Method, device, equipment and storage medium for determining target similarity

Abstract

The present invention discloses a method, apparatus, device and storage medium for determining target similarity. The method includes: responding to the target recommendation request of the current user, traversing each behavioral target that the current user has completed clicking, determining the associated targets with which there are common user clicks with the behavioral target, and the number of common user clicks between the behavioral target and the associated targets; calculating the similarity between the behavioral target and the associated targets based on the number of common user clicks between each behavioral target and its associated targets, as well as the number of independent user clicks of the behavioral target and the associated targets. The technical solution of the present invention comprehensively analyzes the target click behavior of the current user in real time, avoids the problem of omission of target click behaviors referenced in target similarity analysis due to an excessively long offline statistical cycle, calculates the similarity between each behavioral target and each associated target in real time, ensures the timeliness of target similarity calculation, and improves the accuracy of target recommendation.

Description

Technical Field

[0001] Embodiments of the present invention relate to the technical field of information recommendation, and in particular to a method, apparatus, device, and storage medium for determining target similarity. Background Art

[0002] With the rapid development of internet technology, various information recommendation scenarios are becoming increasingly widespread, such as recommendations for user-interested news, short videos, voice chat rooms, and live broadcast rooms. Common collaborative filtering algorithms used in information recommendation scenarios are mainly divided into item-based collaborative filtering (ItemCF), which calculates a list of the most similar items for each item to recommend items similar to the user's favorite items; and user-based collaborative filtering (UserCF), which recommends items that have been liked by users similar to the current user (but the current user has not yet operated on them).

[0003] Currently, the two aforementioned methods are commonly used to combine offline statistics of user behavior lists for each item, and then calculate a list of similar items for each item. This allows subsequent online information recommendations to be made based on the offline similar item lists. However, since offline statistics of similar item lists for each item require calculating the similarities between a large number of items, this incurs a significant computational overhead. This results in each offline statistics being extremely time-consuming, making it impossible to timely update the offline statistics of the item similarity lists. Therefore, when recommending targets with a short lifespan, since offline statistics were collected earlier and users have clicked on new targets, the highly similar targets in the offline similar item lists may no longer be available for recommendation, making it impossible to guarantee the timeliness of target similarity calculations. Summary of the Invention

[0004] The embodiments of the present invention provide a method, apparatus, device and storage medium for determining target similarity, which calculates the similarity between targets in real time during online recommendation, thereby ensuring the timeliness of target similarity calculation and improving the accuracy of target recommendation.

[0005] In a first aspect, an embodiment of the present invention provides a method for determining target similarity, the method comprising:

[0006] In response to a target recommendation request from a current user, traverse each behavioral target that the current user has clicked, determine associated targets that have common user clicks with the behavioral target, and determine the number of common user clicks between the behavioral target and the associated targets;

[0007] The similarity between each behavior target and its associated targets is calculated based on the number of common user clicks between each behavior target and its associated targets, and the number of independent user clicks between the behavior target and the associated targets.

[0008] In a second aspect, an embodiment of the present invention provides a device for determining target similarity, the device comprising:

[0009] an associated target determination module, configured to, in response to a target recommendation request of a current user, traverse each behavioral target that the current user has clicked on, determine associated targets that have a common user click with the behavioral target, and determine the number of common user clicks between the behavioral target and the associated target;

[0010] The similarity calculation module is used to calculate the similarity between each behavior target and its associated targets based on the number of common user clicks between each behavior target and its associated targets, as well as the number of independent user clicks between the behavior target and the associated targets.

[0011] In a third aspect, an embodiment of the present invention provides an electronic device, the electronic device comprising:

[0012] one or more processors;

[0013] a storage device for storing one or more programs;

[0014] When the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining target similarity described in any embodiment of the present invention.

[0015] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for determining target similarity described in any embodiment of the present invention.

[0016] The embodiments of the present invention provide a method, apparatus, device and storage medium for determining target similarity. After receiving a target recommendation request from the current user, the method traverses each behavioral target that the current user has clicked on in real time to determine each associated target that has common user clicks with the behavioral target, as well as the number of common user clicks between the behavioral target and each associated target, thereby comprehensively analyzing the current user's target click behavior in real time to avoid the problem of omission of target click behaviors referenced in target similarity analysis due to a long offline statistical period. Then, during online recommendation, the similarity between each behavioral target and each associated target is calculated in real time based on the number of common user clicks between each behavioral target and each associated target of the behavioral target, as well as the number of independent user clicks between each behavioral target and each associated target, thereby ensuring the timeliness of the target similarity calculation, solving the problem of invalidation during offline statistical target similarity list, and improving the accuracy of target recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings:

[0018] Figure 1A A flowchart of a method for determining target similarity provided in Example 1 of the present invention;

[0019] Figure 1B A schematic diagram illustrating the principle of a process for determining target similarity provided in the first embodiment of the present invention;

[0020] Figure 2A A flowchart of a method for determining target similarity provided in the second embodiment of the present invention;

[0021] Figure 2B A schematic diagram illustrating the principle of a process for determining target similarity provided in the second embodiment of the present invention;

[0022] Figure 3 A schematic diagram of the structure of a device for determining target similarity provided in the third embodiment of the present invention;

[0023] Figure 4 This is a structural diagram of an electronic device provided in Example 4 of the present invention. DETAILED DESCRIPTION

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all of the structures. Furthermore, the embodiments of the present invention and the features of the embodiments may be combined with one another unless otherwise specified.

[0025] Example 1

[0026] Figure 1A This is a flowchart of a method for determining target similarity provided in the first embodiment of the present invention. This embodiment is applicable to any target recommendation scenario with a short life cycle, such as a voice room or a live broadcast room. The method for determining target similarity provided in this embodiment can be executed by a target similarity determination device provided in an embodiment of the present invention. The device can be implemented in software and / or hardware and integrated into an electronic device that executes this method. The device can be a background server of an application where the target, such as a voice room or a live broadcast room, is located.

[0027] Specifically, refer to Figure 1A , the method may include the following steps:

[0028] S110, in response to the target recommendation request of the current user, traverse each behavior target that the current user has completed clicking, determine the associated targets that have common user clicks with the behavior target, and the number of common user clicks between the behavior target and the associated targets.

[0029] Among them, in this embodiment, the targets of user recommendation requests are mainly objects with certain survival cycle requirements. For example, after the user logs in to the live broadcast application, the user will be recommended voice rooms or live broadcast rooms that are of interest and in broadcasting status based on the user's historical viewing behavior.

[0030] At this time, considering that when recommending targets with survival cycle requirements, if the similar target list of each target is counted offline based on user behavior at a long interval, then the user may have new target click behavior during the interval period, so that the targets with higher similarity in the target similarity list counted offline each time the target is recommended may have become invalid, and the timeliness of the target similarity calculation cannot be guaranteed.

[0031] Therefore, for target recommendation scenarios with certain survival cycle requirements, this embodiment provides a new target similarity calculation method based on real-time collaborative filtering. According to the target click behavior of each user, the user's behavior target and the associated targets of the behavior target are updated in a timely manner. When the user requests a recommended target, the corresponding target similarity is calculated online through the timely updated user behavior target and the associated targets of the behavior target, thereby ensuring the timeliness of the target similarity calculation.

[0032] Optionally, in order to accurately analyze the user's interest level in each recommended target, this embodiment will continuously analyze the historical click behavior of each user on the recommended targets in real time or at very short intervals, so as to timely update the various behavioral targets that each user has clicked on, and ensure the integrity of the user's behavioral targets, where the behavioral target can be a voice room or live broadcast room that the user has clicked to watch.

[0033] Moreover, in order to ensure the convenience and efficiency of target similarity, this embodiment will update each behavioral target that the user has clicked on each time, taking into account that there is usually a certain similarity between multiple targets that the same user is interested in and clicks on. Therefore, it will also analyze the common behavioral characteristics between various targets, and pre-screen various related targets that have been clicked by the same user for each behavioral target from all existing targets. At this time, each behavioral target and the related targets of the behavioral target will be clicked by the same user, so when the user's various behavioral targets are updated in real time or at very short intervals, the related targets of each behavioral target are also updated accordingly. Subsequently, the similarity between each behavioral target and the related targets can be directly analyzed, thereby ensuring the convenience and efficiency of target similarity calculation.

[0034] In this embodiment, when the current user refreshes the target, a corresponding target recommendation request will be generated, such as when the user pulls down or refreshes the corresponding live broadcast room. At this time, after receiving the target recommendation request of the current user, such as Figure 1B As shown, it will find out in real time the various behavioral targets that the current user has clicked, and then traverse each behavioral target to determine the various associated targets that have been clicked by common users for the behavioral target. That is, one behavioral target can correspond to multiple associated targets, and determine the number of common user clicks between each behavioral target and each associated target of the behavioral target. Since the more similar two targets are, the more common users will click on them, the number of common user clicks can roughly represent the similarity between the two targets.

[0035] It should be noted that since each behavioral target that the current user has completed clicking and the associated targets of each behavioral target will be continuously updated regularly or at very small intervals, it is possible to ensure the comprehensiveness of the behavioral targets and associated targets referenced when calculating target similarity, avoiding the problem of omissions in target click behaviors referenced in target similarity analysis due to a long offline statistical cycle, thereby ensuring the accuracy of target similarity calculation.

[0036] S120 , calculating the similarity between each behavior target and its associated targets based on the number of common user clicks between each behavior target and its associated targets, and the number of independent user clicks between the behavior target and the associated targets.

[0037] Optionally, after determining the various behavioral targets that the current user has clicked on, as well as the associated targets of each behavioral target, this embodiment will obtain the number of common user clicks between each behavioral target and the associated targets of the behavioral target. At this time, the number of common user clicks can roughly represent the degree of similarity between the behavioral target and the associated targets. Therefore, in order to further ensure the accuracy of the target similarity, this embodiment, after determining the number of common user clicks between the behavioral target and the associated targets of the behavioral target for each behavioral target, will also analyze the behavior of the behavioral target and each of its associated targets being clicked by different users, thereby determining the number of independent user clicks for each behavioral target and each of its associated targets. Then, for each behavioral target and each of its associated targets, the number of common user clicks between the behavioral target and each of its associated targets is analyzed separately, and the proportion of the combined number of independent user clicks of the behavioral target and each of its associated targets under the comprehensive number of independent user clicks is used to obtain the similarity between the behavioral target and each of its associated targets. At this time, when recommending targets online, the similarity between each behavioral target and each related target is calculated in real time to ensure the timeliness of the target similarity calculation, solve the problem of invalidation when calculating the target similarity list offline, and improve the accuracy of target recommendation.

[0038] For example, for each behavior target ROOM_i and each associated target ROOM_j of the behavior target ROOM_i, the similarity between the behavior target ROOM_i and the associated target ROOM_j can be calculated using the following formula: Among them, CNT is the number of common user clicks between the behavior target ROOM_i and the associated target ROOM_j, n i The number of unique user clicks for the behavioral target ROOM_i, n j The number of unique user clicks on the associated target ROOM_j.

[0039] In addition, since the recommended targets in this embodiment all have certain survival cycle requirements, and when calculating the similarity between each behavioral target and each associated target, the corresponding associated targets will be recommended to the current user according to the similarity level, in order to ensure that the recommended associated targets are clickable and viewable by the current user, this embodiment will further determine the survival status of each associated target of each behavioral target after determining the associated targets that have common users clicking on each behavioral target, so that subsequently only the similarity calculation method provided above needs to be used to calculate the similarity between each behavioral target and each surviving associated target of the behavioral target, without calculating the similarity between each behavioral target and the non-surviving associated targets of the behavioral target, thereby ensuring the efficiency of the target similarity calculation.

[0040] Then, after calculating the similarity between each behavior target and each surviving associated target of the behavior target, each surviving associated target can be sorted according to the similarity between each behavior target and each surviving associated target of the behavior target, and then the corresponding surviving associated targets can be recommended to the current user in order to ensure that the recommended targets are all in a surviving state, thereby improving the accuracy of target recommendation.

[0041] The technical solution provided in this embodiment, after receiving the target recommendation request of the current user, will traverse in real time each behavioral target that the current user has completed clicking, to determine the various associated targets that have common user clicks with the behavioral target, and the number of common user clicks between the behavioral target and each associated target, so as to comprehensively analyze the current user's target click behavior in real time, and avoid the problem of omission of target click behaviors referenced by target similarity analysis due to too long offline statistical cycle. Then, when making online recommendations, the similarity between each behavioral target and each associated target is calculated in real time based on the number of common user clicks between each behavioral target and each associated target of the behavioral target, as well as the number of independent user clicks between each behavioral target and each associated target, thereby ensuring the timeliness of the target similarity calculation, solving the problem of invalidation of offline statistical target similarity lists, and improving the accuracy of target recommendations.

[0042] Example 2

[0043] Figure 2A This is a flowchart of a method for determining target similarity provided in the second embodiment of the present invention. Figure 2B The schematic diagram of the principle of the target similarity determination process provided in the second embodiment of the present invention. This embodiment is optimized based on the above embodiment. Specifically, in order to ensure the integrity and unity of the behavioral target that the user has clicked and the associated targets of each behavioral target, Figure 2B As shown, this embodiment maintains a behavior target queue for each user, which records each behavior target that user has clicked. Furthermore, for each behavior target, an associated target queue is maintained, which records each associated target that has been clicked by a common user with the behavior target, as well as the number of common user clicks between each associated target and the behavior target. This embodiment primarily explains the specific process of traversing the current user's behavior targets and the specific process of updating user behavior targets and associated targets.

[0044] Optional, such as Figure 2A As shown, this embodiment may include the following steps:

[0045] S210 , regularly updating the current user's behavior target queue and the associated target queue of each behavior target in the behavior target queue based on the current user's near-line click behavior toward the recommended target.

[0046] Since the recommended targets in this embodiment have a certain life cycle, users may continuously click on multiple targets within a short time interval. Therefore, in order to ensure the integrity of the behavioral targets that users have clicked, this embodiment will use near-line behavioral analysis to regularly update the behavioral targets of each user.

[0047] In this embodiment, the user's near-line click behavior towards the recommended target is searched once every short time interval (such as every 2 minutes). The near-line click behavior may refer to the click behavior performed by the user on each recommended target during the current interval. Exemplarily, the near-line click behavior in this embodiment may be the user's click-to-enter-a-room behavior event during the current interval. Then, by analyzing the various behavioral targets pointed to by each user's near-line click behavior, the user's behavioral target queue is updated, so that the current user's behavioral target queue can fully record the various behavioral targets that the current user has clicked. At the same time, after the new behavioral target is updated in the current user's behavioral target queue, the associated target queue of each behavioral target will be updated accordingly based on the situation in which each behavioral target and the associated target of the behavioral target are clicked by a common user, so that the associated target queue of each behavioral target can accurately record the various associated targets of the behavioral target and the number of common user clicks on the behavioral target and each associated target by the same user.

[0048] As an optional solution of this embodiment, this embodiment updates the current user's behavior target queue and the associated target queue of each behavior target in the following manner.

[0049] 1) Update the current user's behavior target queue: Regularly determine the current user's recent behavior target in the current time period based on the current user's recent click behavior towards the recommended target; update the current user's historical behavior target queue according to the recent behavior target to obtain an updated behavior target queue.

[0050] In this embodiment, the current user's recent click behaviors on the recommended targets are periodically obtained, and the behavioral targets triggered by each recent click behavior are analyzed to obtain the recent behavioral targets clicked by the current user within the current time period. A historical behavioral target queue consisting of the behavioral targets clicked by the current user within the historical time period is then retrieved, and it is determined whether the current user's recent behavioral targets have been cached in the historical behavioral target queue. This is used to analyze whether each recent behavioral target has been clicked for the first time by the current user, and the behavioral target that was first clicked by the current user among the recent behavioral targets is added to the historical behavioral target queue to obtain an updated behavioral target queue. This updated behavioral target queue then comprehensively records the behavioral targets that the current user has clicked.

[0051] For example, a flink process is used to search for the current user's recent click behavior in the current interval every 2 minutes, and cache it in the kafka message queue to consume each recent click behavior, thereby analyzing the current user's various recent behavior targets in the current period and forming the corresponding recent room entry list new_lst. Then, the historical behavior target queue old_lst of the current user USER is searched in the remote dictionary service (Remote Dictionary Server, Redis) database using the KEY1_{USER} method, and it is determined whether each recent behavior target (ROOM_i) in the recent room entry list new_lst belongs to the historical behavior target queue old_lst, thereby adding each behavior target that the current user clicked for the first time in the recent behavior target to the new recent room entry list new_lst_v2, and updating each behavior target in the new recent room entry list new_lst_v2 to the historical behavior target queue old_lst, and obtaining the updated behavior target queue, recorded as KEY1_{USER}:list <room>and save it to the Redis database.

[0052] 2) Update of the associated target queue of each behavior target: For each first behavior target that the current user clicks for the first time among the near-line behavior targets, the target pair consisting of the first behavior target and each second behavior target in the behavior target queue except the first behavior target will be added to the preset cross queue; for each target pair in the cross queue, the associated target queue of the first behavior target in the target pair is obtained, and the associated target queue is updated according to the second behavior target in the target pair.

[0053] Specifically, if a near-line behavior target of the current user is in the historical behavior target queue, it means that the near-line behavior target is not clicked for the first time by the current user, that is, the behavior of the near-line behavior target and each behavior target in the historical behavior target queue being clicked together by the current user has been counted in the number of common user clicks. There is no need to count it again this time. It is only necessary to count the near-line behavior target clicked for the first time by the current user and each behavior target in the historical behavior target queue being clicked together by the current user, so as to avoid repeated counting of the number of common user clicks between the same behavior target and the associated targets.

[0054] Therefore, in order to ensure the accuracy of the number of common user clicks on the behavior target and each associated target recorded in the associated target queue of each behavior target, this embodiment takes the near-line behavior target clicked for the first time by the current user as the first behavior target. At this time, for each first behavior target, the first behavior target and each other behavior target in the updated behavior target queue except the first behavior target will be respectively formed into a target pair. At this time, each other behavior target in the behavior target queue except the first behavior target will be taken as the second behavior target in this embodiment, that is, each first behavior target and each behavior target in the historical behavior target queue will be formed into a target pair in pairs, and each first behavior target will also be formed into a target pair in pairs, and then each formed target pair will be added to the preset cross queue.

[0055] Then, since the first behavior target in each target pair in the cross queue is clicked for the first time by the current user, the number of common user clicks between each first behavior target and the associated target of the first behavior target needs to be updated. At this time, for each target pair in the cross queue, first obtain the associated target queue of the first behavior target in the target pair to determine the various associated targets of the first behavior target. Then, determine whether the second behavior target in the target pair is the associated target of the first behavior target; if so, it means that the first behavior target and the second behavior target have been clicked together by other users. After being clicked together by the current user, the number of common user clicks between the associated target represented by the second behavior target and the first behavior target in the associated target queue of the first behavior target is updated accordingly. However, if the second behavior target within the target pair is not the associated target of the first behavior target, it means that the first behavior target and the second behavior target have not been clicked together by any user before, and this time they are clicked together by the current user is the first time they are clicked together by the same user. Therefore, it is necessary to add the second behavior target as the associated target of the first behavior target in the associated target queue of the first behavior target, and correspondingly increase the record of the number of common user clicks between the first behavior target and the second behavior target. At this time, the number of common user clicks is 1.

[0056] For example, the current user's near-line behavior target is added to the new near-line room entry list new_lst_v2. Each first behavior target is recorded as ROOM_i. The first behavior target ROOM_i is combined with each behavior target ROOM_j in the historical behavior target queue old_lst in sequence and added to the cross queue pair_list. At the same time, each first behavior target in the new near-line room entry list new_lst_v2 is combined in pairs and also added to the cross queue pair_list to obtain the cross queue pair_list. <pair<ROOM_i,ROOM_j> >, where ROOM_j represents the second row target. Then, traverse each target pair in the cross queue<ROOM_i,ROOM_j> , use KEY2_{ROOM_i} to determine the associated target queue list for each first line of target ROOM_i <pair<ROOM_s,CNT> >, where ROOM_s represents the associated targets of the first row target ROOM_i, and CNT represents the number of common user clicks between the first row target ROOM_i and the associated target ROOM_s.<ROOM_i,ROOM_j> The second line in the target does not belong to the associated target ROOM_s in the associated target queue, so the pair<ROOM_j,1> Add to the associated target queue list of the first line target <pair<ROOM_s,CNT> >Inside; and if the target pair<ROOM_i,ROOM_j> The second line in the target belongs to the associated target ROOM_s in the associated target queue, so the associated target queue list <pair<ROOM_s,CNT> >The second line represents the associated target pair<ROOM_j,CNT> Corresponding update is pair<ROOM_j,CNT+1> , thereby realizing the update of the associated target queue of each behavior target, and recording the associated target queue of each first behavior target after update as KEY2_{ROOM}:list <pair<ROOM,CNT> > and save it to the Redis database.

[0057] According to the above method of regular updates based on the recent click behavior of each user, the behavior target queue KEY1_{USER}: list of each user can be accurately obtained. <room>And the associated target queue KEY2_{ROOM} for each line: list <pair<ROOM,CNT> >.

[0058] S220 , in response to the target recommendation request of the current user, obtaining a behavior target queue of the current user oriented towards the recommended target.

[0059] When receiving the target recommendation request of the current user, it is necessary to analyze the various behavioral targets that the current user has completed clicking. At this time, the KEY1_{USER} method is used to find the current user's behavioral target queue from the above-mentioned behavioral target queue that is regularly updated based on the user's recent click behavior. The behavioral target queue accurately records the various behavioral targets that the current user has completed clicking.

[0060] S230, traverse each behavior target in the behavior target queue, determine the associated target queue of the behavior target, and record the associated targets that have common user clicks with the behavior target and the number of common user clicks between the behavior target and the associated targets.

[0061] After obtaining the current user's behavior target queue for the recommended target, each behavior target ROOM_i in the behavior target queue can be traversed, and then the associated target queue of each behavior target ROOM_i that the current user has clicked on can be found from the associated target queues of each behavior target that is regularly updated using the KEY2_{ROOM_i} method. The associated target queue accurately records each associated target ROOM_j that has a common user click with the behavior target ROOM_i, as well as the number of common user clicks between the behavior target ROOM_i and the associated target ROOM_j. Subsequently, the number of common user clicks between the behavior target ROOM_i and the associated target ROOM_j, as well as the number of independent user clicks between the behavior target ROOM_i and the associated target ROOM_j, can be directly used to calculate the similarity between the behavior target ROOM_i and the associated target ROOM_j. Among them, the number of independent user clicks between the behavior target ROOM_i and the associated target ROOM_j can be represented by the list length of the associated target queues of the behavior target ROOM_i and the associated target ROOM_j.

[0062] S240 , calculating the similarity between the behavior target and the associated targets based on the number of common user clicks between each behavior target and the associated targets, and the number of independent user clicks between the behavior target and the associated targets.

[0063] The technical solution provided in this embodiment regularly updates each user's behavioral target queue and each behavioral target's associated target queue according to each user's near-line click behavior towards the recommended target, ensuring the accuracy and comprehensiveness of the behavioral targets and associated targets referenced when calculating target similarity. Then, when making online recommendations, the similarity between each behavioral target and each associated target is calculated in real time based on the number of common user clicks between each behavioral target in the current user's behavioral target queue and each associated target in the associated target queue of the behavioral target, as well as the number of independent user clicks between each behavioral target and each associated target, thereby ensuring the timeliness of the target similarity calculation, solving the problem of invalidation of offline statistical target similarity lists, and improving the accuracy of target recommendations.

[0064] Example 3

[0065] Figure 3 This is a schematic diagram of a device for determining target similarity according to the third embodiment of the present invention. Specifically, Figure 3 As shown, the device may include:

[0066] The associated target determination module 310 is configured to, in response to a target recommendation request of the current user, traverse each behavioral target that the current user has clicked on, determine associated targets that have a common user click with the behavioral target, and determine the number of common user clicks between the behavioral target and the associated targets;

[0067] The similarity calculation module 320 is used to calculate the similarity between each behavior target and its associated targets based on the number of common user clicks between each behavior target and its associated targets, as well as the number of independent user clicks on each behavior target and its associated targets.

[0068] The technical solution provided in this embodiment, after receiving the target recommendation request of the current user, will traverse in real time each behavioral target that the current user has completed clicking, to determine the various associated targets that have common user clicks with the behavioral target, and the number of common user clicks between the behavioral target and each associated target, so as to comprehensively analyze the current user's target click behavior in real time, and avoid the problem of omission of target click behaviors referenced by target similarity analysis due to too long offline statistical cycle. Then, when making online recommendations, the similarity between each behavioral target and each associated target is calculated in real time based on the number of common user clicks between each behavioral target and each associated target of the behavioral target, as well as the number of independent user clicks between each behavioral target and each associated target, thereby ensuring the timeliness of the target similarity calculation, solving the problem of invalidation of offline statistical target similarity lists, and improving the accuracy of target recommendations.

[0069] The device for determining target similarity provided in this embodiment can be applied to the method for determining target similarity provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0070] Example 4

[0071] Figure 4 This is a structural diagram of an electronic device provided in the fourth embodiment of the present invention, such as Figure 4 As shown, the electronic device includes a processor 40, a storage device 41 and a communication device 42; the number of processors 40 in the electronic device can be one or more. Figure 4 In the figure, a processor 40 is used as an example; the processor 40, the storage device 41 and the communication device 42 in the electronic device can be connected by a bus or other means. Figure 4 The bus connection is taken as an example.

[0072] The electronic device provided in this embodiment can be used to execute the target similarity determination method provided in any of the above embodiments, and has corresponding functions and beneficial effects.

[0073] Example 5

[0074] The fifth embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for determining target similarity in any of the above embodiments can be implemented.

[0075] Of course, the computer executable instructions of a storage medium containing computer executable instructions provided by an embodiment of the present invention are not limited to the method operations described above, and can also execute related operations in the target similarity determination method provided by any embodiment of the present invention.

[0076] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general-purpose hardware, and of course it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment of the present invention.

[0077] It is worth noting that in the embodiment of the above-mentioned device for determining target similarity, the various units and modules included are only divided according to functional logic, but are not limited to the above-mentioned division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0078] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.< / room> < / room>

Claims

1. A method for determining target similarity, characterized in that: include: In response to a target recommendation request from a current user, traverse each behavioral target that the current user has clicked, determine associated targets that have common user clicks with the behavioral target, and determine the number of common user clicks between the behavioral target and the associated targets; Calculate the similarity between each behavioral target and its associated targets based on the number of common user clicks between the behavioral target and its associated targets, and the number of independent user clicks between the behavioral target and the associated targets; Regularly update the current user's behavior target queue and the associated target queue of each behavior target in the behavior target queue based on the current user's recent click behavior towards the recommended target. The recent click behavior refers to the click behavior performed by the user on each recommended target within the current interval period.

2. The method according to claim 1, characterized in that The traversing each behavior target that the current user has clicked on, determining associated targets that have common user clicks with the behavior target and the number of common user clicks between the behavior target and the associated targets, includes: Obtaining a behavior target queue of the current user for the recommended target, wherein the behavior target queue records each behavior target that the current user has clicked; Traverse each behavior target in the behavior target queue to determine the associated target queue of the behavior target, wherein the associated target queue records each associated target that has common user clicks with the behavior target and the number of common user clicks between the behavior target and the associated target.

3. The method according to claim 1, characterized in that The updating of the behavior target queue of the current user according to the recent click behavior of the current user towards the recommendation target on a regular basis includes: regularly determining a recent behavior target of the current user in a current time period based on the recent click behavior of the current user towards the recommended target; The historical behavior target queue of the current user is updated according to the near-line behavior target to obtain an updated behavior target queue.

4. The method according to claim 3, characterized in that Updating the associated target queue of each behavior target in the behavior target queue includes: For each first behavior target that is first clicked by the current user in the near-line behavior targets, add a target pair consisting of the first behavior target and each second behavior target in the behavior target queue other than the first behavior target to a preset cross queue; For each target pair in the cross queue, an associated target queue of the first row as the target in the target pair is obtained, and the associated target queue is updated according to the second row as the target in the target pair.

5. The method according to claim 1, wherein After determining the associated targets that have common user clicks with the behavior target and the number of common user clicks between the behavior target and the associated targets, the method further includes: The survival status of each associated target of each behavior target is determined to calculate the similarity between each behavior target and each surviving associated target of the behavior target.

6. The method according to claim 5, characterized in that After calculating the similarity between each behavior target and each existing associated target of the behavior target, it also includes: According to the similarity between each behavior target and each existing associated target of the behavior target, the corresponding existing associated targets are recommended to the current user.

7. A device for determining target similarity, characterized in that: The method for determining target similarity according to claim 1 comprises: an associated target determination module, configured to, in response to a target recommendation request of a current user, traverse each behavioral target that the current user has clicked on, determine associated targets that have a common user click with the behavioral target, and determine the number of common user clicks between the behavioral target and the associated target; The similarity calculation module is used to calculate the similarity between each behavior target and its associated targets based on the number of common user clicks between each behavior target and its associated targets, as well as the number of independent user clicks between the behavior target and the associated targets.

8. An electronic device, characterized in that: The electronic device comprises: one or more processors; a storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining target similarity as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for determining target similarity as claimed in any one of claims 1 to 6 is implemented.

Images