Object recommendation method and apparatus, electronic device, and storage medium

Abstract

This invention relates to an object recommendation method, apparatus, electronic device, and storage medium. The method includes: determining a first object recommendation sequence and determining the feature vector of each first object in the first object recommendation sequence, wherein the first objects are objects in a recommendation pool; determining a second object recommendation sequence and determining the feature vector of each second object in the second object recommendation sequence, wherein the second objects are objects recommended to a target user within a preset historical time period; determining a recommendation score for each first object based on the feature vectors of each first object and each second object; shuffling the first object recommendation sequence based on the recommendation scores of each first object to obtain a third object recommendation sequence; and recommending objects to the target user according to the third object recommendation sequence. This achieves the goal of distributing highly similar recommended objects during object recommendation, reducing user fatigue when browsing recommended objects, thereby improving the user experience.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of computer technology, and in particular to an object recommendation method, apparatus, electronic device and storage medium. Background Technology

[0002] With the development of internet technology, intelligent recommendation services have become widely used because they can quickly present users with recommended items that they may be interested in based on their preferences, habits, and browsing history. These recommended items can be products, information, or audio and video data. For example, when a user plays a video using video playback software, videos that the user may be interested in can be recommended based on the user's historical playback records; or when a user browses web pages, news that the user may be interested in can be recommended based on the user's browsing history.

[0003] However, since the intelligent recommendation service recommends objects to users based on data such as user preferences, habits, and browsing history, there may be a high degree of similarity between different recommended objects. In this case, if highly similar objects are recommended in a concentrated manner, it will increase user fatigue when browsing recommended objects.

[0004] To address this, existing technologies propose using similarity assessments based on the attributes of the recommended objects themselves, such as titles and content, to distribute highly similar recommendations and reduce user fatigue. However, judging similarity based solely on object attributes suffers from low accuracy due to a simplistic criterion. This fails to effectively distribute highly similar recommendations, resulting in a concentration of similar objects and increasing user fatigue, thus negatively impacting the user experience. Summary of the Invention

[0005] In view of this, in order to solve the technical problem that the existing technology cannot effectively distribute the recommendations of highly similar objects, resulting in a concentration of highly similar objects when recommending objects, which increases user fatigue when browsing objects and affects user experience, the embodiments of the present invention provide an object recommendation method, device, electronic device and storage medium.

[0006] In a first aspect, embodiments of the present invention provide an object recommendation method, the method comprising:

[0007] Determine a first object recommendation sequence and determine the feature vector of each first object in the first object recommendation sequence, wherein the first object is an object in the recommendation pool;

[0008] Determine the second object recommendation sequence and the feature vector of each second object in the second object recommendation sequence, wherein the second object is an object recommended to the target user within a preset historical time period;

[0009] The recommended score for each of the first objects is determined based on the feature vectors of each of the first objects and the feature vectors of each of the second objects.

[0010] Based on the recommendation scores of each of the first objects, the recommendation sequence of the first objects is shuffled to obtain the recommendation sequence of the third objects;

[0011] Recommend objects to the target user according to the third object recommendation sequence.

[0012] As one possible implementation, determining the recommendation score for each of the first objects based on the feature vectors of each of the first objects and the feature vectors of each of the second objects includes:

[0013] For each first object, the similarity between the feature vector of the first object and the feature vector of each second object is calculated to obtain the first similarity value of the first object corresponding to each second object;

[0014] Based on each of the first similarity values, a recommended score for the first object is determined.

[0015] As one possible implementation, determining the recommendation score of the first object based on each of the first similarity values ​​includes:

[0016] Determine the number of first similarity values ​​that satisfy the first preset condition among all the first similarity values;

[0017] Based on the quantity, the recommended score for the first object is determined.

[0018] As one possible implementation, determining the recommendation score of the first object based on each of the first similarity values ​​includes:

[0019] Determine the average value of each of the first similarity values;

[0020] Based on the average value, a recommended score for the first object is determined.

[0021] As one possible implementation, after determining the recommendation score of the first object, the method further includes:

[0022] The similarity between the feature vector of each first object and the feature vector of each other first object in the recommendation sequence of the first object is calculated to obtain the second similarity value of the first object to each other first object;

[0023] The recommendation score of the first object is adjusted based on the second similarity value, and the step of shuffling the recommendation sequence of the first object based on the recommendation score of each first object to obtain the recommendation sequence of the third object is performed based on the adjusted recommendation score.

[0024] As one possible implementation, after determining the recommendation score for each of the first objects, the method further includes:

[0025] Determine the historical recommendation time of each of the second objects in the second object recommendation sequence;

[0026] Based on the historical recommendation time, determine the weight value of each of the second objects;

[0027] The recommendation score of the first object is adjusted according to the weight value, and the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object is performed according to the adjusted recommendation score.

[0028] As one possible implementation, after the recommendation score for each of the first objects, the method further includes:

[0029] Determine the predicted click-through rate of the target user for the first object;

[0030] The recommended score of the first object is adjusted based on the predicted click-through rate, and the step of shuffling the recommended sequence of the first object based on the recommended score of each first object is performed to obtain the recommended sequence of the third object.

[0031] As one possible implementation, determining the feature vector of each first object in the first object recommendation sequence includes:

[0032] The feature vectors of each first object in the first object recommendation sequence are obtained from the preset storage service, which stores the feature vectors of each object in the recommendation pool.

[0033] As one possible implementation, the step of shuffling the recommendation sequence of the first objects based on their recommendation scores to obtain the recommendation sequence of the third objects includes:

[0034] In the first object recommendation sequence, delete the first object whose recommendation score does not meet the second preset condition to obtain the first object recommendation sequence after deletion.

[0035] For the first object recommendation sequence after deletion, the first objects are reordered according to their recommendation scores to obtain the third object recommendation sequence.

[0036] Secondly, embodiments of the present invention provide an object recommendation device, the device comprising:

[0037] The first determining module is used to determine the first object recommendation sequence and to determine the feature vector of each first object in the first object recommendation sequence, wherein the first object is an object in the recommendation pool;

[0038] The second determining module is used to determine the second object recommendation sequence and to determine the feature vector of each second object in the second object recommendation sequence, wherein the second object is an object recommended to the target user within a preset historical time period;

[0039] The third determining module is used to determine the recommended score of each first object based on the feature vector of each first object and the feature vector of each second object;

[0040] The shuffling module is used to shuffle the recommendation sequence of the first object according to the recommendation score of each of the first objects to obtain the recommendation sequence of the third object;

[0041] The recommendation module is used to recommend objects to the target user according to the third object recommendation sequence.

[0042] As one possible implementation, the third determining module includes:

[0043] The similarity calculation submodule is used to calculate the similarity between the feature vector of the first object and the feature vector of each second object for each first object, so as to obtain the first similarity value of the first object corresponding to each second object.

[0044] The determination submodule is used to determine the recommended score of the first object based on each of the first similarity values.

[0045] As one possible implementation, the determining submodule is specifically used for:

[0046] Determine the number of first similarity values ​​that satisfy the first preset condition among all the first similarity values;

[0047] Based on the quantity, the recommended score for the first object is determined.

[0048] As one possible implementation, the determining submodule is specifically used for:

[0049] Determine the average value of each of the first similarity values;

[0050] Based on the average value, a recommended score for the first object is determined.

[0051] As one possible implementation, the device further includes:

[0052] The calculation module is used to calculate the similarity between the feature vector of each first object and the feature vector of each other first object in the recommendation sequence of the first object after the recommendation score of each first object is determined, so as to obtain the second similarity value of the first object corresponding to each other first object.

[0053] The first adjustment module is used to adjust the recommendation score of the first object according to the second similarity value, so as to perform the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object.

[0054] As one possible implementation, the device further includes:

[0055] The time determination module is used to determine the historical recommendation time of each of the second objects in the recommendation sequence after determining the recommendation score of each of the first objects;

[0056] The weight value determination module is used to determine the weight value of each of the second objects based on the historical recommendation time.

[0057] The second adjustment module is used to adjust the recommendation score of the first object according to the weight value, so as to perform the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object.

[0058] As one possible implementation, the device further includes:

[0059] The click-through rate determination module is used to determine the predicted click-through rate of the target user for the first object after the recommendation score of each of the first objects is determined;

[0060] The third adjustment module is used to adjust the recommendation score of the first object according to the predicted click-through rate, so as to perform the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object.

[0061] As one possible implementation, the first determining module is specifically used for:

[0062] The feature vectors of each first object in the first object recommendation sequence are obtained from the preset storage service, which stores the feature vectors of each object in the recommendation pool.

[0063] As one possible implementation, the scattering module is specifically used for:

[0064] In the first object recommendation sequence, delete the first object whose recommendation score does not meet the second preset condition to obtain the first object recommendation sequence after deletion.

[0065] For the first object recommendation sequence after deletion, the first objects are reordered according to their recommendation scores to obtain the third object recommendation sequence.

[0066] Thirdly, embodiments of the present invention provide an electronic device, including: a processor and a memory, wherein the processor is configured to execute an object recommendation program stored in the memory to implement the object recommendation method described in any one of the first aspects.

[0067] Fourthly, embodiments of the present invention provide a storage medium storing one or more programs, which can be executed by one or more processors to implement the object recommendation method described in any one aspect.

[0068] The technical solution provided by this invention determines a first object recommendation sequence and the feature vectors of each first object in the first object recommendation sequence (the first objects being objects in the recommendation pool), a second object recommendation sequence and the feature vectors of each second object in the second object recommendation sequence, determines the recommendation score of each first object based on the feature vectors of each first object and each second object, and then shuffles the first object recommendation sequence based on the recommendation scores of each first object to obtain a third object recommendation sequence. Subsequently, objects can be recommended to target users according to the third object recommendation sequence. This technical solution improves the accuracy of the determined recommendation scores compared to existing technologies that determine the recommendation scores of the first and second objects using a single dimension. Therefore, the recommendation score of each first object obtained through the feature vectors of each first object and each second object in the first object recommendation sequence is more accurate. Shuffling the first object recommendation sequence based on the recommendation scores of each first object effectively disperses highly similar first objects within the first object recommendation sequence. This approach distributes highly similar recommended objects more widely to reduce user fatigue when browsing recommendations, thereby improving the user experience. Attached Figure Description

[0069] Figure 1 A flowchart illustrating an embodiment of an object recommendation method provided by this invention;

[0070] Figure 2 A flowchart illustrating an embodiment of another object recommendation method provided by the present invention;

[0071] Figure 3 A flowchart illustrating another embodiment of the object recommendation method provided by this invention;

[0072] Figure 4 A flowchart illustrating another embodiment of the object recommendation method provided by this invention;

[0073] Figure 5 A flowchart illustrating another embodiment of the object recommendation method provided by this invention;

[0074] Figure 6 A flowchart illustrating another embodiment of the object recommendation method provided by this invention;

[0075] Figure 7 A block diagram illustrating an embodiment of an object recommendation device provided by this invention;

[0076] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] To facilitate understanding of the object recommendation method provided by this invention, the background technology involved in this invention will be described below:

[0079] With the development of internet technology, intelligent recommendation services have been widely used because they can quickly show users recommended items that they may be interested in based on their preferences, habits, and browsing history.

[0080] The recommendation process of intelligent recommendation services can be mainly divided into several stages, including recall, coarse ranking, fine ranking, and fragmentation. In the recall stage, multiple recommendation objects are retrieved from the recommendation pool based on user behavior data. In the coarse ranking stage, these multiple recommendation objects are roughly sorted to ensure a certain level of accuracy while reducing the number of recommendation objects.

[0081] Subsequently, in the fine-ranking stage, a ranking model is trained based on user impressions, clicks, and other sample data to rank the recommended objects obtained from the recall or coarse-ranking process. The ranking results are highly correlated with user behavior, leading to high similarity between ranked recommended objects and between recommended objects and previously recommended objects. If highly similar recommended objects are concentrated in one area during the recommendation process, it increases user fatigue when browsing. Therefore, it is necessary to disperse the ranked recommended objects after fine-ranking.

[0082] Currently, in recommendation scenarios, taking videos as the recommended object as an example, the "dispersing" operation mainly relies on the video's own attributes, including title, IP address (video type), and other data, to determine similarity and distribute highly similar recommended objects to reduce user fatigue when browsing recommendations. For example, by calculating the edit distance between different object names, the similarity between two objects can be determined, and the objects can be reordered based on this similarity.

[0083] However, judging the similarity between different recommended objects based on their own attributes is a rather simplistic approach, resulting in low accuracy. This makes it difficult to effectively distribute highly similar recommended objects, leading to a concentration of similar objects in the recommended list. This increases user fatigue when browsing recommended objects and negatively impacts the user experience.

[0084] To address this, embodiments of the present invention provide an object recommendation method to improve the similarity accuracy between the determined recommended object and other recommended objects, thereby distributing the recommendations of highly similar objects, reducing user fatigue when browsing recommended objects, and improving user experience.

[0085] The object recommendation method provided by the present invention will be further explained and described below with reference to the accompanying drawings and specific embodiments. The embodiments do not constitute a limitation on the embodiments of the present invention.

[0086] See Figure 1 This is a flowchart illustrating an embodiment of an object recommendation method provided by the present invention. Figure 1 As shown, the process may include the following steps:

[0087] Step 101: Determine the first object recommendation sequence and the feature vector of each first object in the first object recommendation sequence, wherein the first object is an object in the recommendation pool.

[0088] The first object mentioned above can be an object in the recommendation pool.

[0089] The aforementioned first object recommendation sequence can be the object recommendation sequence output by the ranking model corresponding to the fine ranking stage in the intelligent recommendation service after ranking the recommended objects in the recall stage or the coarse ranking stage.

[0090] In this embodiment of the invention, objects can be mapped to a high-dimensional space using deep learning. Each object has a unique feature vector, and similar content in spatial distribution will be located close to each other in the feature space.

[0091] In this embodiment of the invention, the feature vectors of each first object in the first object recommendation sequence can be obtained through a preset vector model.

[0092] Optionally, the above vector model can be a word2vec (word to vector, a related model that generates word vectors) model or a DSSM (Deep Structured Semantic Models) model. This embodiment of the invention does not limit this.

[0093] As one possible implementation, the execution entity of this embodiment can input all recommended objects in the recommendation pool into the aforementioned preset vector model to obtain the feature vector corresponding to each object in the recommendation pool. Then, the feature vector corresponding to each object in the recommendation pool can be updated to a preset storage service, such as a real-time caching service.

[0094] Based on this, when determining the feature vector of each first object in the first object recommendation sequence, the feature vector of each first object in the first object recommendation sequence can be directly obtained from the aforementioned preset storage service.

[0095] As another possible implementation, the executing entity of this embodiment of the invention can directly input each first object in the first object recommendation sequence into the above vector model to obtain the feature vector of each first object in the first object recommendation sequence.

[0096] Step 102: Determine the second object recommendation sequence and the feature vector of each second object in the second object recommendation sequence. The second object is the object recommended to the target user within a preset historical time period.

[0097] The second object mentioned above can be any object in the recommendation pool recommended to the target user within a historical time period. The preset historical time period can be any historical time period before the current moment, such as the past 24 hours, the past 48 hours, etc., and this embodiment of the invention does not impose any restrictions on this.

[0098] In one embodiment, a second object recommendation sequence can be determined from an already recommended list.

[0099] In this embodiment of the invention, the feature vectors of each second object in the aforementioned second object recommendation sequence can be obtained through a preset vector model. The vector model can be a word2vec model or a DSSM model; this embodiment of the invention does not impose any limitations on this.

[0100] As one possible implementation, the execution entity of this embodiment can input the objects in the recommendation pool into the aforementioned preset vector model to obtain the feature vector corresponding to each object in the recommendation pool. Then, the feature vector corresponding to each object in the recommendation pool can be updated to a preset storage service, such as a real-time caching service.

[0101] Based on this, when determining the feature vector of each second object in the second object recommendation sequence, the feature vector of each second object can be directly obtained from the aforementioned preset storage service.

[0102] As another possible implementation, the executing entity of this embodiment of the invention can directly input each of the second objects in the second object recommendation sequence into the above-mentioned vector model to obtain the feature vector of each of the second objects in the second object recommendation sequence.

[0103] Step 103: Determine the recommended score for each first object based on the feature vectors of each first object and the feature vectors of each second object.

[0104] The aforementioned recommendation scores can be used to measure the recommendation level corresponding to each first object, and this recommendation level can be used to characterize the recommendation order of each first object.

[0105] Optionally, the higher the recommendation score of the first object, the higher the recommendation level of the first object, and the more likely it is to be recommended.

[0106] Based on the description of the feature vectors in step 101 above, in step 103, the execution entity of this embodiment of the invention can determine the similarity between the first object and each second object according to the feature vectors of each first object and each second object, thereby determining the recommended score for each first object. Compared with the prior art, which determines the recommended scores of the first and second objects through a single dimension, this can improve the accuracy of the determined recommended scores.

[0107] As for how the similarity between each first object and the second object is determined based on the feature vectors of each first object, and thus the recommendation score for each first object is determined, this will be explained in the following text. Figure 2 The process shown will be explained in detail here.

[0108] Step 104: Based on the recommendation scores of each first object, the recommendation sequence of the first object is shuffled to obtain the recommendation sequence of the third object.

[0109] The above-mentioned shuffling process may include reordering the recommended sequence of the first object, or it may include deleting the first object that is highly similar to the second object, or it may include deleting the first object that is highly similar to the second object and then reordering the recommended sequence of the first object after the deletion operation. The embodiments of the present invention do not limit this.

[0110] In one embodiment, the executing entity of this invention can reorder the first objects according to their recommendation scores in the first object recommendation sequence to obtain a third object recommendation sequence. For example, the first object with a higher recommendation score in the first object recommendation sequence will rank higher in the third object recommendation sequence.

[0111] In another embodiment, the executing entity of this invention can delete the first object in the first object recommendation sequence whose recommendation score does not meet the second preset condition, thereby obtaining the third object recommendation sequence. The second preset condition can be that the recommendation score is less than a preset score threshold; for example, the first object with a recommendation score less than 80 can be deleted. The preset score threshold can be 80 or 90; this embodiment of the invention does not limit this.

[0112] In another embodiment, the executing entity of this invention can delete first objects in the first object recommendation sequence whose recommendation scores do not meet the second preset condition, thus obtaining a first object recommendation sequence after deletion. Then, for the first object recommendation sequence after deletion, the first object recommendation sequence can be reordered according to the recommendation scores of each first object therein to obtain a third object recommendation sequence.

[0113] Step 105: Recommend objects to the target user according to the third object recommendation sequence.

[0114] The aforementioned target users can be users recommended to the object to be evaluated, such as logged-in users of video playback software or logged-in users of websites, etc., and this embodiment of the invention does not impose any restrictions on this.

[0115] In one embodiment, the executing entity of this invention may recommend objects to the target user according to the aforementioned third object recommendation sequence, either when the target user has already browsed all the currently recommended objects or when the target user is browsing the last N recommended objects.

[0116] The technical solution provided by this invention determines a first object recommendation sequence and the feature vectors of each first object in the first object recommendation sequence (the first objects being objects in the recommendation pool), a second object recommendation sequence and the feature vectors of each second object in the second object recommendation sequence, determines the recommendation score of each first object based on the feature vectors of each first object and each second object, and then shuffles the first object recommendation sequence based on the recommendation scores of each first object to obtain a third object recommendation sequence. Subsequently, objects can be recommended to target users according to the third object recommendation sequence. This technical solution improves the accuracy of the determined recommendation scores compared to existing technologies that determine the recommendation scores of the first and second objects using a single dimension. Therefore, the recommendation score of each first object obtained through the feature vectors of each first object and each second object in the first object recommendation sequence is more accurate. Shuffling the first object recommendation sequence based on the recommendation scores of each first object effectively disperses highly similar first objects within the first object recommendation sequence. This approach distributes highly similar recommended objects more widely to reduce user fatigue when browsing recommendations, thereby improving the user experience.

[0117] See Figure 2 The above is a flowchart of another embodiment of the object recommendation method provided by the present invention. Figure 2 The process shown is in Figure 1 Based on the illustrated process, this section describes how to determine the recommended score for each first object based on the feature vectors of each first object and each second object. For example... Figure 2 As shown, the process may include the following steps:

[0118] Step 201: For each first object, calculate the similarity between the feature vector of the first object and the feature vector of each second object to obtain the first similarity value of the first object to each second object.

[0119] As described above, spatially similar content tends to be located close to each other in the feature space. Therefore, in this embodiment of the invention, by mapping the first object and the second object to a high-dimensional space respectively, the similarity between the first object and the second object is determined based on their respective spatial positions, thereby improving the accuracy of the determined similarity between the first object and the second object. That is, the similarity between the first object and the second object can be determined based on the feature vectors of each first object and each second object.

[0120] Specifically, for each first object in the first object recommendation sequence, the execution entity of this embodiment of the invention can calculate the similarity between the feature vector of the first object and the feature vector of the second object to obtain the similarity value between the first object and the second object (hereinafter referred to as the first similarity value for ease of description).

[0121] Optionally, the execution entity of this embodiment of the invention can determine the first similarity value between the first object and the second object by calculating the inner product between the feature vector of the first object and the feature vector of the second object.

[0122] Step 202: Determine the recommended score for the first object based on each first similarity value.

[0123] In this embodiment of the invention, the recommended score of the first object can be determined based on the first similarity value between the first object and each of the second objects. As described above, the higher the first similarity value of the first object, the lower the recommended score corresponding to that first object.

[0124] As one possible implementation, the number of first similarity values ​​between the first object and the second object that satisfy a first preset condition can be further determined, and the recommended score of the first object can be determined based on the number of such values. The first preset condition may be that the first similarity value is greater than a preset similarity threshold, which may be 90% or 95%, and this embodiment of the invention does not impose any limitation on this.

[0125] Specifically, the larger the number mentioned above, the more similar recommended objects there are. Therefore, when recommending the first object, the probability of the first object and its similar recommended objects being recommended together is greater. Thus, to avoid the first object and its similar recommended objects being recommended together, the probability of the first object being recommended is reduced, and correspondingly, the recommendation score received by the first object is lower.

[0126] Conversely, the smaller the number mentioned above, the fewer similar recommended objects there are. Therefore, when recommending the first object, the probability of the first object and its similar recommended objects being recommended together is lower. Consequently, the probability of the first object being recommended is higher, and correspondingly, the recommendation score received by the first object is higher.

[0127] For example, suppose that step 302 yields 200 first similarity values ​​for each first object. Continuing to assume the first preset condition is a similarity value exceeding 90 points, the number of first similarity values ​​exceeding 90 points can be determined. If 100 of these first similarity values ​​exceed 90 points, the probability of this first object and similar recommended objects being recommended together is relatively high, resulting in a lower recommendation score of 80 points. If only 10 of these first similarity values ​​exceed 90 points, the probability of this first object and similar recommended objects being recommended together is relatively low, resulting in a higher recommendation score of 90 points.

[0128] As another possible implementation, the average value of each first similarity value can be further determined, and the recommended score of the first object can be determined based on the average value.

[0129] Specifically, the reciprocal of the aforementioned average can be used as the recommendation score for the first object. A larger average indicates a higher similarity between the first object and the second object, meaning a greater probability that the first object and similar objects will be recommended together. Therefore, to avoid this concentration of recommendations for the first object and its similar objects, the probability of the first object being recommended should be reduced, resulting in a lower recommendation score for that object.

[0130] Conversely, the smaller the average value, the lower the similarity between the first object and the second object. Therefore, when recommending the first object, the probability of it and similar objects being recommended together is lower. Consequently, the probability of the first object being recommended is higher, and correspondingly, the recommendation score it receives is higher.

[0131] The technical solution provided by this invention calculates the similarity between the feature vector of each first object and the feature vector of each second object to obtain a first similarity value for each second object. Based on these first similarity values, a recommendation score for each first object is determined. This solution improves the accuracy of the similarity values ​​determined between the first and second objects compared to existing technologies that determine recommendation scores based on a single dimension, since each recommended object has a unique feature vector and similar content is spatially close in the feature space. This results in a more accurate recommendation score for the first object. It accurately determines the recommendation score for each first object in the first recommendation sequence, effectively dispersing highly similar first objects within the recommended sequence when shuffling the sequence based on these scores.

[0132] See Figure 3 This is a flowchart of another embodiment of the object recommendation method provided by the present invention. Figure 3 The process shown is in Figure 1 Based on the illustrated process, this section describes how to further adjust the recommended scores for each first object, given the determined recommended scores. For example... Figure 3 As shown, the process may include the following steps:

[0133] Step 301: After determining the recommendation score for each first object, determine the historical recommendation time for each second object in the second object recommendation sequence.

[0134] Step 302: Determine the weight value of each second object based on the historical recommendation time.

[0135] The following provides a unified explanation of steps 301 and 302:

[0136] The aforementioned historical recommendation time refers to the time when each of the second objects in the aforementioned second object recommendation sequence was recommended to the target user.

[0137] As described in step 202, the executing entity of this embodiment can determine the recommendation score of the first object based on the first similarity value. Therefore, when the recommendation score of the first object is determined based on the first similarity value, the historical recommendation time of the aforementioned second object can be further determined.

[0138] Then, the weight value of each second object can be determined based on the historical recommendation time of the second object mentioned above.

[0139] Preferably, the closer the historical recommendation time of the second object is to the current moment, the greater its reference value, and therefore the greater its corresponding weight value.

[0140] For example, suppose the second object recommendation sequence mentioned above includes 5 recommended objects, whose historical recommendation times are 5 hours, 6 hours, 7 hours, 8 hours and 9 hours away from the current time, respectively. Then, according to the historical recommendation time, their corresponding weights can be 0.9, 0.8, 0.7, 0.6 and 0.5 respectively.

[0141] Step 303: Adjust the recommendation score of the first object according to the above weight values, and then perform the above step of shuffling the recommendation sequence of the first object according to the recommendation scores of each first object to obtain the recommendation sequence of the third object.

[0142] In one embodiment, in order to reduce the similarity between the first object and the most recently recommended object, the recommendation score of the first object can be adjusted according to the weight value mentioned above, so as to shuffle the recommendation sequence of the first object according to the adjusted recommendation score and obtain the recommendation sequence of the third object.

[0143] Optionally, the product of the aforementioned weight value and the recommended score of the first object can be used as the adjusted recommended score.

[0144] The technical solution provided by this invention, after determining the recommendation score of a first object, determines the historical recommendation time of each second object in the second object recommendation sequence, determines the weight value of each second object based on the historical recommendation time, and then adjusts the recommendation score of the first object according to the aforementioned weight value. Based on the adjusted recommendation score, the step of shuffling the first object recommendation sequence according to the recommendation scores of each first object is performed to obtain the third object recommendation sequence. This technical solution, by determining the weight value of the second object based on its historical recommendation time and adjusting the recommendation score of the first object based on this weight value, reduces the similarity between recommended objects and recently recommended objects, thereby reducing user fatigue and improving user experience.

[0145] See Figure 4 The above is a flowchart of another embodiment of the object recommendation method provided by the present invention. Figure 4 The process shown is in Figure 1 Based on the illustrated process, after determining the recommended score for each first object, how can these recommended scores be further adjusted? For example... Figure 4 As shown, the process may include the following steps:

[0146] Step 401: After determining the recommendation score for each first object, determine the predicted click-through rate of the target user for the first object.

[0147] As described in step 202, the executing entity of this embodiment can determine the recommended score of the first object based on the first similarity value. Based on this, after determining the recommended score of the first object, the predicted click-through rate of the target user for the first object can be further determined.

[0148] As one possible implementation, the predicted click-through rate of the first object can be determined by finding the click-through rate of the target users for objects of the same type as the first object.

[0149] As another possible implementation, the first object can be input into a preset click-through rate prediction model, and the predicted click-through rate of the first object can be determined based on the output value of the click-through rate prediction model.

[0150] Step 402: Adjust the recommendation score of the first object according to the predicted click-through rate, and then perform the above steps of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object.

[0151] In one embodiment, in order to further ensure that the target user is interested in the recommended object, the recommendation score of the first object can be further adjusted by the predicted click-through rate of the target user on the first object, so as to improve the user experience.

[0152] As one possible implementation, the product of the predicted click-through rate of the target user for the first object and the recommended score can be used as the adjusted recommended score.

[0153] As another possible implementation, the sum of the predicted click-through rate of the target user for the first object and the recommended score can be used as the adjusted recommended score.

[0154] Then, based on the adjusted recommendation scores, the recommendation sequence of the first object can be shuffled to obtain the recommendation sequence of the third object.

[0155] The technical solution provided by this invention, after determining the recommendation score of a first object, determines the predicted click-through rate of the target user for the first object, and adjusts the recommendation score of the first object based on the predicted click-through rate. Then, based on the adjusted recommendation score, the step of shuffling the recommendation sequence of the first objects according to their respective recommendation scores is performed to obtain the recommendation sequence of the third object. This technical solution, by adjusting the recommendation score of the first object based on the predicted click-through rate of the target user, achieves the goal of recommending objects of interest to the target user while reducing the similarity between recommended objects and other objects, thereby improving the user experience.

[0156] See Figure 5 The following is a flowchart of an embodiment of an object recommendation method provided by the present invention.Figure 5 The process shown is in Figure 1 Based on the illustrated process, after determining the recommended score for each first object, how can these recommended scores be further adjusted? For example... Figure 5 As shown, the process may include the following steps:

[0157] Step 501: Calculate the similarity between the feature vector of each first object and the feature vectors of other first objects in the recommendation sequence of the first object, and obtain the second similarity value of the first object to each other first object.

[0158] As described in step 202, the execution entity of this embodiment can determine the recommendation score of the first object based on the first similarity value. Therefore, after determining the recommendation score of the first object, in order to further reduce the probability of the objects to be recommended being centrally recommended, the execution entity of this embodiment can further determine the similarity between each first object and each other first object in the first object recommendation sequence.

[0159] Specifically, the second similarity value of the first object to each other first object can be obtained by calculating the similarity between the feature vector of each first object and the feature vector of each other first object in the recommendation sequence of the first object.

[0160] Optionally, the execution entity of this embodiment of the invention can determine the second similarity value of the first object corresponding to each of the other first objects by calculating the inner product between the feature vector of the first object and the feature vector of each other first object in the recommendation sequence of the first object.

[0161] Step 502: Adjust the recommendation score of the first object according to the second similarity value, and then perform the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object.

[0162] In one embodiment, in order to reduce the probability that the first object is recommended in a concentrated manner, a second similarity value between the first object and each other first object in the first object recommendation sequence can be determined, and the recommendation score of the first object can be adjusted according to the second similarity value, so as to break up the first object recommendation sequence according to the adjusted recommendation score to obtain a third object recommendation sequence.

[0163] Specifically, firstly, a similarity value (hereinafter referred to as the third similarity value) can be determined based on the second similarity values ​​corresponding to each first object.

[0164] As a possible implementation, for each first object, the sum of the second similarity values ​​corresponding to that first object can be used to determine the third similarity value.

[0165] As another possible implementation, for each first object, the average of the second similarity values ​​corresponding to that first object can be determined as the third similarity value.

[0166] As another possible implementation, for each first object, the number of second similarity values ​​that satisfy the third preset condition can be determined, and the third similarity value can be determined based on this number. The aforementioned third preset condition may be that the second similarity value is greater than a preset similarity threshold, which may be 90% or 95%, and the embodiments of the present invention do not limit this.

[0167] Optionally, the above quantities are positively correlated with the third similarity value mentioned above.

[0168] Then, the recommended score for the first object can be adjusted based on the aforementioned third similarity value.

[0169] Optionally, a weight value can be determined based on the third similarity value, and the product of the weight value and the recommended score of the first object can be used as the adjusted recommended score of the first object.

[0170] Specifically, the larger the third similarity value mentioned above, the more similar first objects there are to the first object. Therefore, when recommending the first object, the probability that the first object and its similar recommended objects will be recommended together is greater. Thus, to avoid the first object and its similar recommended objects being recommended together, the probability of the first object being recommended is reduced, and correspondingly, the weight value obtained by the first object is smaller.

[0171] Conversely, the smaller the third similarity value, the fewer recommended objects are similar to the first object. Therefore, when recommending the first object, the probability of the first object and its similar recommended objects being recommended together is lower. Consequently, the probability of the first object being recommended is higher, and correspondingly, the weight value received by the first object is greater.

[0172] The technical solution provided by this invention, after determining the recommended score of each first object, calculates the similarity between the feature vector of each first object and the feature vectors of other first objects in the recommended sequence of the first object, obtaining a second similarity value for each first object corresponding to each other first object. The recommended score of the first object is then adjusted based on the second similarity value. The recommended sequence of the first objects is then shuffled based on the adjusted recommended scores to obtain a third object recommended sequence. This technical solution, by further determining the second similarity value of each first object corresponding to each other in the recommended sequence of the first object, and adjusting the recommended score of the first object based on this second similarity value, can improve the accuracy of the determined recommended scores of the first objects, making the determined recommended scores of the first objects more accurate. It achieves accurate determination of the recommended scores of each first object in the first recommended sequence, thereby effectively dispersing the first objects with high similarity in the first object recommended sequence when shuffling the first object sequence based on the recommended scores.

[0173] See Figure 6 The following is a flowchart of an embodiment of an object recommendation method provided by the present invention. Figure 6 The illustrated process uses video recommendation as an example to illustrate the object recommendation method provided in this embodiment of the invention. Figure 6 As shown, the process may include the following steps:

[0174] First, regarding the recommendation pool, the execution entity of this embodiment can construct training samples based on user operation data such as historical display or clicks of historical recommended objects to obtain a video feature vector model. Each video in the recommendation pool is then input into this video feature vector model to obtain the feature vector of each video in the recommendation pool. The real-time storage service is then updated, updating the feature vector of each video in the recommendation pool to the real-time cache service.

[0175] Then, upon receiving a user's video recommendation request, the recall model outputs multiple recommended videos related to the user's behavior information, and the ranking model sorts the recommended videos to output a ranked list, which is the first object recommendation sequence.

[0176] Then, the executing entity of this embodiment of the invention can obtain the recommended videos currently displayed by the user, as well as the recommended videos displayed by the user in the past, to obtain a real-time display video list, and click on the cache service to obtain the feature vectors of the above-mentioned finely ranked videos and the videos already displayed by the user.

[0177] Finally, according to Figure 1 and Figure 2The rearrangement and shuffling algorithm provided in the process shown (that is, the object recommendation method provided in the embodiment of the present invention) determines the similarity between each recommended video in the fine-ranked list and other videos (including other recommended videos in the fine-ranked list besides the current recommended video and the displayed videos), and shuffles the fine-ranked list according to the above similarity to obtain the final recommendation list.

[0178] See Figure 7 This is a block diagram illustrating an embodiment of an object recommendation device provided by an embodiment of the present invention. Figure 7 As shown, the device includes:

[0179] The first determining module 71 is used to determine the first object recommendation sequence and to determine the feature vector of each first object in the first object recommendation sequence, wherein the first object is an object in the recommendation pool;

[0180] The second determining module 72 is used to determine the second object recommendation sequence and to determine the feature vector of each second object in the second object recommendation sequence, wherein the second object is an object recommended to the target user within a preset historical time period;

[0181] The third determining module 73 is used to determine the recommended score of each first object based on the feature vector of each first object and the feature vector of each second object;

[0182] The scattering module 74 is used to scatter the recommendation sequence of the first object according to the recommendation score of each of the first objects to obtain the recommendation sequence of the third object;

[0183] The recommendation module 75 is used to recommend objects to the target user according to the third object recommendation sequence.

[0184] As one possible implementation, the third determining module 73 includes:

[0185] The similarity calculation submodule is used to calculate the similarity between the feature vector of the first object and the feature vector of each second object for each first object, so as to obtain the first similarity value of the first object corresponding to each second object.

[0186] The determination submodule is used to determine the recommended score of the first object based on each of the first similarity values.

[0187] As one possible implementation, the determining submodule is specifically used for:

[0188] Determine the number of first similarity values ​​that satisfy the first preset condition among all the first similarity values;

[0189] Based on the quantity, the recommended score for the first object is determined.

[0190] As one possible implementation, the determining submodule is specifically used for:

[0191] Determine the average value of each of the first similarity values;

[0192] Based on the average value, a recommended score for the first object is determined.

[0193] As one possible implementation, the device further includes (not shown in the figure):

[0194] The calculation module is used to calculate the similarity between the feature vector of each first object and the feature vector of each other first object in the recommendation sequence of the first object after the recommendation score of each first object is determined, so as to obtain the second similarity value of the first object corresponding to each other first object.

[0195] The first adjustment module is used to adjust the recommendation score of the first object according to the second similarity value, so as to perform the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object.

[0196] As one possible implementation, the device further includes (not shown in the figure):

[0197] The time determination module is used to determine the historical recommendation time of each of the second objects in the recommendation sequence after determining the recommendation score of each of the first objects;

[0198] The weight value determination module is used to determine the weight value of each of the second objects based on the historical recommendation time.

[0199] The second adjustment module is used to adjust the recommendation score of the first object according to the weight value, so as to perform the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object.

[0200] As one possible implementation, the device further includes (not shown in the figure):

[0201] The click-through rate determination module is used to determine the predicted click-through rate of the target user for the first object after the recommendation score of each of the first objects is determined;

[0202] The third adjustment module is used to adjust the recommendation score of the first object according to the predicted click-through rate, so as to perform the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object.

[0203] As one possible implementation, the first determining module 71 is specifically used for:

[0204] The feature vectors of each first object in the first object recommendation sequence are obtained from the preset storage service, which stores the feature vectors of each object in the recommendation pool.

[0205] As one possible implementation, the dispersing module 74 is specifically used for:

[0206] In the first object recommendation sequence, delete the first object whose recommendation score does not meet the second preset condition to obtain the first object recommendation sequence after deletion.

[0207] For the first object recommendation sequence after deletion, the first objects are reordered according to their recommendation scores to obtain the third object recommendation sequence.

[0208] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Figure 8 The illustrated electronic device 800 includes at least one processor 801, a memory 802, at least one network interface 804, and a user interface 803. The various components in the electronic device 800 are coupled together via a bus system 805. It is understood that the bus system 805 is used to implement communication between these components. In addition to a data bus, the bus system 805 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 8 The general labeled all buses as Bus System 805.

[0209] The user interface 803 may include a display, keyboard, or clicking device (e.g., mouse, trackball, touchpad, or touchscreen).

[0210] It is understood that the memory 802 in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous DRAM (SDRAM), Double Data Rate Synchronous DRAM (DDRSDRAM), Enhanced Synchronous DRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 802 described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0211] In some implementations, memory 802 stores elements, executable units or data structures, or subsets thereof, or extended sets thereof: operating system 8021 and application programs 8022.

[0212] The operating system 8021 includes various system programs, such as the framework layer, core library layer, and driver layer, used to implement various basic business functions and handle hardware-based tasks. The application program 8022 includes various applications, such as a media player and a browser, used to implement various application functions. The program implementing the method of this embodiment can be included in the application program 8022.

[0213] In this embodiment of the invention, by calling the program or instructions stored in the memory 802, specifically the program or instructions stored in the application program 8022, the processor 801 executes the method steps provided in each method embodiment, including, for example:

[0214] Determine a first object recommendation sequence and determine the feature vector of each first object in the first object recommendation sequence, wherein the first object is an object in the recommendation pool;

[0215] Determine the second object recommendation sequence and the feature vector of each second object in the second object recommendation sequence, wherein the second object is an object recommended to the target user within a preset historical time period;

[0216] The recommended score for each of the first objects is determined based on the feature vectors of each of the first objects and the feature vectors of each of the second objects.

[0217] Based on the recommendation scores of each of the first objects, the recommendation sequence of the first objects is shuffled to obtain the recommendation sequence of the third objects;

[0218] Recommend objects to the target user according to the third object recommendation sequence.

[0219] The methods disclosed in the above embodiments of the present invention can be applied to or implemented by processor 801. Processor 801 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 801 or by instructions in the form of software. The processor 801 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software units in the decoding processor. The software units may be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 802. Processor 801 reads the information in memory 802 and, in conjunction with its hardware, completes the steps of the above method.

[0220] It is understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described herein, or combinations thereof.

[0221] For software implementation, the techniques described herein can be implemented by units that perform the functions described herein. The software code can be stored in memory and executed by a processor. The memory can be implemented in the processor or external to the processor.

[0222] The electronic device provided in this embodiment may be as follows: Figure 8 The electronic device shown can perform the following: Figures 1-6 All steps of the object recommendation method, thereby achieving Figures 1-6 For details on the technical effectiveness of the object recommendation method shown, please refer to [link / reference]. Figures 1-6 The relevant descriptions are presented concisely and will not be elaborated upon here.

[0223] This invention also provides a storage medium (computer-readable storage medium). This storage medium stores one or more programs. The storage medium may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; the memory may also include combinations of the above types of memory.

[0224] When one or more programs in the storage medium can be executed by one or more processors to implement the above-mentioned recommended method of executing objects on the electronic device side.

[0225] The processor is used to execute an object recommendation program stored in memory to implement the following steps of an object recommendation method executed on the electronic device side:

[0226] Determine a first object recommendation sequence and determine the feature vector of each first object in the first object recommendation sequence, wherein the first object is an object in the recommendation pool;

[0227] Determine the second object recommendation sequence and the feature vector of each second object in the second object recommendation sequence, wherein the second object is an object recommended to the target user within a preset historical time period;

[0228] The recommended score for each of the first objects is determined based on the feature vectors of each of the first objects and the feature vectors of each of the second objects.

[0229] Based on the recommendation scores of each of the first objects, the recommendation sequence of the first objects is shuffled to obtain the recommendation sequence of the third objects;

[0230] Recommend objects to the target user according to the third object recommendation sequence.

[0231] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0232] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0233] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An object recommendation method, characterized in that, The method includes: Determine a first object recommendation sequence and determine the feature vector of each first object in the first object recommendation sequence, wherein the first object is an object in the recommendation pool; Determine the second object recommendation sequence and the feature vector of each second object in the second object recommendation sequence, wherein the second object is an object recommended to the target user within a preset historical time period; The recommended score for each first object is determined based on the feature vectors of each first object and the feature vectors of each second object, including: for each first object, calculating the similarity between the feature vector of the first object and the feature vectors of each second object to obtain a first similarity value of the first object corresponding to each second object; and determining the recommended score of the first object based on the first similarity values. After determining the recommendation score of each of the first objects, the method further includes: determining the historical recommendation time of each of the second objects in the second object recommendation sequence; determining the weight value of each of the second objects based on the historical recommendation time; adjusting the recommendation score of the first objects based on the weight value, so as to perform the step of shuffling the first object recommendation sequence based on the recommendation score of each of the first objects to obtain the third object recommendation sequence based on the adjusted recommendation score; Based on the recommendation scores of each of the first objects, the recommendation sequence of the first objects is shuffled to obtain a third object recommendation sequence, including: deleting the first objects in the first object recommendation sequence whose recommendation scores do not meet the second preset condition, to obtain a first object recommendation sequence after deletion; and reordering each of the first objects in the first object recommendation sequence after deletion based on the recommendation scores of each of the first objects, to obtain a third object recommendation sequence. Recommend objects to the target user according to the third object recommendation sequence.

2. The method according to claim 1, characterized in that, The step of determining the recommendation score for the first object based on each of the first similarity values ​​includes: Determine the number of first similarity values ​​that satisfy the first preset condition among all the first similarity values; Based on the quantity, the recommended score for the first object is determined.

3. The method according to claim 1, characterized in that, The step of determining the recommendation score for the first object based on each of the first similarity values ​​includes: Determine the average value of each of the first similarity values; Based on the average value, a recommended score for the first object is determined.

4. The method according to claim 1, characterized in that, After determining the recommendation score for each of the first objects, the method further includes: The similarity between the feature vector of each first object and the feature vector of each other first object in the recommendation sequence of the first object is calculated to obtain the second similarity value of the first object to each other first object; The recommendation score of the first object is adjusted based on the second similarity value, and the step of shuffling the recommendation sequence of the first object based on the recommendation score of each first object to obtain the recommendation sequence of the third object is performed based on the adjusted recommendation score.

5. The method according to claim 1, characterized in that, After determining the recommendation score for each of the first objects, the method further includes: Determine the predicted click-through rate of the target user for the first object; The recommended score of the first object is adjusted based on the predicted click-through rate, and the step of shuffling the recommended sequence of the first object based on the recommended score of each first object is performed to obtain the recommended sequence of the third object.

6. The method according to claim 1, characterized in that, Determining the feature vector of each first object in the first object recommendation sequence includes: The feature vectors of each first object in the first object recommendation sequence are obtained from the preset storage service, which stores the feature vectors of each object in the recommendation pool.

7. An object recommendation device, characterized in that, The device includes: The first determining module is used to determine the first object recommendation sequence and to determine the feature vector of each first object in the first object recommendation sequence, wherein the first object is an object in the recommendation pool; The second determining module is used to determine the second object recommendation sequence and to determine the feature vector of each second object in the second object recommendation sequence, wherein the second object is an object recommended to the target user within a preset historical time period; The third determining module is used to determine the recommended score of each first object based on the feature vector of each first object and the feature vector of each second object; The third determining module includes: The similarity calculation submodule is used to calculate the similarity between the feature vector of the first object and the feature vector of each second object for each first object, so as to obtain the first similarity value of the first object corresponding to each second object. The determination submodule is used to determine the recommendation score of the first object based on each of the first similarity values; The time determination module is used to determine the historical recommendation time of each of the second objects in the second object recommendation sequence; The weight value determination module is used to determine the weight value of each of the second objects based on the historical recommendation time. The second adjustment module is used to adjust the recommendation score of the first object according to the weight value, so as to perform the step of shuffling the recommendation sequence of the first object according to the recommendation score of each first object to obtain the recommendation sequence of the third object. The shuffling module is used to shuffle the recommendation sequence of the first objects according to the recommendation scores of each of the first objects to obtain a third object recommendation sequence. Specifically, it is used to: delete the first objects in the first object recommendation sequence whose recommendation scores do not meet the second preset condition to obtain a deleted first object recommendation sequence; and reorder each of the first objects in the deleted first object recommendation sequence according to the recommendation scores of each of the first objects to obtain a third object recommendation sequence. The recommendation module is used to recommend objects to the target user according to the third object recommendation sequence.

8. An electronic device, characterized in that, include: A processor and a memory, the processor being configured to execute an object recommendation program stored in the memory to implement the object recommendation method according to any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium stores one or more programs, which can be executed by one or more processors to implement the object recommendation method according to any one of claims 1 to 6.

Images