Training method, device and electronic equipment for satisfactory descent model

Abstract

The present disclosure provides a training method, device and electronic device for a satisfaction decline model, which relates to the field of artificial intelligence technology, especially to the field of deep learning, natural language processing, computer vision and intelligent search technology. The specific implementation scheme is: obtaining a video browsing behavior record; determining a satisfaction decline label corresponding to the video browsing behavior record based on the browsing behavior description information in the video browsing behavior record; the satisfaction decline label indicates whether the browsing object has a satisfactory decline behavior for browsing the video; determining training data based on the first attribute information of the browsing object and the second attribute information of the browsing video in multiple video browsing behavior records, as well as the satisfaction decline label corresponding to the video browsing behavior record; and then training to obtain a satisfaction decline model, so that when recommending videos, the probability of the object's satisfaction decline for the video can be considered, the efficiency of video recommendation can be improved, and the continuous consumption of videos can be achieved.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, in particular to the field of deep learning, natural language processing, computer vision, and intelligent search technology, and in particular to a training method, device, and electronic device for a satisfaction decline model. Background Art

[0002] In the current video recommendation method, the object's preference information for the video is determined based on the object's historical browsing behavior for the video, such as the browsing time of the video, the interaction behavior with the video, etc.; the video is recalled based on the preference information and the object's attribute information for video recommendation processing.

[0003] In the above video recommendation method, whether the subject is satisfied with the video when scrolling down the video is not considered, that is, whether the subject is satisfied with the video and scrolls down. The information considered when recommending videos is relatively one-sided, resulting in a low possibility that the subject is satisfied with the recommended video after browsing and scrolls down. The video recommendation efficiency is poor, it is difficult to achieve continuous consumption of videos, and the subject's consumption stickiness to the video is reduced. Summary of the invention

[0004] The present invention provides a training method, device and electronic equipment for a satisfactory descent model.

[0005] According to one aspect of the present disclosure, a method for training a satisfaction slump model is provided, the method comprising: obtaining a video browsing behavior record; the video browsing behavior record comprising first attribute information of a browsing object, second attribute information of a browsing video corresponding to the browsing object, and browsing behavior description information; determining a satisfaction slump label corresponding to the browsing video behavior record based on the browsing behavior description information; the satisfaction slump label indicates whether the browsing object has a satisfactory slump behavior for the browsing video; determining training data based on the first attribute information and the second attribute information in a plurality of browsing video behavior records and the satisfaction slump labels corresponding to the browsing video behavior records; and using the training data to train an initial satisfaction slump model to obtain a trained satisfaction slump model for video recommendation processing.

[0006] According to another aspect of the present disclosure, a video recommendation method is provided, the method comprising: obtaining first attribute information of a target object and second attribute information of a plurality of candidate videos; for each candidate video, inputting the second attribute information of the candidate video and the first attribute information into a preset satisfaction decline model to obtain a satisfaction decline probability output by the satisfaction decline model; the satisfaction decline model is trained by combining the first attribute information of a browsing object in a plurality of browsing video behavior records, the second attribute information of a browsing video corresponding to the browsing object, and a satisfaction decline label corresponding to the browsing video behavior records; selecting a video to be recommended from the plurality of candidate videos in combination with the satisfaction decline probabilities of the plurality of candidate videos; and recommending the video to be recommended to the target object.

[0007] According to another aspect of the present disclosure, a training device for a satisfaction slump model is provided, the device comprising: an acquisition module for acquiring browsing video behavior records; the browsing video behavior records comprising first attribute information of a browsing object, second attribute information of a browsing video corresponding to the browsing object, and browsing behavior description information; a first determination module for determining a satisfaction slump label corresponding to the browsing video behavior record according to the browsing behavior description information; the satisfaction slump label indicates whether the browsing object has a satisfactory slump behavior for browsing the video; a second determination module for determining training data according to the first attribute information and the second attribute information in a plurality of browsing video behavior records, and the satisfaction slump label corresponding to the browsing video behavior records; a training processing module for using the training data to perform training processing on an initial satisfaction slump model to obtain a trained satisfaction slump model for video recommendation processing.

[0008] According to another aspect of the present disclosure, a video recommendation device is provided, the device comprising: a first acquisition module, used to acquire first attribute information of a target object, and second attribute information of a plurality of candidate videos; a second acquisition module, used to input the second attribute information of the candidate video and the first attribute information into a preset satisfaction decline model for each candidate video, and acquire the satisfaction decline probability output by the satisfaction decline model; the satisfaction decline model is trained by combining the first attribute information of the browsing object in a plurality of browsing video behavior records, the second attribute information of the browsing video corresponding to the browsing object, and the satisfaction decline label corresponding to the browsing video behavior records; a selection module, used to select a video to be recommended from a plurality of candidate videos in combination with the satisfaction decline probabilities of the plurality of candidate videos; and a recommendation module, used to recommend the video to be recommended to the target object.

[0009] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the training method of the satisfactory decline model or the video recommendation method proposed above in the present disclosure.

[0010] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute the training method of the satisfactory decline model or the video recommendation method proposed in the present disclosure.

[0011] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, which, when executed by a processor, implements the steps of the training method of the satisfaction decline model or the steps of the video recommendation method proposed above in the present disclosure.

[0012] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0014] Figure 1 is a schematic diagram according to a first embodiment of the present disclosure;

[0015] Figure 2 is a schematic diagram according to a second embodiment of the present disclosure;

[0016] Figure 3 is a schematic diagram according to a third embodiment of the present disclosure;

[0017] Figure 4 is a schematic diagram according to a fourth embodiment of the present disclosure;

[0018] Figure 5 is a schematic diagram according to a fifth embodiment of the present disclosure;

[0019] Figure 6 is a schematic diagram according to a sixth embodiment of the present disclosure;

[0020] Figure 7 It is a block diagram of an electronic device used to implement the training method of the satisfaction decline model or the video recommendation method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0021] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0022] In the current video recommendation method, the object's preference information for the video is determined based on the object's historical browsing behavior for the video, such as the browsing time of the video, the interaction behavior with the video, etc.; the video is recalled based on the preference information and the object's attribute information for video recommendation processing.

[0023] In the above video recommendation method, whether the subject is satisfied with the video when scrolling down the video is not considered, that is, whether the subject is satisfied with the video and scrolls down. The information considered when recommending videos is relatively one-sided, resulting in a low possibility that the subject is satisfied with the recommended video after browsing and scrolls down. The video recommendation efficiency is poor, it is difficult to achieve continuous consumption of videos, and the subject's consumption stickiness to the video is reduced.

[0024] In view of the above problems, the present disclosure provides a training method, device and electronic equipment for a satisfactory descent model.

[0025] Figure 1 It is a schematic diagram according to the first embodiment of the present disclosure. It should be noted that the training method of the satisfactory descent model of the embodiment of the present disclosure can be applied to a training device of the satisfactory descent model, and the device can be configured in an electronic device so that the electronic device can perform the training function of the satisfactory descent model. In the following embodiments, the execution subject is an electronic device as an example for explanation.

[0026] Among them, the electronic device can be any device with computing capabilities, such as a personal computer (PC), a mobile terminal, a server, etc. The mobile terminal can be, for example, a vehicle-mounted device, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, a smart speaker, and other hardware devices with various operating systems, touch screens and / or display screens.

[0027] like Figure 1 As shown, the training method of the satisfaction decline model may include the following steps:

[0028] Step 101, obtaining a video browsing behavior record; the video browsing behavior record includes first attribute information of a browsing object, second attribute information of a browsing video corresponding to the browsing object, and browsing behavior description information.

[0029] In the disclosed embodiment, the video browsing behavior record is a record formed by the browsing object browsing the video displayed in the software. The software here refers to software that can provide videos for browsing, such as video software, etc., which can be set according to actual needs. The video browsing behavior record can be generated by the software during the browsing process of the browsing object and provided to the electronic device for training processing of the satisfaction decline model.

[0030] The browsing object, for example, is a user browsing a video, etc. The first attribute information of the browsing object, for example, is the age, gender, education level, etc. of the browsing object.

[0031] The browsed video refers to the video resource browsed by the browse object. The second attribute information of the browsed video includes, for example, the total duration, publisher, content summary, etc. of the browsed video.

[0032] In one example, the browsing behavior description information may include at least one of the following: browsing time, whether there is a swiping behavior after browsing, whether there is an exit behavior after browsing, whether there is a video like behavior, whether there is a video evaluation behavior, whether there is a video follow behavior, etc. In another example, the browsing behavior description information may include the operation type and operation time point of each operation of the browsing object. According to the operation type and operation time point of each operation, it can be determined whether there is a swiping behavior after browsing, whether there is an exit behavior after browsing, whether there is a video like behavior, whether there is a video evaluation behavior, whether there is a video follow behavior, etc.

[0033] Step 102, determining the satisfied sliding label corresponding to the video browsing behavior record according to the browsing behavior description information; the satisfied sliding label indicates whether the browsing object has satisfied sliding behavior for browsing the video.

[0034] In an embodiment of the present disclosure, in one example, the browsing behavior description information may include browsing time and whether there is a post-browsing swiping behavior. Correspondingly, the process of the electronic device executing step 102 may be, for example, that when the browsing time is greater than or equal to the first time threshold and there is a post-browsing swiping behavior, the value of the satisfactory swiping label corresponding to the browsing video behavior record is determined to be a first value; the first value indicates that the browsing object has a satisfactory swiping behavior for browsing the video; when the browsing time is less than the first time threshold, or there is no post-browsing swiping behavior, the value of the satisfactory swiping label corresponding to the browsing video behavior record is determined to be a second value; the second value indicates that the browsing object has no satisfactory swiping behavior for the video.

[0035] The first duration threshold may be determined based on the average playback duration of each video in the application to which the video belongs. The average playback duration of each video refers to the average playback duration of each video. For example, the first duration threshold may be 10 seconds.

[0036] The first value and the second value are different; the first value and the second value can be any values ​​and are not specifically limited here.

[0037] Among them, in this example, the electronic device can automatically determine the satisfied slump label corresponding to the browsing video behavior record according to the browsing time and whether there is a slump behavior after browsing, avoiding manual processing, reducing the cost and time of obtaining training data, and improving the training speed of the satisfied slump model.

[0038] In another example of the disclosed embodiment, the browsing behavior description information may include browsing time, whether there is a swiping behavior after browsing, and whether there is an exit behavior after browsing. Correspondingly, the process of the electronic device executing step 102 may be, for example, that when the browsing time is greater than or equal to the first time threshold and there is a swiping behavior after browsing, the value of the satisfactory swiping label corresponding to the browsing video behavior record is determined to be a first value; the first value indicates that the browsing object has a satisfactory swiping behavior for browsing the video; when the browsing time is less than the first time threshold and there is a swiping behavior after browsing, or when the browsing time is less than or equal to the second time threshold and there is an exit behavior after browsing, the value of the satisfactory swiping label corresponding to the browsing video behavior record is determined to be a second value; the second value indicates that the browsing object has no satisfactory swiping behavior for the video; the second time threshold is less than the first time threshold.

[0039] Wherein, when the browsing time is greater than or equal to the first time threshold and there is a swiping behavior after browsing, it indicates that there is a satisfied browsing and swiping behavior. Wherein, when the browsing time is less than or equal to the second time threshold and there is an exit behavior after browsing, it indicates that there is an exit behavior after a quick swipe, indicating that the browsing object is not satisfied with the browsing video and exits. Wherein, when the browsing time is less than the first time threshold and there is a swiping behavior after browsing, it indicates that there is an unsatisfied browsing and swiping behavior.

[0040] The first duration threshold may be determined based on the average playback duration of each video in the application to which the video belongs. The average playback duration of each video refers to the average playback duration of each video. For example, the first duration threshold may be 10 seconds, etc.; the second duration threshold may be 3 seconds, etc.

[0041] Among them, in this example, the electronic device can automatically determine the satisfied decline label corresponding to the browsing video behavior record according to the browsing time, whether there is a decline behavior after browsing, and whether there is an exit behavior after browsing, thereby avoiding manual processing, reducing the cost and time of obtaining training data, and improving the training speed of the satisfied decline model.

[0042] Step 103 : determining training data according to the first attribute information and the second attribute information in the plurality of video browsing behavior records, and the satisfaction slide labels corresponding to the video browsing behavior records.

[0043] In the disclosed embodiment, in order to allow the satisfied decline model to consider more information when predicting and improve the prediction accuracy of the satisfied decline model, before step 103, the electronic device may also perform the following process: determine the predicted probability of the browsing object performing a specified interactive behavior on the browsing video according to the first attribute information and the second attribute information. Correspondingly, the process of the electronic device performing step 103 may be, for example, determining the training data according to the first attribute information and the second attribute information in a plurality of browsing video behavior records, the predicted probability of the browsing object performing a specified interactive behavior on the browsing video in the browsing video behavior records, and the satisfied decline label corresponding to the browsing video behavior records.

[0044] The specified interactive behavior may include at least one of the following: video like behavior, video evaluation behavior, video follow behavior, video fast scrolling behavior, video exit behavior, and video long browsing behavior. In the video fast scrolling behavior, the browsing time of the browsing object for browsing the video is less than or equal to the second time threshold; in the video long browsing behavior, the browsing time of the browsing object for browsing the video is greater than or equal to the first time threshold.

[0045] The predicted probability of the browsing object performing a specified interactive behavior on the browsed video can be predicted by other models, such as a like prediction model, a review prediction model, a follow prediction model, etc. The input of the above model can at least include the first attribute information of the browsing object and the second attribute information of the browsed video.

[0046] In the disclosed embodiment, in order to allow the satisfied decline model to consider more information when predicting and further improve the prediction accuracy of the satisfied decline model, the browsing video behavior record also includes browsing scene information indicating the video browsing batch of the browsing object. Correspondingly, the process of the electronic device executing step 103 may be, for example, determining the training data according to the first attribute information, the second attribute information and the browsing scene information in the plurality of browsing video behavior records, and the satisfied decline labels and weights corresponding to the browsing video behavior records.

[0047] The video browsing batch refers to the batch of videos browsed by the browsing object in the software. For example, after the software is started, the first batch of videos displayed are the first-viewed videos; after the first batch of videos are browsed and scrolled, the second batch of videos displayed are the second-viewed videos, and so on.

[0048] Step 104 , using the training data, the initial satisfaction decline model is trained to obtain a trained satisfaction decline model for video recommendation processing.

[0049] In the disclosed embodiment, the satisfaction decline model can be a model based on the extreme gradient boosting xgboost algorithm. Among them, the xgboost algorithm is a classification algorithm based on a decision tree, which has the characteristics of high adaptability and high robustness. The model based on the xgboost algorithm can obtain a satisfactory decline model with high accuracy after training with training data, which can further improve the efficiency of video recommendation when used for video recommendation processing.

[0050] The training data may be determined based on the video browsing behavior record of the software within a preset time period. The preset time period may be, for example, one day, one week, etc. The video browsing behavior record of the software within the preset time period may be the video browsing behavior record of all objects using the software within the preset time period.

[0051] The training method of the satisfied decline model of the embodiment of the present disclosure is through obtaining the browsing video behavior record; the browsing video behavior record includes the first attribute information of the browsing object, the second attribute information of the browsing video corresponding to the browsing object, and the browsing behavior description information; according to the browsing behavior description information, the satisfied decline label corresponding to the browsing video behavior record is determined; the satisfied decline label indicates whether the browsing object has satisfied decline behavior on the browsing video; according to the first attribute information and the second attribute information in a plurality of browsing video behavior records, and the satisfied decline label corresponding to the browsing video behavior record, the training data is determined; the initial satisfied decline model is trained with the training data to obtain the trained satisfied decline model for video recommendation processing, so that when recommending videos, the probability of satisfied decline of the object on the video can be considered, the efficiency of video recommendation can be improved, the continuous consumption of the video can be realized, and the consumption stickiness of the object on the video can be improved.

[0052] In order to further improve the accuracy of the training data obtained, and thus improve the accuracy of the trained satisfaction decline model, the weight corresponding to the browsing video behavior record can be adjusted according to whether there is a specified interactive behavior in the browsing behavior description information during the training of the satisfaction decline model, and the training data can be determined in combination with the weight. Figure 2 As shown, Figure 2 is a schematic diagram according to a second embodiment of the present disclosure, Figure 2The illustrated embodiment may include the following steps:

[0053] Step 201, obtaining a video browsing behavior record; the video browsing behavior record includes first attribute information of a browsing object, second attribute information of a browsing video corresponding to the browsing object, and browsing behavior description information.

[0054] Step 202, according to the browsing behavior description information, determine the satisfied sliding label corresponding to the browsing video behavior record; the satisfied sliding label indicates whether the browsing object has satisfied sliding behavior for browsing the video.

[0055] Step 203: Determine whether there is a specified interactive behavior in the browsing behavior description information.

[0056] Step 204, when there is a designated interactive behavior, the initial weight is adjusted according to the weight deviation value corresponding to the designated interactive behavior to obtain the weight corresponding to the browsing video behavior record.

[0057] In the embodiment of the present disclosure, the process of the electronic device executing step 204 may, for example, be to determine the sum of the initial weight and the weight deviation value corresponding to the specified interactive behavior; and use the sum as the weight corresponding to the browsing video behavior record.

[0058] The specified interactive behavior may include at least one of the following: video like behavior, video evaluation behavior, video follow behavior, video fast scrolling behavior, video exit behavior, and video long browsing behavior. In the video fast scrolling behavior, the browsing time of the browsing object for browsing the video is less than or equal to the second time threshold; in the video long browsing behavior, the browsing time of the browsing object for browsing the video is greater than or equal to the first time threshold.

[0059] Among them, the weight deviation value corresponding to the video like behavior, the weight deviation value corresponding to the video comment behavior, the weight deviation value corresponding to the video follow behavior, and the weight deviation value corresponding to the video long browsing behavior are positive values; the weight deviation value corresponding to the video fast scrolling behavior and the weight deviation value corresponding to the video exit behavior are negative values.

[0060] Among them, the weight corresponding to the browsing video behavior record is determined in combination with the weight deviation value corresponding to the specified interactive behavior and the specified interactive behavior existing in the browsing behavior description information. Different weights can be determined for different situations of the browsing behavior description information, so that the weight can reflect the relevant situations of the interactive behavior in the browsing behavior description information, thereby improving the accuracy of the determined training data.

[0061] Among them, for the video like behavior, video comment behavior and video follow behavior in the interactive behavior, the weight of the browsing video behavior records is increased, so that the satisfaction decline model can focus on training this part of the browsing video behavior records; for the video fast scroll behavior and video exit behavior in the interactive behavior, the weight of the browsing video behavior records is reduced, so that the satisfaction decline model can non-focus on training this part of the browsing video behavior records, thereby further improving the accuracy of the trained satisfaction decline model.

[0062] Step 205 , determining training data according to the first attribute information and the second attribute information in the plurality of video browsing behavior records, and the satisfaction slide labels and weights corresponding to the video browsing behavior records.

[0063] Step 206: Use the training data to train the initial satisfaction decline model to obtain a trained satisfaction decline model for video recommendation processing.

[0064] It should be noted that the details of step 201 and step 202 can be found in Figure 1 Step 101 and step 102 in the illustrated embodiment will not be described in detail herein.

[0065] The training method of the satisfied decline model of the embodiment of the present disclosure is through obtaining browsing video behavior records; the browsing video behavior records include the first attribute information of the browsing object, the second attribute information of the browsing video corresponding to the browsing object, and the browsing behavior description information; according to the browsing behavior description information, the satisfied decline label corresponding to the browsing video behavior record is determined; the satisfied decline label indicates whether the browsing object has satisfied decline behavior on the browsing video; whether there is a specified interactive behavior in the browsing behavior description information; in the case of the existence of the specified interactive behavior, the initial weight is adjusted according to the weight deviation value corresponding to the specified interactive behavior to obtain the weight corresponding to the browsing video behavior record; the training data is determined according to the first attribute information and the second attribute information in a plurality of browsing video behavior records, and the satisfied decline label and weight corresponding to the browsing video behavior records; the training data is used to train the initial satisfied decline model to obtain the trained satisfied decline model for video recommendation processing, so that when recommending videos, the probability of the object's satisfied decline on the video can be considered, the video recommendation efficiency can be improved, the continuous consumption of the video can be achieved, and the consumption stickiness of the object on the video can be improved.

[0066] In order to quickly train the satisfactory descent model and improve the training speed of the satisfactory descent model, the satisfactory descent model can be tested in time after the satisfactory descent model is trained. Figure 3 As shown, Figure 3 is a schematic diagram according to a third embodiment of the present disclosure, Figure 3The illustrated embodiment may include the following steps:

[0067] Step 301, obtaining a video browsing behavior record; the video browsing behavior record includes first attribute information of a browsing object, second attribute information of a browsing video corresponding to the browsing object, and browsing behavior description information.

[0068] Step 302, determining the satisfied sliding label corresponding to the video browsing behavior record according to the browsing behavior description information; the satisfied sliding label indicates whether the browsing object has satisfied sliding behavior for browsing the video.

[0069] Step 303: determine training data according to the first attribute information and the second attribute information in the plurality of video browsing behavior records, and the satisfaction slide labels corresponding to the video browsing behavior records.

[0070] Step 304: Acquire test data corresponding to the training data.

[0071] In the embodiment of the present disclosure, the test data may also include the first attribute information and the second attribute information in a plurality of browsing video behavior records, and the satisfied slide label corresponding to the browsing video behavior record. The determination process of the test data may refer to steps 301 to 303. The difference between the test data and the training data is that the browsing video behavior record used in determining the training data is different from the browsing video behavior record used in determining the test data.

[0072] In one example, after the electronic device executes steps 301 to 303 to obtain training data, it can separate a portion of the training data as test data and the rest as training data.

[0073] Step 305: Use the training data to train the initial satisfactory decline model, and use the test data to test the trained satisfactory decline model to obtain the test index value of the satisfactory decline model.

[0074] In the disclosed embodiment, the training process of the satisfaction decline model by the electronic device may be, for example, inputting the first attribute information and the second attribute information in the browsing video behavior record into the initial satisfaction decline model to obtain the satisfaction decline prediction probability output by the satisfaction decline model; combining the satisfaction decline prediction probability, the satisfaction decline label corresponding to the browsing video behavior record, and the loss function of the satisfaction decline model to determine the value of the loss function; and adjusting the parameters of the satisfaction decline model according to the value of the loss function to obtain a trained satisfaction decline model.

[0075] Among them, when the satisfied swipe label indicates that the browsing object has satisfied swipe behavior for browsing the video, the satisfied swipe probability is determined to be 1; when the satisfied swipe label indicates that the browsing object has not satisfied swipe behavior for browsing the video, the satisfied swipe probability is determined to be 0.

[0076] In the disclosed embodiment, the training data may further include at least one of the following: the predicted probability of the browsing object performing a specified interactive behavior on the browsing video in the browsing video behavior record, and the browsing scene information. Where the training data also includes the predicted probability of the browsing object performing a specified interactive behavior on the browsing video in the browsing video behavior record, the input of the satisfaction decline model during the training process may be the first attribute information and the second attribute information in the browsing video behavior record, and the predicted probability of the browsing object performing a specified interactive behavior on the browsing video in the browsing video behavior record.

[0077] In the case where the training data also includes browsing scene information, the input of the satisfaction decline model during the training process may be the first attribute information and the second attribute information in the browsing video behavior record, and the browsing scene information.

[0078] In an embodiment of the present disclosure, when the test data includes the first attribute information and the second attribute information in the browsing video behavior record, and the satisfied decline label, the testing process of the electronic device on the trained satisfied decline model can be, for example, inputting the first attribute information and the second attribute information in the browsing video behavior record into the trained satisfied decline model to obtain the satisfied decline prediction probability output by the satisfied decline model; and determining the test indicator value by combining the satisfied decline prediction probability, the satisfied decline label corresponding to the browsing video behavior record, and the test indicator function of the satisfied decline model.

[0079] The test indicator corresponding to the test indicator value may be, for example, the area under the curve (AUC), etc. Where the test indicator is AUC, the process of determining the test indicator value may be, for example, when the predicted probability of satisfactory decline in the browsing video behavior record is greater than a preset probability threshold, and the satisfactory decline label indicates that there is a satisfactory decline behavior, the count value is increased by 1; when the predicted probability of satisfactory decline in the browsing video behavior record is less than or equal to the preset probability threshold, and the satisfactory decline label indicates that there is no satisfactory decline behavior, the count value is increased by 1; the ratio of the count value to the total number of satisfactory decline indicators in the test data is determined as the test indicator value.

[0080] Step 306, determining whether the training is completed based on multiple test indicator values ​​obtained through multiple test processes.

[0081] In the disclosed embodiment, the test indicator corresponding to the test indicator value may be, for example, an area under a curve (AUC), etc. Wherein, when the test indicator is an AUC indicator, the process of the electronic device executing step 306 may be, for example, determining an absolute change value between adjacent test indicator values ​​among a plurality of test indicator values; determining that the training is completed when there are a first number of consecutive absolute change values ​​that are less than a preset change threshold; and determining that the training is not completed when there are no first number of consecutive absolute change values ​​that are less than the preset change threshold.

[0082] Among them, if there are a first number of consecutive absolute change values ​​that are less than the preset change threshold, it means that the test indicator value is stable, that is, the accuracy of the satisfactory decline model is stable, and it can be determined that the training is completed, so that the training process of the satisfactory decline model can be stopped in time, reducing the training time of the satisfactory decline model and improving the training speed of the satisfactory decline model.

[0083] Step 307: When the training is not completed, repeat the training and testing process.

[0084] In the embodiment of the present disclosure, repeating the training and testing process means repeatedly executing step 305 to step 306.

[0085] Step 308: After the training is completed, the satisfactory descent model obtained by the last training is used as the trained satisfactory descent model.

[0086] It should be noted that the details of step 301 and step 303 can be found in Figure 1 Step 101 and step 103 in the illustrated embodiment will not be described in detail herein.

[0087] The training method of the satisfied decline model of the embodiment of the present disclosure is as follows: obtaining a video browsing behavior record; the video browsing behavior record includes first attribute information of a browsing object, second attribute information of a browsing video corresponding to the browsing object, and browsing behavior description information; determining a satisfied decline label corresponding to the video browsing behavior record according to the browsing behavior description information; the satisfied decline label indicates whether the browsing object has satisfied decline behavior on the browsing video; determining training data according to the first attribute information and the second attribute information in a plurality of video browsing behavior records and the satisfied decline label corresponding to the video browsing behavior records; obtaining test data corresponding to the training data; using the training data to train an initial satisfied decline model, and using the test data to test the trained satisfied decline model to obtain a test index value of the satisfied decline model; determining whether the training is completed according to a plurality of test index values ​​obtained by multiple test processes; repeating the training and test processes when the training is not completed; after the training is completed, using the satisfied decline model obtained by the last training as the trained satisfied decline model, so that when recommending videos, the probability of satisfied decline of the object to the video can be considered, the efficiency of video recommendation can be improved, the continuous consumption of the video can be achieved, and the consumption stickiness of the object to the video can be improved.

[0088] Figure 4 4 is a schematic diagram according to the fourth embodiment of the present disclosure. It should be noted that the video recommendation method of the embodiment of the present disclosure can be applied to a video recommendation device, which can be configured in an electronic device so that the electronic device can perform a video recommendation function. In the following embodiments, the execution subject is an electronic device as an example for description.

[0089] Among them, the electronic device can be any device with computing capabilities, such as a personal computer (PC), a mobile terminal, a server, etc. The mobile terminal can be, for example, a vehicle-mounted device, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, a smart speaker, and other hardware devices with various operating systems, touch screens and / or display screens.

[0090] The electronic device may be provided with software for providing videos for browsing, such as video software, etc. The video recommendation device may also be the software provided in the electronic device.

[0091] like Figure 4 As shown, the video recommendation method may include the following steps:

[0092] Step 401, obtaining first attribute information of a target object and second attribute information of a plurality of candidate videos.

[0093] In the embodiment of the present disclosure, the target object, for example, a user who needs to browse videos, etc. The first attribute information of the target object, for example, the age, gender, education level, etc. of the target object.

[0094] The multiple candidate videos may be videos recalled by combining all or part of the first attribute information of the target object; or may be videos obtained by roughly sorting and stage-processing the recalled videos according to other sorting indicators. The other sorting indicators may include part or all of the first attribute information, and information determined based on the first attribute information and the second attribute information.

[0095] Step 402, for each candidate video, input the second attribute information and the first attribute information of the candidate video into a preset satisfaction decline model to obtain the satisfaction decline probability output by the satisfaction decline model; the satisfaction decline model is trained by combining the first attribute information of the browsing object in multiple browsing video behavior records, the second attribute information of the browsing video corresponding to the browsing object, and the satisfaction decline label corresponding to the browsing video behavior record.

[0096] In an embodiment of the present disclosure, in one example, the electronic device may further perform the following process: determining the predicted probability of the target object performing a specified interactive behavior on the candidate video based on the second attribute information and the first attribute information of the candidate video. Correspondingly, the process of the electronic device performing step 402 may, for example, be inputting the second attribute information, the first attribute information of the candidate video, and the predicted probability of the target object performing a specified interactive behavior on the candidate video into the satisfaction decline model to obtain the satisfaction decline probability.

[0097] The predicted probability of the target object performing a specified interactive behavior on the candidate video can be obtained by other models. Other models include, for example, a like prediction model, a review prediction model, a follow prediction model, etc. The input of the above model can at least include the first attribute information of the target object and the second attribute information of the candidate video.

[0098] The specified interactive behavior may include at least one of the following: video like behavior, video evaluation behavior, video follow behavior, video fast scrolling behavior, video exit behavior, and video long browsing behavior. In the video fast scrolling behavior, the browsing time of the browsing object for browsing the video is less than or equal to the second time threshold; in the video long browsing behavior, the browsing time of the browsing object for browsing the video is greater than or equal to the first time threshold.

[0099] Among them, the predicted probability of the target object performing a specified interactive behavior on the candidate video is input into the satisfaction decline model, and the satisfaction decline probability prediction processing is performed. This can enable the satisfaction decline model to consider more information when predicting, improve the accuracy of the predicted satisfaction decline probability, and further improve the efficiency of video recommendation.

[0100] In an embodiment of the present disclosure, in one example, the electronic device may further perform the following process: obtaining browsing scene information of the target object, the browsing scene information indicating the video browsing batch of the target object. Correspondingly, the process of the electronic device performing step 402 may be, for example, inputting the second attribute information, the first attribute information and the browsing scene information of the candidate video into the satisfaction decline model to obtain the satisfaction decline probability.

[0101] The video browsing batch refers to the batch of videos browsed by the target object in the software. For example, after the software is started, the first batch of videos displayed are the first-viewed videos; after the first batch of videos are browsed and scrolled, the second batch of videos displayed are the second-viewed videos, and so on.

[0102] Among them, inputting browsing scene information into the satisfaction decline model and performing satisfaction decline probability dependency processing can enable the satisfaction decline model to consider more information when predicting, improve the accuracy of the predicted satisfaction decline probability, and further improve the efficiency of video recommendation.

[0103] Step 403 , selecting a video to be recommended from the multiple candidate videos based on the satisfactory decline probabilities of the multiple candidate videos.

[0104] In an embodiment of the present disclosure, in one example, an electronic device may sort multiple candidate videos according to their satisfactory decline probabilities to obtain a sorting result; and use a preset number of videos ranked at the top of the sorting result as videos to be recommended.

[0105] In another example, the electronic device can perform a comprehensive sorting process on the multiple candidate videos according to their satisfactory decline probabilities and other indicators to obtain a sorting result; and use a preset number of videos ranked first in the sorting result as the videos to be recommended. Among them, other indicators, such as part or all of the information in the first attribute information, part or all of the information in the second attribute information, etc., are not specifically limited here and can be set according to actual needs.

[0106] Step 404: recommend the video to be recommended to the target object.

[0107] In the disclosed embodiment, the electronic device may display the recommended video to facilitate browsing by the target object.

[0108] The video recommendation method of the embodiment of the present disclosure obtains the first attribute information of the target object and the second attribute information of multiple candidate videos; for each candidate video, inputs the second attribute information and the first attribute information of the candidate video into a preset satisfaction decline model to obtain the satisfaction decline probability output by the satisfaction decline model; the satisfaction decline model is obtained by training by combining the first attribute information of the browsing object in multiple browsing video behavior records, the second attribute information of the browsing video corresponding to the browsing object, and the satisfaction decline label corresponding to the browsing video behavior records; combining the satisfaction decline probabilities of multiple candidate videos, a video to be recommended is selected from the multiple candidate videos; the video to be recommended is recommended to the target object, so that when recommending the video, the object's satisfaction decline probability for the video can be considered, thereby improving the efficiency of video recommendation, achieving continuous consumption of the video, and improving the object's consumption stickiness to the video.

[0109] In order to implement the above embodiment, the present disclosure also provides a training device for a satisfactory descent model. Figure 5 As shown, Figure 5 The fifth embodiment of the present disclosure is a schematic diagram of the training device 50 of the satisfactory descent model, which may include: an acquisition module 501 , a first determination module 502 , a second determination module 503 and a training processing module 504 .

[0110] Among them, the acquisition module 501 is used to obtain the browsing video behavior record; the browsing video behavior record includes the first attribute information of the browsing object, the second attribute information of the browsing video corresponding to the browsing object, and the browsing behavior description information; the first determination module 502 is used to determine the satisfaction slide label corresponding to the browsing video behavior record according to the browsing behavior description information; the satisfaction slide label indicates whether the browsing object has a satisfactory slide behavior for the browsing video; the second determination module 503 is used to determine the training data according to the first attribute information and the second attribute information in multiple browsing video behavior records, and the satisfaction slide label corresponding to the browsing video behavior record; the training processing module 504 is used to use the training data to train the initial satisfaction slide model to obtain the trained satisfaction slide model for video recommendation processing.

[0111] As a possible implementation method of the embodiment of the present disclosure, the browsing behavior description information includes browsing time and whether there is a post-browsing swiping behavior; the first determination module 502 is specifically used to, when the browsing time is greater than or equal to a first time threshold and there is a post-browsing swiping behavior, determine that the value of the satisfactory swiping label corresponding to the browsing video behavior record is a first value; the first value indicates that the browsing object has a satisfactory swiping behavior for browsing the video; when the browsing time is less than the first time threshold, or there is no post-browsing swiping behavior, determine that the value of the satisfactory swiping label corresponding to the browsing video behavior record is a second value; the second value indicates that the browsing object has no satisfactory swiping behavior for the video.

[0112] As a possible implementation method of the embodiment of the present disclosure, the browsing behavior description information includes browsing time, whether there is a sliding behavior after browsing, and whether there is an exit behavior after browsing; the first determination module 502 is specifically used to, when the browsing time is greater than or equal to a first time threshold and there is a sliding behavior after browsing, determine that the value of the satisfactory sliding label corresponding to the browsing video behavior record is a first value; the first value indicates that the browsing object has a satisfactory sliding behavior on the browsing video; when the browsing time is less than the first time threshold and there is a sliding behavior after browsing, or when the browsing time is less than or equal to the second time threshold and there is an exit behavior after browsing, determine that the value of the satisfactory sliding label corresponding to the browsing video behavior record is a second value; the second value indicates that the browsing object has no satisfactory sliding behavior on the video; the second time threshold is less than the first time threshold.

[0113] As a possible implementation method of the embodiment of the present disclosure, the device also includes: a third determination module, used to determine the weight corresponding to the browsing video behavior record according to the browsing behavior description information; the second determination module 503 is specifically used to determine the training data according to the first attribute information and the second attribute information in multiple browsing video behavior records, and the satisfaction slide label and weight corresponding to the browsing video behavior record.

[0114] As a possible implementation method of an embodiment of the present disclosure, the third determination module is specifically used to determine whether there is a specified interactive behavior in the browsing behavior description information; when there is a specified interactive behavior, the initial weight is adjusted according to the weight deviation value corresponding to the specified interactive behavior to obtain the weight corresponding to the browsing video behavior record.

[0115] As a possible implementation method of the embodiment of the present disclosure, the third determination module is further specifically used to determine the sum of the initial weight and the weight deviation value corresponding to the specified interactive behavior; and use the sum as the weight corresponding to the browsing video behavior record.

[0116] As a possible implementation method of an embodiment of the present disclosure, the specified interactive behavior includes at least one of the following: video like behavior, video evaluation behavior, video attention behavior, video fast scrolling behavior, video exit behavior, and video long browsing behavior; in the video fast scrolling behavior, the browsing time of the browsing object for browsing the video is less than or equal to a second time threshold; in the long video browsing behavior, the browsing time of the browsing object for browsing the video is greater than or equal to a first time threshold; the weight deviation value corresponding to the video like behavior, the weight deviation value corresponding to the video evaluation behavior, the weight deviation value corresponding to the video attention behavior, and the weight deviation value corresponding to the long video browsing behavior are positive values; the weight deviation value corresponding to the video fast scrolling behavior and the weight deviation value corresponding to the video exit behavior are negative values.

[0117] As a possible implementation of the embodiment of the present disclosure, the device also includes: a fourth determination module, used to determine the predicted probability of the browsing object performing a specified interactive behavior on the browsing video based on the first attribute information and the second attribute information; the second determination module 503 is specifically used to determine the training data based on the first attribute information and the second attribute information in multiple browsing video behavior records, the predicted probability of the browsing object performing a specified interactive behavior on the browsing video in the browsing video behavior records, and the satisfied slide label corresponding to the browsing video behavior record.

[0118] As a possible implementation method of an embodiment of the present disclosure, the browsing video behavior record also includes browsing scene information indicating a video browsing batch of the browsing object; the second determination module 503 is specifically used to determine the training data based on the first attribute information, the second attribute information and the browsing scene information in multiple browsing video behavior records, and the satisfaction decline label and weight corresponding to the browsing video behavior records.

[0119] As a possible implementation method of the embodiment of the present disclosure, the training processing module 504 is specifically used to obtain test data corresponding to the training data; use the training data to train the initial satisfaction decline model, and use the test data to test the trained satisfaction decline model to obtain test index values ​​of the satisfaction decline model; determine whether the training is completed based on multiple test index values ​​obtained by multiple test processes; when the training is not completed, repeat the training and testing process; after the training is completed, use the satisfaction decline model obtained from the last training as the trained satisfaction decline model.

[0120] As a possible implementation method of the embodiment of the present disclosure, the test indicator corresponding to the test indicator value is the area under the curve AUC indicator; the training processing module 504 is also specifically used to determine the absolute change value between adjacent test indicator values ​​among multiple test indicator values; when there are a first number of consecutive absolute change values ​​that are less than a preset change threshold, it is determined that the training is completed; when there are not a first number of consecutive absolute change values ​​that are less than the preset change threshold, it is determined that the training is not completed.

[0121] As a possible implementation of the embodiment of the present disclosure, the satisfied descent model is a model based on the extreme gradient boosting xgboost algorithm.

[0122] The training device of the satisfied decline model of the embodiment of the present disclosure obtains the browsing video behavior record; the browsing video behavior record includes the first attribute information of the browsing object, the second attribute information of the browsing video corresponding to the browsing object, and the browsing behavior description information; according to the browsing behavior description information, the satisfied decline label corresponding to the browsing video behavior record is determined; the satisfied decline label indicates whether the browsing object has satisfied decline behavior on the browsing video; according to the first attribute information and the second attribute information in a plurality of browsing video behavior records, and the satisfied decline label corresponding to the browsing video behavior record, the training data is determined; the initial satisfied decline model is trained with the training data to obtain the trained satisfied decline model for video recommendation processing, so that when recommending videos, the probability of satisfied decline of the object on the video can be considered, the video recommendation efficiency can be improved, the continuous consumption of the video can be realized, and the consumption stickiness of the object on the video can be improved.

[0123] In order to implement the above embodiment, the present disclosure also provides a video recommendation device. Figure 6 As shown, Figure 6 The sixth embodiment of the present disclosure is shown in FIG. 6. The video recommendation device 60 may include: a first acquisition module 601 , a second acquisition module 602 , a selection module 603 and a recommendation module 604 .

[0124] Among them, the first acquisition module 601 is used to obtain the first attribute information of the target object and the second attribute information of multiple candidate videos; the second acquisition module 602 is used to input the second attribute information of the candidate video and the first attribute information into a preset satisfaction decline model for each candidate video, and obtain the satisfaction decline probability output by the satisfaction decline model; the satisfaction decline model is trained by combining the first attribute information of the browsing object in multiple browsing video behavior records, the second attribute information of the browsing video corresponding to the browsing object, and the satisfaction decline label corresponding to the browsing video behavior records; the selection module 603 is used to select a video to be recommended from multiple candidate videos in combination with the satisfaction decline probabilities of multiple candidate videos; the recommendation module 604 is used to recommend the video to be recommended to the target object.

[0125] As a possible implementation method of the embodiment of the present disclosure, the device also includes: a determination module, used to determine the predicted probability of the target object performing a specified interactive behavior on the candidate video based on the second attribute information of the candidate video and the first attribute information; the second acquisition module 602 is specifically used to input the second attribute information of the candidate video, the first attribute information and the predicted probability of the target object performing a specified interactive behavior on the candidate video into the satisfaction decline model to obtain the satisfaction decline probability.

[0126] As a possible implementation method of the embodiment of the present disclosure, the device also includes: a third acquisition module, used to obtain browsing scene information of the target object, and the browsing scene information indicates a video browsing batch of the target object; the second acquisition module 602 is specifically used to input the second attribute information, the first attribute information and the browsing scene information of the candidate video into the satisfaction decline model to obtain the satisfaction decline probability.

[0127] The video recommendation device of the embodiment of the present disclosure obtains the first attribute information of the target object and the second attribute information of multiple candidate videos; for each candidate video, inputs the second attribute information and the first attribute information of the candidate video into a preset satisfaction decline model to obtain the satisfaction decline probability output by the satisfaction decline model; the satisfaction decline model is obtained by training by combining the first attribute information of the browsing object in multiple browsing video behavior records, the second attribute information of the browsing video corresponding to the browsing object, and the satisfaction decline label corresponding to the browsing video behavior records; selects a video to be recommended from the multiple candidate videos in combination with the satisfaction decline probabilities of the multiple candidate videos; recommends the video to be recommended to the target object, so that when recommending the video, the object's satisfaction decline probability for the video can be considered, thereby improving the efficiency of video recommendation, achieving continuous consumption of the video, and improving the object's consumption stickiness to the video.

[0128] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information are all carried out with the user's consent, comply with the relevant laws and regulations, and do not violate public order and good morals.

[0129] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0130] Figure 7 A schematic block diagram of an example electronic device 700 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0131] like Figure 7 As shown, the device 700 includes a computing unit 701, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0132] A number of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0133] The computing unit 701 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 701 performs the various methods and processes described above, such as the training method of the satisfactory descent model or the video recommendation method. For example, in some embodiments, the training method of the satisfactory descent model or the video recommendation method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the training method of the satisfactory descent model or the video recommendation method described above may be executed. Alternatively, in other embodiments, the computing unit 701 may be configured to execute the training method of the satisfaction decline model or the video recommendation method in any other appropriate manner (for example, by means of firmware).

[0134] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0135] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0136] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0137] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0138] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0139] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0140] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0141] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for training a satisfaction decline model, the method comprising: Obtain video browsing behavior records; The video browsing behavior record includes first attribute information of the browsing object, second attribute information of the browsing video corresponding to the browsing object, and browsing behavior description information; Determine, according to the browsing behavior description information, a satisfaction slide label corresponding to the browsing video behavior record; The satisfied slide tag indicates whether the browsing object has satisfied slide behavior when browsing the video; wherein the browsing behavior description information includes browsing time and whether there is slide behavior after browsing; Determining training data according to first attribute information and second attribute information in a plurality of video browsing behavior records, and satisfaction slide labels corresponding to the video browsing behavior records; The training data is used to train the initial satisfaction decline model to obtain a trained satisfaction decline model for video recommendation processing.

2. The method according to claim 1, Determining the satisfactory slide label corresponding to the video browsing behavior record according to the browsing behavior description information includes: When the browsing time is greater than or equal to the first time threshold and there is a sliding behavior after browsing, determining that the value of the satisfied sliding label corresponding to the browsing video behavior record is a first value; the first value indicates that the browsing object has a satisfied sliding behavior for browsing the video; When the browsing time is less than the first time threshold, or there is no sliding behavior after browsing, the value of the satisfied sliding label corresponding to the browsing video behavior record is determined to be a second value; the second value indicates that the browsing object has no satisfied sliding behavior on the video.

3. The method according to claim 1, wherein: The browsing behavior description information includes browsing time, whether there is a swiping behavior after browsing, and whether there is an exit behavior after browsing; Determining the satisfactory slide label corresponding to the video browsing behavior record according to the browsing behavior description information includes: When the browsing time is greater than or equal to the first time threshold and there is a sliding behavior after browsing, determining that the value of the satisfied sliding label corresponding to the browsing video behavior record is a first value; the first value indicates that the browsing object has a satisfied sliding behavior for browsing the video; When the browsing time is less than the first time threshold and there is a swiping behavior after browsing, or when the browsing time is less than or equal to the second time threshold and there is an exit behavior after browsing, it is determined that the value of the satisfied swiping label corresponding to the browsing video behavior record is the second value; the second value indicates that the browsing object has no satisfied swiping behavior on the video; the second time threshold is less than the first time threshold.

4. The method according to claim 1, wherein: The method further comprises: Determining a weight corresponding to the video browsing behavior record according to the browsing behavior description information; The determining of training data according to the first attribute information and the second attribute information in the plurality of video browsing behavior records and the satisfaction slide labels corresponding to the video browsing behavior records includes: The training data is determined according to first attribute information and second attribute information in a plurality of video browsing behavior records, and satisfaction slide labels and weights corresponding to the video browsing behavior records.

5. The method according to claim 4, wherein: Determining the weight corresponding to the video browsing behavior record according to the browsing behavior description information includes: Determining whether a specified interactive behavior exists in the browsing behavior description information; In the case of a specified interactive behavior, the initial weight is adjusted according to the weight deviation value corresponding to the specified interactive behavior to obtain the weight corresponding to the video browsing behavior record.

6. The method according to claim 5, wherein: The adjusting the initial weight according to the weight deviation value corresponding to the specified interactive behavior to obtain the weight corresponding to the browsing video behavior record includes: Determine a sum of the initial weight and the weight deviation value corresponding to the specified interaction behavior; The summed result is used as the weight corresponding to the video browsing behavior record.

7. The method according to claim 5 or 6, wherein: The specified interactive behavior includes at least one of the following: video like behavior, video evaluation behavior, video follow behavior, video fast scrolling behavior, video exit behavior, video long browsing behavior; In the video fast scrolling behavior, the browsing time of the browsing object for browsing the video is less than or equal to the second time threshold; in the video long browsing behavior, the browsing time of the browsing object for browsing the video is greater than or equal to the first time threshold; The weight deviation value corresponding to the video like behavior, the weight deviation value corresponding to the video evaluation behavior, the weight deviation value corresponding to the video attention behavior, and the weight deviation value corresponding to the video long browsing behavior are positive values; The weight deviation value corresponding to the video fast-swiping behavior and the weight deviation value corresponding to the video exit behavior are negative values.

8. The method according to claim 1 or 4, wherein: The method further comprises: Determining, according to the first attribute information and the second attribute information, a predicted probability that the browsing object performs a specified interactive behavior on the browsing video; The determining of training data according to the first attribute information and the second attribute information in the plurality of video browsing behavior records and the satisfaction slide labels corresponding to the video browsing behavior records includes: The training data is determined based on first attribute information and second attribute information in multiple browsing video behavior records, predicted probabilities of browsing objects performing specified interactive behaviors on browsing videos in the browsing video behavior records, and satisfied slide labels corresponding to the browsing video behavior records.

9. The method according to claim 4, wherein: The video browsing behavior record also includes browsing scene information indicating the video browsing batch of the browsing object; The determining of training data according to the first attribute information and the second attribute information in the plurality of video browsing behavior records and the satisfaction slide labels corresponding to the video browsing behavior records includes: The training data is determined according to the first attribute information, the second attribute information and the browsing scene information in a plurality of browsing video behavior records, and the satisfaction slide labels and weights corresponding to the browsing video behavior records.

10. The method according to claim 1, wherein: The training data is used to train the initial satisfactory decline model to obtain a trained satisfactory decline model, including: Obtaining test data corresponding to the training data; The training data is used to train the initial satisfaction decline model, and the test data is used to test the trained satisfaction decline model to obtain the test index value of the satisfaction decline model; Determine whether the training is completed based on multiple test indicator values ​​obtained through multiple test processing; When training is not completed, repeat the training and testing process; After the training is completed, the satisfactory descent model obtained by the last training is used as the trained satisfactory descent model.

11. The method according to claim 10, wherein: The test indicator corresponding to the test indicator value is the area under the curve AUC indicator; Determining whether the training is completed according to the multiple test index values ​​obtained by the multiple test processes includes: Determine the absolute change value between adjacent test indicator values ​​among multiple test indicator values; In the case where there are a first number of consecutive absolute change values ​​that are smaller than a preset change threshold, determining that the training is completed; If there are not a first number of consecutive absolute change values ​​that are smaller than the preset change threshold, it is determined that the training is not completed.

12. The method according to claim 1, wherein: The satisfactory decline model is a model based on the extreme gradient boosting xgboost algorithm.

13. A video recommendation method, the method comprising: Acquire first attribute information of a target object and second attribute information of a plurality of candidate videos; For each candidate video, input the second attribute information and the first attribute information of the candidate video into a preset satisfaction decline model to obtain the satisfaction decline probability output by the satisfaction decline model; the satisfaction decline model is obtained by training in combination with the method described in any one of claims 1 to 12; Combining the satisfaction decline probabilities of multiple candidate videos, selecting a video to be recommended from the multiple candidate videos; Recommend the video to be recommended to the target object.

14. The method according to claim 13, wherein: The method further comprises: Determining, based on the second attribute information of the candidate video and the first attribute information, a predicted probability that the target object performs a specified interactive behavior on the candidate video; The step of inputting the second attribute information and the first attribute information of the candidate video into a preset satisfactory decline model to obtain the satisfactory decline probability output by the satisfactory decline model includes: The second attribute information of the candidate video, the first attribute information, and the predicted probability of the target object performing a specified interactive behavior on the candidate video are input into the satisfaction decline model to obtain the satisfaction decline probability.

15. The method according to claim 13 or 14, wherein: The method further comprises: Acquire browsing scene information of the target object, where the browsing scene information indicates a video browsing batch of the target object; The step of inputting the second attribute information and the first attribute information of the candidate video into a preset satisfactory decline model to obtain the satisfactory decline probability output by the satisfactory decline model includes: The second attribute information, the first attribute information and the browsing scene information of the candidate video are input into the satisfactory decline model to obtain the satisfactory decline probability.

16. A training device for a satisfactory descent model, the device comprising: The acquisition module is used to obtain the browsing video behavior records; The video browsing behavior record includes first attribute information of the browsing object, second attribute information of the browsing video corresponding to the browsing object, and browsing behavior description information; A first determination module is used to determine a satisfactory slide label corresponding to the video browsing behavior record according to the browsing behavior description information; The satisfied slide tag indicates whether the browsing object has satisfied slide behavior when browsing the video; wherein the browsing behavior description information includes browsing time and whether there is slide behavior after browsing; A second determination module is used to determine training data according to first attribute information and second attribute information in a plurality of browsing video behavior records, and a satisfaction slide label corresponding to the browsing video behavior records; The training processing module is used to use the training data to train the initial satisfaction decline model to obtain a trained satisfaction decline model for video recommendation processing.

17. The device according to claim 16, wherein: The browsing behavior description information includes browsing time and whether there is a post-browsing swiping behavior; The first determination module is specifically configured to: When the browsing time is greater than or equal to the first time threshold and there is a sliding behavior after browsing, determining that the value of the satisfied sliding label corresponding to the browsing video behavior record is a first value; the first value indicates that the browsing object has a satisfied sliding behavior for browsing the video; When the browsing time is less than the first time threshold, or there is no sliding behavior after browsing, the value of the satisfied sliding label corresponding to the browsing video behavior record is determined to be a second value; the second value indicates that the browsing object has no satisfied sliding behavior on the video.

18. The device according to claim 16, wherein: The browsing behavior description information includes browsing time, whether there is a sliding behavior after browsing, and whether there is an exit behavior after browsing; the first determination module is specifically used to: When the browsing time is greater than or equal to the first time threshold and there is a sliding behavior after browsing, determining that the value of the satisfied sliding label corresponding to the browsing video behavior record is a first value; the first value indicates that the browsing object has a satisfied sliding behavior for browsing the video; When the browsing time is less than the first time threshold and there is a swiping behavior after browsing, or when the browsing time is less than or equal to the second time threshold and there is an exit behavior after browsing, it is determined that the value of the satisfied swiping label corresponding to the browsing video behavior record is a second value; the second value indicates that the browsing object has no satisfied swiping behavior on the video; The second duration threshold is smaller than the first duration threshold.

19. The device according to claim 16, wherein: The device further comprises: a third determination module, configured to determine a weight corresponding to the browsing video behavior record according to the browsing behavior description information; The second determination module is specifically configured to determine the training data according to the first attribute information and the second attribute information in a plurality of video browsing behavior records, and the satisfaction slide labels and weights corresponding to the video browsing behavior records.

20. The device according to claim 19, wherein The third determination module is specifically used to: Determining whether a specified interactive behavior exists in the browsing behavior description information; In the case of a specified interactive behavior, the initial weight is adjusted according to the weight deviation value corresponding to the specified interactive behavior to obtain the weight corresponding to the video browsing behavior record.

21. The device according to claim 20, wherein: The third determination module is further specifically configured to: Determine a sum of the initial weight and the weight deviation value corresponding to the specified interaction behavior; The summed result is used as the weight corresponding to the video browsing behavior record.

22. The device according to claim 20 or 21, wherein: The specified interactive behavior includes at least one of the following: video like behavior, video evaluation behavior, video follow behavior, video fast scrolling behavior, video exit behavior, video long browsing behavior; In the video fast scrolling behavior, the browsing time of the browsing object for browsing the video is less than or equal to the second time threshold; in the video long browsing behavior, the browsing time of the browsing object for browsing the video is greater than or equal to the first time threshold; The weight deviation value corresponding to the video like behavior, the weight deviation value corresponding to the video evaluation behavior, the weight deviation value corresponding to the video attention behavior, and the weight deviation value corresponding to the video long browsing behavior are positive values; The weight deviation value corresponding to the video fast-swiping behavior and the weight deviation value corresponding to the video exit behavior are negative values.

23. The device according to claim 16 or 19, wherein: The device further includes: a fourth determination module, configured to determine, based on the first attribute information and the second attribute information, a predicted probability that the browsing object performs a specified interactive behavior on the browsing video; The second determination module is specifically used to determine the training data based on the first attribute information and the second attribute information in multiple browsing video behavior records, the predicted probability of the browsing object performing a specified interactive behavior on the browsing video in the browsing video behavior records, and the satisfaction slide label corresponding to the browsing video behavior record.

24. The device according to claim 19, wherein The video browsing behavior record also includes browsing scene information indicating the video browsing batch of the browsing object; the second determination module is specifically used to: The training data is determined according to the first attribute information, the second attribute information and the browsing scene information in a plurality of browsing video behavior records, and the satisfaction slide labels and weights corresponding to the browsing video behavior records.

25. The device according to claim 16, wherein: The training processing module is specifically used to: Obtaining test data corresponding to the training data; The training data is used to train the initial satisfaction decline model, and the test data is used to test the trained satisfaction decline model to obtain the test index value of the satisfaction decline model; Determine whether the training is completed based on multiple test indicator values ​​obtained through multiple test processing; When training is not completed, repeat the training and testing process; After the training is completed, the satisfactory descent model obtained by the last training is used as the trained satisfactory descent model.

26. The device according to claim 25, wherein The test indicator corresponding to the test indicator value is the area under the curve AUC indicator; the training processing module is also specifically used for: Determine the absolute change value between adjacent test indicator values ​​among multiple test indicator values; In the case where there are a first number of consecutive absolute change values ​​that are smaller than a preset change threshold, determining that the training is completed; If there are not a first number of consecutive absolute change values ​​that are smaller than the preset change threshold, it is determined that the training is not completed.

27. The device according to claim 16, wherein: The satisfactory decline model is a model based on the extreme gradient boosting xgboost algorithm.

28. A video recommendation device, comprising: A first acquisition module, used to acquire first attribute information of a target object and second attribute information of a plurality of candidate videos; A second acquisition module is used to input the second attribute information of each candidate video and the first attribute information of the candidate video into a preset satisfaction decline model to obtain the satisfaction decline probability output by the satisfaction decline model; The satisfaction decline model is obtained by training in combination with the method described in any one of claims 1 to 12; A selection module, used to select a video to be recommended from multiple candidate videos by combining the satisfactory decline probabilities of multiple candidate videos; The recommendation module is used to recommend the video to be recommended to the target object.

29. The device according to claim 28, wherein The device further includes: a determination module, configured to determine a predicted probability of the target object performing a specified interactive behavior on the candidate video based on the second attribute information of the candidate video and the first attribute information; The second acquisition module is specifically used to input the second attribute information of the candidate video, the first attribute information and the predicted probability of the target object performing a specified interactive behavior on the candidate video into the satisfaction decline model to obtain the satisfaction decline probability.

30. The device according to claim 28, wherein The device further comprises: a third acquisition module, configured to acquire browsing scene information of the target object, wherein the browsing scene information indicates a video browsing batch of the target object; The second acquisition module is specifically used to input the second attribute information of the candidate video, the first attribute information and the browsing scene information into the satisfaction decline model to obtain the satisfaction decline probability.

31. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 12; Alternatively, the method according to any one of claims 13 to 15 is performed.

32. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 12; or to execute the method according to any one of claims 13 to 15.

33. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements the method according to any one of claims 1 to 12; or executes the method according to any one of claims 13 to 15.

Images