user will have one or two exit directions after exiting a turnstile, and each exit direction has an exit that can be accessed.

[0157] After obtaining the probability of the user's exit direction, step S302 can be used to determine the user's passage direction after exiting the gate group, i.e., the exit direction corresponding to the high probability.

[0158] The exit on the high-probability exit direction is determined. In an exemplary embodiment, if there is only one exit on the high-probability exit direction, the exit closest to the user on that exit direction is determined based on the distance relationship between the high-probability exit direction and the gate group entrance / exit, and this exit is designated as the user's assigned exit.

[0159] In another exemplary embodiment, there are more than two exits in the exit direction corresponding to a high probability. For example, the exit channel in this exit direction branches into more than two exits. In this case, the exit to which the user belongs will be determined from the more than two exits based on the traffic flow ratio of each exit and the obtained random number.

[0160] Therefore, it is no longer necessary to obtain the user's location, that is, it is no longer necessary to obtain location permissions to perform GPS positioning. The user's journey and trajectory can be accurately mapped to the predicted exit, thus enabling accurate prediction of the user's trajectory after exiting the station.

[0161] In another exemplary embodiment of this application, Figure 3The step S1122 shown in the corresponding embodiment, which determines the exit gate to which a user belongs during a time-sharing period based on the distribution of the number of times a user exits at each gate of the subway station gate group, further includes:

[0162] Determine whether the number of times a user exits the subway station at any of the turnstiles during the designated time period is less than a set threshold. If it is less than the set threshold, proceed to step S301; otherwise, proceed to step S301.

[0163] Redirect to execute the steps to determine the user's assigned exit based on exit bias statistics.

[0164] Specifically, it should be noted that if a user frequently exits through a gate group, it means that the gate corresponding to that gate group in the exit direction is the one the user prefers. For such subway users, the exit direction will no longer be predicted. The exit direction of the subway user can be determined simply by statistically analyzing the historical exit directions, thereby reducing computational complexity and improving execution efficiency.

[0165] The number of times a user exits through the turnstiles at a subway station is used to determine whether they frequently exit through those turnstiles. This number can be obtained based on the user's travel data and turnstile information. Furthermore, it can be obtained from user exit data constructed from travel data and turnstile information. However, it is not limited to this; it can also be obtained by statistically analyzing the user's exit data at each turnstile in the subway station. This is not a specific limitation here.

[0166] For example, the threshold can be set to 5. That is, if a user exits a turnstile group at a subway station at least 5 times within a set time range, it means that the exit direction corresponding to the turnstile group and the exit in that exit direction are consistent with the user's exit passage. Thus, it can be determined that the exit is the user's frequently used exit.

[0167] To further explain, for the number of exits not less than a set threshold, the user's preferred gate will be determined. That is, based on the distribution of the user's exits in the gate group, the user's preferred gate group will be counted as either the left or right half of the gate group. Regardless of whether it is the left or right half of the gate group, it corresponds to an exit direction. The exit direction after the user exits can be determined by the preferred gate, and the exit in that exit direction is the exit to which the user belongs.

[0168] It should be added that if the number of turnstiles in a turnstile group is odd, the middle turnstile will belong to both the left and right half of the turnstile group. Therefore, for users exiting the middle turnstile, the number of times the user exits the turnstile group will be counted as the number of times the user exits the turnstile group on the left half and the number of times the user exits the turnstile group on the right half.

[0169] If a user exits the subway station turnstile less than a set threshold, it indicates that neither the subway station nor the turnstile is frequently used by the user. Therefore, it is necessary to construct features based on richer data, such as the distribution of user exit times at each turnstile and user profiles, as mentioned above, and then implement predictions through a pre-built model.

[0170] Please also see Figure 5 , Figure 5 This is a flowchart illustrating a method for determining a user's assigned exit by exit bias statistics according to an exemplary embodiment.

[0171] The steps for determining a user's assigned exit based on exit bias statistics provided in this application embodiment include:

[0172] Step S401: Based on the statistical distribution of the number of times a user exits the subway station at each gate in the turnstile group at different times, determine the turnstile that the user tends to use in the turnstile group.

[0173] Step S402: Determine the exit on the same side as the user's exit during the time-sharing rotation period based on the location of the turnstile.

[0174] The following is a detailed explanation of these two steps.

[0175] For a given turnstile group, the distribution of a user's exit times at each turnstile in the group is statistically analyzed to determine whether the user prefers to exit through the left or right half of the turnstile group.

[0176] Based on the statistically determined user's bias towards the turnstile group, the exit direction and exit can be determined. The exit on the same side is the user's frequently used exit. This exit is designated as the user's assigned exit, and the user's itinerary and travel trajectory are correlated with this exit.

[0177] After determining the station to which a user's travel time belongs through the aforementioned step S112, the users belonging to the same station during the time-sharing broadcast period can be aggregated through step S113. This means that users belonging to the same station are grouped together, and the aggregated users form the target user group of the media information position mapped by that station during the time-sharing broadcast period.

[0178] The user aggregation referred to here is to determine the corresponding subway user group for the location at the exit, and to describe and represent the subway user group obtained by aggregating the subway users through user profiles.

[0179] The aggregated user profiles will be assigned to the site, as well as the media information positions mapped to the site's exit points, to serve as the profiles of those media information positions during the corresponding time-sharing broadcast periods.

[0180] After determining the target user group for each media information position in each time-sharing period under step S110, the expected click value of the product to be advertised in each time-sharing period of the media information position can be determined based on the target user group under step S120.

[0181] In other words, for a product, in order to improve the conversion rate of product placement, it is necessary to determine the recommended media placements and time slots for the product by estimating the expected clicks of each media placement during each time slot.

[0182] Expected click value (CDR) refers to the degree to which a product receives attention from users passing through a media slot during a one-minute carousel. The CDR for a product displayed on that media slot during a one-minute carousel measures the degree to which users passing through that media slot are attracted to the product's media content and the product itself if the product's media content is displayed there during the one-minute carousel.

[0183] It should be understood that the higher the expected click value of a product placed on that media information position, the more suitable the product is for placement on that media information position and the corresponding time-sharing rotation period, and the higher the likelihood that users passing by will be attracted to the product.

[0184] The click expectation of a target user group for a product in a specific media information position and a specific time-sharing carousel period corresponds to the click expectation of all users in the target user group for the product, which is calculated by a pre-trained click-through rate prediction model.

[0185] For example, a click-through rate (CTR) prediction model can be used to predict the CTR of a product among users in the target user group, thereby obtaining the CTR of each user in the target user group.

[0186] The goods referred to are those intended to be placed on media information slots. For example, goods include those recommended for media information slots. When a product is recommended for media information slots so that it can be placed on media during a specific time slot, that product is the goods referred to in this application.

[0187] It should be noted that the expected click value is estimated for a specific time slot on a media information position and for a product targeting all users within the target user group corresponding to that media information position during that time slot.

[0188] The expected clicks are estimated, including the click-through rate (CTR) between users and products, and the execution process of weighted summation of the CTR of all users in the target user group, in order to finally obtain the expected click value of the product when it is displayed in the one-minute carousel of the media information slot.

[0189] In one exemplary embodiment, the click-through rate prediction model for estimating the click-through rate between users and products is trained using the click records of users on exposed products in a travel application.

[0190] Specifically, a training dataset is first constructed for the click-through rate prediction model. Each record in the training dataset includes a user profile of the user and product features constructed for the product. The product is exposed on the travel application and clicked by the user, which corresponds to the user's click record on the exposed product on the travel application.

[0191] Thus, the training dataset is constructed.

[0192] The constructed training dataset D train single record in y ij ,Right now The user profile is composed of multiple data points, including but not limited to data collected from travel applications.

[0193] Specifically,

[0194] Corresponding to the user's own travel data, and This corresponds to supplementary data from a third party. For example,

[0195] For example, as a product feature,

[0196] y ij ∈{0,1}, when the value is 0, it means that no click event occurred after product j was exposed to user i; when the value is 1, it means that a click event occurred after product j was exposed to user i.

[0197] Therefore, travel applications are in the near n d For each user i, Tianzhong considers all n that have been exposed to them. i One product generated These records form the training dataset for the click-through rate prediction model.

[0198] Correspondingly, the click-through rate (CTR) prediction between users and products will be obtained through a trained CTR prediction model. CTR prediction using this model requires constructing a corresponding prediction dataset. Specifically, for each user corresponding to a media information segment, a prediction dataset is built using user profiles and product features. Each record in the prediction dataset represents the constructed user-product pair.

[0199] The constructed prediction dataset is Each record, i.e., user's product pair The feature list is the same as the training dataset.

[0200] Each record in the prediction dataset represents a recommendation of product j to all users corresponding to a media information segment during a time-sharing carousel period, i.e., users 1, 2, ..., n. user Each record will be used as input to the click-through rate prediction model to obtain the click-through rate of each user for the product corresponding to the media information position in the one-minute carousel period. Similarly, the click-through rate of each user for the product corresponding to the media information position in the one-minute carousel period will be obtained, and finally the expected click value of the product in the media information position in the one-minute carousel period will be obtained.

[0201] In another exemplary embodiment, for the click-through rate prediction model, the problem loss function constructed for the t-th iteration is:

[0202]

[0203] in, f is the predicted value of the ij-th record in the t-th round. t For the t-th tree, f t (x)=w q (x) represents tree f t The structure, w is a tree f t The weight of the middle leaf, w i It is the weight of the i-th leaf, Ω(f t ) is about tree f t The regularization term with complexity L is then L (t) At point For second-order approximation, we have:

[0204]

[0205] in,

[0206] The optimal solution is:

[0207]

[0208] Where q is a tree with a fixed structure, and γ and λ are the coefficients of the first-order regularization term and the second-order regularization term, respectively.

[0209] Click-through rate prediction model recommend It is obtained by iterative convergence of the optimal solution in each round on the training dataset.

[0210] The constructed prediction dataset D predict Every record Each item is predicted using a binary classification model, i.e., a click-through rate (CTR) prediction model, to obtain the CTR of all users for the product within a time-sharing carousel period for each media information position. Then, based on the number of times users pass through the media information position during the time-sharing carousel period, a weighted sum of the CTRs is calculated to obtain the expected click value of the media information position for the target user group for the product during the time-sharing carousel period. Where, n user It refers to the number of users in the target user group corresponding to the media information slot during the time-sharing broadcast period. As a weighted sum, the weight is essentially the number of times each user in the target group passes through the media information position during the time-sharing period, and also the number of times a user exits the station at the exit gate to which they belong during the i-time-sharing period, i.e., the number of time-sharing exits.

[0211] In summary, in an exemplary embodiment, the specific execution process of step S120 is as follows: Figure 6 As shown, Figure 6 It is based on Figure 1 The flowchart described in the corresponding embodiment describes the steps of estimating the click expectation of products placed on media information positions for target user groups and obtaining the click expectation value of products on media information positions during time-sharing periods.

[0212] The step S120 provided in this application, which estimates the click expectation of products placed on media information positions for a target user group and obtains the click expectation value of products on media information positions during time-sharing carousels, includes:

[0213] Step S121: For each product, estimate the click-through rate of each user for the product within the target user group during the time-sharing rotation period of the media information position;

[0214] Step S122: Based on the number of times the user passes through the media information during the time-sharing carousel period, the weighted sum of the estimated click-through rate of the product by the user is obtained to determine the media information position during the time-sharing carousel period.

[0215] The media information is positioned within a time-sharing carousel segment targeting a specific user group, i.e., all users within that target group. Each user forms a user-product pair with the product to estimate that user's click-through rate (CTR). CTR represents the product's attractiveness to the user; a higher CTR indicates greater product attractiveness, and vice versa.

[0216] For a user-product pair consisting of a user and a product, the click-through rate (CTR) prediction model described above is used to predict the user's CTR for the product. By analogy, the CTR of each user for the product in the target user group corresponding to a media information position in a time-sharing carousel period can be obtained.

[0217] In other words, each user in a media information slot during a time-sharing period has a corresponding click-through rate for the product. This numerically describes the product's attractiveness to all passing users, and thus allows for the assessment of whether to recommend this media information slot and time-sharing period to the product for media placement, or to evaluate which product is most suitable for placement in this media information slot and its time-sharing period.

[0218] After obtaining the click-through rate of each user for a product, the expected click value of the product corresponding to the media information position and the time-sharing carousel period can be obtained by weighted summation.

[0219] For step S121, please refer to [link / reference needed]. Figure 7 , Figure 7 It is based on Figure 6 The corresponding embodiment shows a flowchart describing the steps of estimating the click-through rate of each product for each user group within the target user group during the time-sharing rotation period of the media information position.

[0220] The step S121 provided in this application embodiment, which estimates the click-through rate of each product for each user within the target user group during the time-sharing rotation period of the media information position, includes:

[0221] Step S1211: For the target user group of the media information position during the time-sharing rotation period, construct user-product pairs by using the user profile of each user and the product features corresponding to the product.

[0222] Step S1212: By predicting the click-through rate of user-product pairs, obtain the click-through rate of each user for a product in the target user group during the time-sharing carousel period of the media information position.

[0223] First, it should be noted that users generate corresponding user trips as they initiate entry and exit behaviors. Numerous users and each user's trips will constitute trip data. In other words, users will exist in the form of user trips in the data. Therefore, it is possible to construct a user profile describing the user's travel based on this data, thereby using the user profile to represent the user and to build user-product pairs with products.

[0224] The media information slots and time-sharing periods used to evaluate product click-through rates belong to a specific website. Therefore, the user journeys used to build user profiles are related to this specific website, thus ensuring the accuracy of the user-product pair features and eliminating useless or even interfering data.

[0225] The system, built around a site-oriented architecture, provides users with travel applications deployed on their personal terminals. Thus, each user only needs to interact with the turnstile via the travel application on their terminal to initiate entry or exit actions, such as scanning a QR code provided by the travel application on the turnstile. Both the travel application running on the user terminal and the systems it interfaces with will obtain the user's own travel data, including their itinerary.

[0226] In other words, the user's own travel data will be used to characterize the user and build a user profile. The user's own travel data consists of multiple parts of data owned by the deployed travel application, such as the user's itinerary data mentioned above.

[0227] In addition, users' own travel data also includes user information on travel apps, data tracking data on travel apps, statistical data based on trip data, user business operation data on travel apps, user-related product data, etc., which will not be listed here.

[0228] For example, the data collected by users on travel apps is used to describe the user's clicks on the exposed products; user information on travel apps includes gender, age, region, and occupation; statistical data based on trip data may include, but is not limited to, the total number of trips in the past 77 / 14 / 30 / 90 days, the total trip duration in the past 7 / 14 / 30 / 90 days, the average arrival time of the first trip on weekdays in the past 7 / 14 / 30 / 90 days, the average departure time of the last trip on weekdays in the past 7 / 14 / 30 / 90 days, the average commuting distance per trip in the past 7 / 14 / 30 / 90 days, the total number of trips, and the top 10 most frequent stations;

[0229] User's commercial operation data on travel apps may include, but is not limited to: total number of times commercial benefits are claimed, total number of times commercial benefits are redeemed, total amount of commercial benefits redeemed, average order value of commercial benefits redeemed, total number of times travel benefits are claimed, number of times each card and coupon is used, and frequency of use of the card.

[0230] User-related product data may include, but is not limited to: the top 10 product categories viewed in the past 1 / 3 / 7 / 14 / 30 / 90 days, the corresponding number of times the top 10 product categories were viewed in the past 1 / 3 / 7 / 14 / 30 / 90 days, and the average price of the top 10 product categories viewed in the past 1 / 3 / 7 / 14 / 30 / 90 days.

[0231] It should be understood that users' own travel data will be adapted to the marketing recommendations output of travel applications as media information slots and their time-sharing rotation periods in order to obtain corresponding user profiles.

[0232] User profiles can be derived from various sources, including users' own travel data and supplementary data from third parties. In short, user profiles correspond to users' own travel data, as well as combinations of their own travel data and supplementary third-party data.

[0233] Third-party supplementary data includes, but is not limited to, whether the individual is pregnant or postpartum, their level of interest in gaming, whether they are frugal, whether they are high-net-worth individuals, whether they frequently travel by air, whether they enjoy financial management, whether they are fans of anime and manga, whether they are frugal, whether they are constantly looking at their phones, whether they enjoy shopping at supermarkets, whether they enjoy taking their children out, whether they intend to buy a car, whether they intend to buy a house or hospital, whether they like to visit museums, historical and cultural sites, and natural scenic areas.

[0234] For each user corresponding to a media information slot and a time-sharing broadcast period, a user profile is obtained through the user's own travel data, or a combination of the user's own travel data and third-party supplementary data. The user profile is then used as the user characteristics in the constructed user-product pair.

[0235] In other words, the user profile, which is numerically derived from the user's own travel data or a combination of the user's own travel data and third-party supplementary data, will serve as the user's characteristics and, together with the product's characteristics, constitute a user-product pair.

[0236] Among them, the product characteristics are used to describe the features of the product. For example, product characteristics include descriptive features, statistical features, and statistical generalization features.

[0237] Specifically, the product characteristics include the product's first / second / third-level category, the product's price, the product's exposure count in the last 1 / 3 / 7 / 14 / 30 days, the product's click count in the last 1 / 3 / 7 / 14 / 30 days, the product's purchase conversion count in the last 1 / 3 / 7 / 14 / 30 days, the product's listing days, the product's exposure count in the last 1 / 3 / 7 / 14 / 30 days in the third-level category, the product's click count in the last 1 / 3 / 7 / 14 / 30 days, and the product's purchase conversion count in the last 1 / 3 / 7 / 14 / 30 days.

[0238] For each user in media information slots and time-sharing rotation periods, user-product pairs are constructed using user profiles and product characteristics. Based on these user-product pairs, the click-through rate of a user for a product can be predicted.

[0239] It should be understood that media information slots, as physical resources that actually exist offline, attract users to the media content on which products are displayed, which is analogous to the clicks on products displayed on the Internet. Therefore, the degree of attraction of products to users through offline exposure of products in time-sharing carousel slots can be accurately measured by the click-through rate of users on products.

[0240] For recommending a media information slot and its time-sharing carousel period, the click expectation value of the product corresponding to the media information slot and carousel period is obtained through the execution of step S120. For multiple products competing for the same media information slot and time-sharing carousel period, the click expectation value of each product on that media information slot and time-sharing carousel period can also be used to determine the product that can obtain the highest click expectation, that is, the product most likely to be attracted by passing users, and thus recommend that media information slot and time-sharing carousel period to that product.

[0241] In summary, by executing step S130, recommendations for media information positions and their time-sharing periods will be made for each media information position and its time-sharing period based on the expected click value of the product on the media information position and its time-sharing period.

[0242] In step S130, based on the expected clicks of the product in the media information position and its time-sharing rotation period, the media information position and its time-sharing rotation period of the product are recommended, so that the media information position and its time-sharing rotation period of the current marketing recommendation can match the most suitable product.

[0243] In an exemplary embodiment, the execution process of step S130 includes:

[0244] By comparing the expected click value of a product across different media information positions and time slots in each time slot, the system recommends the media information positions with the highest expected click value and the time slots in which the product is displayed on those media information positions.

[0245] Based on the expected click value of a product in different media information positions and time-sharing periods, the media information positions and time-sharing periods with the highest expected click value are directly selected for the product.

[0246] For a product, each available media information slot and its available time-sharing period are all objects that can be considered for implementing corresponding media content delivery. Therefore, by executing the aforementioned steps, the expected click value of the product in each media information slot and time-sharing period is obtained. At this time, the expected click value of the product is compared, and the media information slot and time-sharing period with the highest expected click value are recommended to the product.

[0247] This involves selecting media information positions and time-sharing periods with high click-through rates for a product from multiple media information positions and time-sharing periods, and then recommending the selected media information positions and time-sharing periods with high click-through rates to this product.

[0248] In the marketing recommendation of a product for media information slots and time-sharing carousel periods, the expected click value of the product in each media information slot and time-sharing carousel period is estimated for each idle media information slot and time-sharing carousel period, thereby obtaining the expected click value of the product in each media information slot and time-sharing carousel period.

[0249] This is a recommendation of media information slots and time-sharing carousel periods for a single product. If a media information slot and time-sharing carousel period are recommended to more than two products, a situation arises where the same media information slot and the same time-sharing carousel period are competed for by more than two products. In this case, after step S130, the method described above further includes:

[0250] For two or more products in competing media information positions, obtain the click value of each product;

[0251] The conversion value of each product is calculated by combining the click value and the expected click value of the product in the media information slots during the competing time slots of the carousel.

[0252] The time slots for rotating media information displays will be recommended for advertising products with high conversion value.

[0253] When there are two or more competing products for a media information slot and a time-sharing carousel period on it, it is necessary to consider the conversion value of each product based on the expected click value of each product in the media information slot and time-sharing carousel period in order to determine which product the media information slot and time-sharing carousel period should recommend.

[0254] For two or more products competing for the same media placement and time-sharing slot, their corresponding click value is obtained. Click value is related to the product's intrinsic value. For example, a product's click value represents the revenue generated by past users attracted by the exposed product. Specifically, a product's click value can be an estimate of the product's value or the contribution rate of the product's exposure to transactions during media placements and time-sharing slots. Therefore, a product's click value can be an estimate of the product itself or the click value obtained for the product in different media placements and time-sharing slots. This is not limited here and will be determined according to specific operational needs.

[0255] When two or more products are competing for a media information slot and a time-sharing carousel period, that is, when two or more products have calculated the expected click value for that media information slot and time-sharing carousel period, the conversion value of each product in the media information slot and time-sharing carousel period is obtained by multiplying the click value and the expected click value.

[0256] The conversion value of a product in media information slots and time-sharing carousels considers both the degree to which the product is attracted by past users during its exposure in these media information slots and time-sharing carousels, and quantifies the conversion value that exposure in these media information slots and time-sharing carousels can generate. Therefore, the conversion value of a product in media information slots and time-sharing carousels can accurately determine the most suitable products for conversion in terms of media information slots and time-sharing carousels, so that the marketing recommendations in media information slots and time-sharing carousels can prioritize the placement of products with the highest conversion value.

[0257] For example, a media information slot and its time-sharing rotation period are considered premium media, and multiple products compete for it. The click value of each product is then captured by the CTV (Content Viewer). j Then, combine the expected click value of the product in media information positions and time-sharing carousel periods. The conversion value of this product across media placements and time-sharing broadcast periods is calculated as follows:

[0258]

[0259] At this point, the product with the highest conversion value is identified, and media placements and time-sharing rotation periods are recommended for that product.

[0260] Furthermore, it should be noted that in specific examples, the distribution of users' timed exits at station gate groups is obtained through user itinerary tables and gate information tables. When users exit by scanning QR codes using the travel application, the distribution of users' timed exits at station gate groups describes the number of times a user scans their code at each gate during a time-sharing period, thus obtaining the user's gate group scan table.

[0261] In other words, the user's itinerary is linked to the gate code and the gate information table to obtain the user's gate group swipe code table.

[0262] Based on the user gate group's code-swiping table, the distribution of user code swipes in this group of gates can be obtained, and the distribution of user code swipes in this group of gates will indicate the user's exit direction.

[0263] By combining the distance relationships between the entrances and exits of the turnstile group as described in the table of distance relationships between the entrances and exits, we can determine the exits that the turnstile group points in different directions, and thus determine the user's frequently used exits and obtain the user's recently frequently used exit table.

[0264] For each exit, the media information bits mapped to that exit were recorded to construct and maintain a media exit relationship table.

[0265] By mapping the exits to media information locations, we can determine which media information locations a user will pass through when exiting a station. In other words, all users who pass through a media information location during a time-sharing broadcast period are all users corresponding to that media information location, thus forming the target user group.

[0266] For each time-sharing period, user profiles of all users corresponding to the media information position are assigned to the media information position, further enriching its tags and obtaining a media profile table.

[0267] At this point, the most suitable media information positions and time slots for product information can be determined based on the media profile table, ultimately realizing marketing recommendations for media information positions.

[0268] It should be further explained that all users corresponding to the media information position can be determined by statistically analyzing the user's preferred exit direction in the user gate group's code-swiping table, and then using this to determine all users corresponding to the media information position.

[0269] In addition, if a user's trips in the user itinerary are few, for example, if the number of times a user exits the station gate group is less than a set threshold, it will be necessary to use a binary classification model to determine the user's preferred exit direction by analyzing the distribution of the number of times the user exits the station gate group at each gate and the user profile, and then determine the exit gate to which the user belongs in that exit direction.

[0270] The probability of a user belonging to each exit is obtained by using a pre-trained binary classification model, and then the exit to which the user belongs is finally determined by the probability. For example, the pre-trained model can be a binary classification model, which uses the distribution of the number of times a user exits from each gate in the station gate group and the user profile as features. Under these features, the direction in which the user exits is labeled with the corresponding direction value of the feature, and each direction value uniquely corresponds to an exit that can be accessed from the gate group.

[0271] For example, if a user with the feature mapping exits the gate group and moves to the left, the labeled direction value is 0; conversely, if a user with the feature mapping exits the gate group and moves to the right, the labeled direction value is 1.

[0272] Based on this, a training dataset is constructed.

[0273] Wherein, each training dataset D train In a single record This is a characteristic of the record, which describes the user status of the corresponding user.

[0274] Specifically, This includes the distribution of the number of times a user exits the station through each gate in a given time period and the user profile, and the user profile can be represented by multiple user tags. It is the characteristic of user i in gate group j, R m It is the m-dimensional feature of user i.

[0275] For example, x ij ∈R m = [Date(YYYYMM), User i's User Tag 1, User i's User Tag 2, ..., User i's User Tag n] label The number of times user i exits through gate 1 in gate group j, the number of times user i exits through gate 2 in gate group j, ..., the number of times user i exits through gate n in gate group j zj [Number of exits], date (YYYYMM) represents the month, such as 202312, which represents December 2023. The date feature indicates the time range corresponding to other features. User tag n label This refers to the nth term of user i. label n user tags r Represents the number of users, n i n represents the number of turnstiles that user i faces when exiting the station during their trip. label y represents the number of user tags. ij ∈{0,1} represents the direction value of the feature annotation in a single record of the corresponding training dataset. ij ∈{0,1} represents the final exit direction for user i under the corresponding feature (especially in terms of its actual number of exits).

[0276] That is, y in a single record of the training dataset ij This refers to the direction value of the label. For example, if user i exits through gate group j and turns left, then y... ij =0, if to the right, then y ij =1.

[0277] The model is trained using the constructed training dataset to obtain a model that can accurately predict the exit to which a user belongs, providing users with a precise judgment of their exit location without the need for location acquisition, i.e., no longer limited by GPS positioning and location permission acquisition.

[0278] Furthermore, the training for exit prediction will be conducted on turnstile groups that correspond only to a single exit. A training dataset will be constructed for each turnstile group corresponding to a single exit. Features will be constructed based on all records related to that turnstile group, and the constructed features will be labeled with directional values ​​according to the unique orientation of that turnstile group.

[0279] Specifically, the number of times a user exits each gate in the gate group at the site is obtained through travel data and gate information. That is, the user gate group exit data of user-site-gate group-gate-exit count is obtained. It can be in tabular form, that is, a table maintained according to fields such as user, site, gate group, gate, and exit count. The user gate group exit data corresponding only to a single exit is obtained from the table. The number of times a user exits each gate in the gate group at the site is obtained from the user gate group exit data corresponding only to a single exit, and features are constructed based on this.

[0280] In addition, corresponding user profiles will be extracted for feature construction, and the constructed features will be labeled with directional values ​​to finally form a training dataset.

[0281] The constructed training dataset shields the influence of other exits, considering only the prediction of the user's exit in the feature dimension. This ensures that subsequent predictions are not affected by other factors, greatly improving the accuracy and reliability of the predictions.

[0282] For example, the distance relationship between the entrance and exit of the turnstile group will also be constructed. The distance relationship between the entrance and exit of the turnstile group is used to indicate the distance of each turnstile in the turnstile group in a station relative to the exit in each direction. So that after predicting the most likely direction of the user after exiting the turnstile, the user can be directly mapped to the nearest exit according to the distance relationship between the entrance and exit of the turnstile group. This exit is the exit to which the user belongs as predicted.

[0283] As mentioned earlier, the model trained using the constructed training dataset can be a binary classification model. This binary classification model will then be used to predict the user's assigned exit. Specifically, the binary classification model predicts the most likely direction a user will take after exiting the station's turnstiles based on the number of times the user exits each turnstile and their user profile. The exit mapped from this direction is then used as the user's assigned exit.

[0284] To further explain, a predictive dataset is constructed based on the number of times a user exits from each gate in the site's turnstile group and the user profile. Unlike the training dataset, the prediction dataset is not limited to the gate group corresponding to a unique exit. In other words, regardless of whether the gate group corresponds to a unique exit, all of the user's data will be used to construct the prediction dataset, and then the exit to which the user belongs will be predicted based on the features of the constructed prediction dataset.

[0285] Prediction dataset D predict In the middle, each record This represents the characteristics of user i in gate group j, whose exit point has not yet been determined. As mentioned earlier, these characteristics describe the number of times user i exits each gate in gate group j and the user's tag. The list of included characteristics is similar to... similar.

[0286] Based on the obtained binary classification model, the prediction dataset D will be... predict Each record By using a binary classification model to make predictions one by one, we can obtain the probability that user i's features in gate group j are biased in one direction, that is, the probability that it belongs to the exit corresponding to gate group j.

[0287] For example, if the turnstile group j corresponds to two exits. In other words, after exiting the station through turnstile group j, user i has two directions, each corresponding to an exit. The probability that user i will favor the other direction, i.e., belong to the other exit, is...

[0288] Furthermore, if there are more than two exits in one direction of gate group j, that is, the exits in that direction are not unique, then when the binary classification model predicts that user i is inclined to the direction with more than two exits after exiting gate group j, the exit that user i is most likely to pass through will be determined based on the flow of each exit. This exit is the exit to which user i belongs.

[0289] Specifically, each exit in one direction has its own traffic flow, which can be sensed by passenger flow counters installed at each exit. The traffic flow corresponding to each exit is used to determine the traffic proportion of each exit, and then the exit to which user i belongs is determined based on the traffic proportion.

[0290] Therefore, after determining the exit gate to which a user belongs, the user's travel trajectory is mapped to the exit gate to which the user belongs, and the user's itinerary during the trip will also be mapped to this exit gate.

[0291] Based on the exit where the user belongs, the target user group is determined for the media information positions around the exit. Then, the user profile of the target user group is used as the media profile to estimate the expected click value and conversion value of each media information position and its time-sharing rotation period, and obtain the media product conversion value table, which is ultimately used to execute the recommendation of media information positions.

[0292] It should be further explained that the click-through rate prediction recommendation algorithm, i.e. the click-through rate prediction model, is trained based on online product exposure and click data from platforms such as travel applications. The training dataset and product features are obtained from the constructed product tag table, which will not be elaborated here.

[0293] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the method according to the embodiments of this application.

[0294] In an exemplary embodiment of this application, a computer program medium is also provided, on which computer-readable instructions are stored, which, when executed by a computer's processor, cause the computer to perform the methods described in the above method embodiments.

[0295] According to one embodiment of this application, a program product for implementing the methods in the above-described method embodiments is also provided. This product may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of this invention is not limited thereto. In this document, a readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0296] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0297] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0298] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0299] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0300] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0301] Furthermore, although the steps of the method in this application are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0302] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, mobile terminal, or network device, etc.) to execute the method according to the embodiments of this application.

[0303] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the appended claims.

Claims

1. A time-based recommendation method for media information slots, characterized in that, The method includes: Corresponding to the split time-sharing carousel period, the user's travel data is obtained, and the travel data matches the time-sharing carousel period in time; Based on the user's travel data during the time-sharing period, the station to which the user's travel belongs during the time-sharing period is determined. Specifically, this includes: obtaining the number of times the user exits at each gate of the subway station gate group during the time-sharing period based on the user's exit station and exit gate information; determining the exit gate to which the user belongs during the time-sharing period based on the distribution of the number of times the user exits at each gate of the subway station gate group during the time-sharing period, where the exit gate is the exit that can be accessed by the gate group; Users belonging to the site during the time-sharing carousel period are aggregated to form the target user group for the media information positions mapped by the site to display products during the time-sharing carousel period; To estimate the expected clicks of products placed on media information slots for the target user group, and to obtain the expected click value of the products on the media information slots during the time-sharing carousel period, the method specifically includes: for each product, estimating the click-through rate of each user for the product within the target user group during the time-sharing carousel period; and calculating the weighted sum of the estimated click-through rates of users for the products based on the number of times users access the media information during the time-sharing carousel period, to obtain the expected click value of the media information slots during the time-sharing carousel period. Based on the expected click value of the product in each media information position corresponding to each time-sharing carousel period, media information positions with high expected click value and time-sharing carousel periods in the media information positions are recommended for the product.

2. The method according to claim 1, characterized in that, For each product, the step of estimating the click-through rate of each user for that product within the target user group during the time-sharing carousel segment of the media information location includes: For the target user group of the media information position during the time-sharing rotation period, user-product pairs are constructed by using the user profile of each user and the product characteristics corresponding to the product; By predicting the click-through rate of the user's product pair, the click-through rate of the media information position for each user in the target user group during the time-sharing carousel period is obtained.

3. The method according to claim 2, characterized in that, The user profile corresponds to the user's own travel data, as well as a combination of the user's own travel data and supplementary data from third parties.

4. The method according to claim 1, characterized in that, The step of recommending media information positions with high click-through expectations and time-sharing periods for the product based on the click-through expectations of the product in each media information position corresponding to each time-sharing period includes: By comparing the expected click value of the product in each time slot of each media information position, the media information position with the highest expected click value and the time slot of the product in the media information position are recommended to the product.

5. The method according to claim 1, characterized in that, The media information slot during the time-sharing rotation period is competed for by two or more products; The method further includes recommending media information positions with high click-expectation values ​​for the product based on the click-expectation values ​​of the product in each media information position corresponding to each time-sharing carousel period, and recommending media information positions with high click-expectation values ​​after the time-sharing carousel period of the media information positions. For two or more products competing for the media information slot, the click value of each product is obtained; The conversion value of each product is calculated using the click value and the expected click value of the product in the media information position during the competing time-sharing carousel period. The time-sharing rotation period of the media information slot will be recommended for the placement of high-conversion-value products.

6. A time-sharing recommendation system for media information bits, the system comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method as described in any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 5.

8. A computer program product, comprising a computer program, characterized in that, The computer program is executed by a processor to implement the steps of the method as described in any one of claims 1 to 5.

Images