Video content recommendation method and device

Abstract

The invention provides a video content recommendation method and device, and the method comprises the steps: collecting behavior data and equipment state data, and carrying out the preprocessing of the data when a mobile device is detected to play a video; if the behavior data used for indicating that the user is in the video watching state in the moving process exists in the preprocessed behavior data, constructing a feature vector set based on the preprocessed behavior data and the equipment state data; according to a preset feature vector weight, scores of all feature vectors in the feature vector set are calculated and summed, and a comprehensive score is obtained; and if the comprehensive score is higher than or equal to the threshold value, recommending the video content belonging to the same type as the currently played video to the user. The behavior data and the equipment state data of the user are collected in real time, the comprehensive score of the currently played video is calculated according to the behavior data and the equipment state data, the recommended content is dynamically adjusted according to the comprehensive score, the video recommendation accuracy is remarkably improved, the user experience is improved, the user churn rate is reduced, and the user stickiness is improved.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and in particular to a method and device for recommending video content. Background Art

[0002] With the popularization of video streaming services and the continuous improvement of data analysis capabilities, the development of video recommendation methods has become an important means to improve user experience and platform content utilization.

[0003] Current video recommendation methods mainly customize personalized recommendation lists based on users' historical viewing records and static preference information. However, this method has certain limitations in capturing dynamic changes in user interests and real-time preferences, resulting in insufficient accuracy and personalization of recommended content.

[0004] Therefore, how to overcome the limitations of current recommendation methods and improve the accuracy and personalization of recommended content is an urgent problem that needs to be solved. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a method and device for recommending video content to solve the problem that the current recommendation method has low accuracy and personalization.

[0006] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0007] A first aspect of the present invention discloses a method for recommending video content, the method comprising:

[0008] When it is detected that a mobile device is playing a video, the user's behavior data and the device status data of the mobile device are collected and pre-processed;

[0009] If there is behavior data in the preprocessed behavior data indicating that the user is in a video watching state during movement, constructing a feature vector set based on the preprocessed behavior data and the preprocessed device state data, the feature vector set including a behavior feature vector and a device state feature vector;

[0010] According to the preset weights of the feature vectors, the scores of the behavior feature vector and the device state feature vector are calculated and summed to obtain a comprehensive score;

[0011] If the comprehensive score is higher than or equal to the threshold, recommending video content of the same type as the currently playing video to the user;

[0012] If the comprehensive score is lower than a threshold, video content different from the type of the currently playing video is recommended to the user.

[0013] Preferably, the method further comprises:

[0014] The video resources stored in the database are classified using the comprehensive score and the feature vector set, a video recommendation list that meets the comprehensive score is constructed, and the video recommendation list is used as the first recommended video list for starting video playback next time.

[0015] Preferably, the collecting of user behavior data and device status data of the mobile device and preprocessing thereof include:

[0016] Collecting the user's movement data and the user's operation data on the video to obtain behavior data;

[0017] Collecting screen direction change data of the mobile device and playback status data of the video when playing the video, to obtain device status data of the mobile device;

[0018] The behavior data and the device status data are subjected to denoising and smoothing processing to obtain the pre-processed behavior data and the device status data.

[0019] Preferably, after the pretreatment, the method further comprises:

[0020] Determining whether the video is in a continuous playback state according to the preprocessed device status data;

[0021] If the video is in a continuous playback state, determining whether the user is in a continuous moving state based on the pre-processed behavior data;

[0022] If yes, determining that there is behavior data in the preprocessed behavior data indicating that the user is in a video watching state during the moving process;

[0023] If not, it is determined that there is no behavior data in the preprocessed behavior data indicating that the user is in a video watching state during the movement.

[0024] Preferably, after the pretreatment, the method further comprises:

[0025] Determining whether the video is in a continuous playback state according to the preprocessed device status data;

[0026] If the video is in a continuous playback state, whether there is behavior data in the preprocessed behavior data for instructing the user to adjust the playback state of the video;

[0027] If so, determining that the preprocessed behavior data contains behavior data indicating that the user is in a video watching state during the moving process;

[0028] If not, it is determined that there is no behavior data in the preprocessed behavior data for indicating that the user is in a state of watching a video during movement.

[0029] Preferably, after the pretreatment, the method further comprises:

[0030] Determining whether the screen of the mobile device is facing the user according to the preprocessed device status data;

[0031] If yes, determining whether there is data indicating the user's operation on the video in the preprocessed behavior data;

[0032] If so, determining that the preprocessed behavior data contains behavior data indicating that the user is in a video watching state during the moving process;

[0033] If not, it is determined that there is no behavior data in the preprocessed behavior data for indicating that the user is in a state of watching a video during movement.

[0034] Preferably, the step of calculating the scores of the behavior feature vector and the device state feature vector and summing them according to the preset weights of the feature vectors to obtain a comprehensive score includes:

[0035] Calculating a corresponding eigenvalue of the behavior feature vector according to the behavior feature vector, and calculating a corresponding eigenvalue of the device state feature vector according to the device state feature vector;

[0036] Calculating the product of the characteristic value corresponding to the behavior characteristic vector and the preset weight of the behavior characteristic vector to obtain the score of the behavior characteristic vector;

[0037] Calculating the product of the eigenvalue corresponding to the device state feature vector and the preset weight of the device state feature vector to obtain a score of the device state feature vector;

[0038] The score of the behavior feature vector and the score of the device state feature vector are added together to obtain a comprehensive score.

[0039] A second aspect of the present invention discloses a video content recommendation device, the device comprising:

[0040] A collection unit, for collecting user behavior data and device status data of the mobile device and performing preprocessing when detecting that the mobile device plays a video;

[0041] a constructing unit, configured to construct a feature vector set based on the preprocessed behavior data and the preprocessed device state data if there is behavior data indicating that the user is in a video watching state during movement, wherein the feature vector set includes a behavior feature vector and a device state feature vector;

[0042] A calculation unit, configured to calculate scores of the behavior feature vector and the device state feature vector according to preset feature vector weights and sum them up to obtain a comprehensive score;

[0043] A first recommendation unit, configured to recommend video content of the same type as the currently played video to the user if the comprehensive score is higher than or equal to a threshold;

[0044] The second recommendation unit is configured to recommend video content different from the type of the currently played video to the user if the comprehensive score is lower than a threshold.

[0045] Preferably, the device further comprises:

[0046] A list building unit is used to classify the video resources stored in the database using the comprehensive score and the feature vector set, build a video recommendation list that meets the comprehensive score, and use the video recommendation list as the first recommended video list for starting video playback next time.

[0047] Preferably, the collection unit is specifically used for:

[0048] Collect the user's movement data and the user's operation data on the video to obtain behavior data; collect the screen direction change data of the mobile device and the playback status data of the video when playing the video to obtain the device status data of the mobile device; denoise and smooth the behavior data and the device status data to obtain the pre-processed behavior data and the device status data.

[0049] A method and device for recommending video content based on the above-mentioned embodiment of the present invention includes: when a mobile device is detected to be playing a video, collecting behavior data and device status data and performing preprocessing; if there is behavior data in the preprocessed behavior data indicating that the user is in a video watching state during movement, constructing a feature vector set based on the preprocessed behavior data and device status data; according to the preset feature vector weights, calculating the scores of each feature vector in the feature vector set and summing them up to obtain a comprehensive score; if the comprehensive score is higher than or equal to the threshold, recommending to the user video content of the same type as the currently playing video. The user's behavior data and device status data are collected in real time, the comprehensive score of the currently playing video is calculated based on the behavior data and device status data, and the recommended content is dynamically adjusted based on the comprehensive score, which significantly improves the accuracy of video recommendations, improves user experience, reduces user churn rate, and increases user stickiness. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0051] Figure 1 A flowchart of a method for recommending video content provided by an embodiment of the present invention;

[0052] Figure 2 A schematic diagram of data collection and preprocessing provided by an embodiment of the present invention;

[0053] Figure 3 A judgment schematic diagram provided for an embodiment of the present invention;

[0054] Figure 4 A flowchart for judging viewing behavior in a mobile state provided by an embodiment of the present invention;

[0055] Figure 5 A flow chart of the judgment of the operation frequency and type analysis provided by the embodiment of the present invention;

[0056] Figure 6 A flow chart for determining screen direction and interactive behavior provided by an embodiment of the present invention;

[0057] Figure 7 A schematic diagram of a feature vector provided by an embodiment of the present invention;

[0058] Figure 8 This is a structural block diagram of a video content recommendation device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.

[0061] As can be seen from the background technology, the current video recommendation methods have certain limitations in capturing the dynamic changes of user interests and real-time preferences, resulting in insufficient accuracy and personalization of recommended content. Based on this, if combined with the user's actual viewing behavior, it can be further determined whether the user has a high interest in the video content.

[0062] Therefore, an embodiment of the present invention provides a method and device for recommending video content, including: when a mobile device is detected to be playing a video, collecting behavior data and device status data and performing preprocessing; if the preprocessed behavior data contains behavior data indicating that the user is in a video watching state during movement, constructing a feature vector set based on the preprocessed behavior data and device status data; according to the preset feature vector weights, calculating the scores of each feature vector in the feature vector set and summing them up to obtain a comprehensive score; if the comprehensive score is higher than or equal to a threshold, recommending to the user video content of the same type as the currently playing video. The user's behavior data and device status data are collected in real time, the comprehensive score of the currently playing video is calculated based on the behavior data and device status data, and the recommended content is dynamically adjusted based on the comprehensive score, which significantly improves the accuracy of video recommendations, enhances user experience, reduces user churn, and increases user stickiness.

[0063] See also Figure 1 , shows a flow chart of a method for recommending video content provided by an embodiment of the present invention, the method comprising:

[0064] Step S101: When it is detected that a mobile device is playing a video, the user's behavior data and the device status data of the mobile device are collected and pre-processed.

[0065] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0066] In the specific implementation of step S101, when it is detected that a user is playing a video using a mobile device, the built-in sensors and video playback interface of the mobile device are used to collect the user's behavior data and the device status data of the mobile device in real time, and then the collected behavior data and device status data are preprocessed.

[0067] It should be noted that behavior data includes movement data and operation data; device status data includes screen direction change data and playback status data. Figure 2 The data collection and preprocessing schematic diagram is shown, and the specific collection process is as follows (process A1-process A3):

[0068] Process A1: Collect the user's movement data and the user's operation data on the video to obtain behavior data.

[0069] In the specific implementation process A1, the user's movement data is collected, including but not limited to the collection of accelerometer data, gyroscope data and GPS location data. The user's operation data on the video is collected, that is, the user's operation behavior data on the video is collected, and these data (i.e., accelerometer data, gyroscope data, GPS location data and user operation behavior data) are identified as behavior data.

[0070] Process A2: Collect screen direction change data of the mobile device and video playback status data when playing the video, and obtain device status data of the mobile device.

[0071] In the specific implementation process A2, screen direction change data, light sensor data, and video playback status data of the mobile device when playing the video are collected, and these data are identified as device status data of the mobile device.

[0072] Process A3: De-noising and smoothing the behavior data and the device status data to obtain pre-processed behavior data and device status data.

[0073] In the specific implementation process A3, the behavior data and the device status data are subjected to denoising and smoothing processing to reduce the noise in the data and improve the quality and availability of the data, thereby obtaining the preprocessed behavior data and the device status data.

[0074] In some embodiments, after the behavior data and the device status data are preprocessed, it is necessary to label the preprocessed behavior data and the preprocessed device status data, that is, to determine whether there is behavior data in the preprocessed behavior data and the preprocessed device status data that indicates that the user is in a state of watching a video while moving. If so, the preprocessed behavior data and the preprocessed device status data are labeled, such as a "real viewing" label.

[0075] Specific, combined Figure 3 The judgment diagram shown is based on three aspects: "viewing behavior in mobile state", "operation frequency and type analysis" and "screen direction and interactive behavior". The specific judgment process is as follows:

[0076] First, combine Figure 4 The content shown is judged from the perspective of "viewing behavior in mobile state":

[0077] Step S401: Determine whether the video is in a continuous playback state based on the pre-processed device status data. If the video is in a continuous playback state, execute step S402; if the video is not in a continuous playback state, execute step S403. Figure 1 Step S101 is shown.

[0078] It should be noted that if the preprocessed device status data contains device status data indicating that the video is in a continuous playback state, the video is determined to be in a continuous playback state and step S402 is executed; if not, the video is determined not to be in a continuous playback state.

[0079] It is understandable that when it is determined that the video is not in a continuous playback state, it means that the video may have been paused or the user has exited the playback. Figure 1 Step S101 is shown so that when it is detected that a mobile device plays a video, the user's behavior data and the device status data of the mobile device are re-collected and pre-processed.

[0080] Step S402: If the video is in a continuous playback state, determine whether the user is in a continuous moving state based on the preprocessed behavior data. If the user is determined to be in a continuous moving state based on the preprocessed behavior data, execute step S403; if the user is determined not to be in a continuous moving state based on the preprocessed behavior data, execute step S404.

[0081] It should be noted that when it is determined that the video is in a continuous playback state, it is determined whether the user is in a continuous moving state based on the pre-processed behavior data.

[0082] It is understandable that it is possible to determine whether the user is in a continuous moving state based on the accelerometer data, gyroscope data and GPS location data in the preprocessed behavior data. If the user is in a continuous moving state, step S403 is executed; if the user is not in a continuous moving state, step S404 is executed.

[0083] Step S403: If yes, it is determined that there is behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during movement.

[0084] In the process of specifically implementing step S403, if it is determined that the user is in a state of continuous movement, it can be determined that there is behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during the movement.

[0085] In some specific embodiments, after determining that the preprocessed behavior data contains behavior data indicating that the user is in a state of watching a video while moving, the preprocessed behavior data and the preprocessed device state data are labeled, for example, with a "real viewing" label.

[0086] Step S404: If not, it is determined that there is no behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during the movement.

[0087] In the process of specifically implementing step S404, if it is determined that the user is not in a continuous moving state, it is determined that there is no behavior data in the preprocessed behavior data indicating that the user is in a video watching state during the moving process.

[0088] Secondly, combined Figure 5 The content shown is judged from the aspect of "operation frequency and type analysis":

[0089] Step S501: Determine whether the video is in a continuous playback state based on the pre-processed device status data. If the video is in a continuous playback state, execute step S502; if the video is not in a continuous playback state, execute step S503. Figure 1 Step S101 is shown.

[0090] It should be noted that the specific implementation process of step S501 is the same as Figure 4 The specific implementation process of step S401 is consistent and will not be repeated here.

[0091] Step S502: If the video is in a continuous playback state, it is determined whether there is any behavior data in the preprocessed behavior data for instructing the user to adjust the playback state of the video. When there is any behavior data in the preprocessed behavior data for instructing the user to adjust the playback state of the video, step S503 is executed; when there is no behavior data in the preprocessed behavior data for instructing the user to adjust the playback state of the video, step S504 is executed.

[0092] It should be noted that the behavior data of the user adjusting the playback status of the video includes but is not limited to fast forward adjustment, rewind adjustment and volume adjustment of the video.

[0093] Step S503: If yes, it is determined that there is behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during movement.

[0094] In the process of specifically implementing step S503, when it is determined that the video is in a continuous playback state and the preprocessed behavior data contains behavior data for indicating that the user adjusts the playback state of the video, it is determined that the preprocessed behavior data contains behavior data for indicating that the user is in a state of watching the video during movement.

[0095] In some specific embodiments, after determining that the video is in a continuous playback state and the preprocessed behavioral data contains behavioral data for indicating that the user has adjusted the playback state of the video, the preprocessed behavioral data and the preprocessed device state data are labeled, for example, with a "real viewing" label.

[0096] Step S504: If not, it is determined that there is no behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during movement.

[0097] In the process of implementing step S504, when it is determined that the video is in a continuous playback state, but there is no behavioral data in the preprocessed behavioral data for indicating that the user has adjusted the playback state of the video, it indicates that the user may not be actually watching the video (such as the video is continuously playing but the user is not watching the video). Therefore, it can be determined that there is no behavioral data in the preprocessed behavioral data for indicating that the user is in a state of watching the video during movement.

[0098] Finally, combine Figure 6 The content displayed is judged from the aspects of "screen direction and interactive behavior":

[0099] Step S601: Determine whether the screen of the mobile device is facing the user based on the pre-processed device status data. If the screen of the mobile device is facing the user, execute step S602; if the screen of the mobile device is not facing the user, execute step S603. Figure 1 Step S101 shown.

[0100] In the specific implementation of step S601, it is determined whether the screen of the mobile device continues to face the user based on the screen orientation change data and light sensor data in the pre-processed device status data. If it is determined that the screen of the mobile device continues to face the user, step S602 is executed; if it is determined that the screen of the mobile device does not continue to face the user, it indicates that the user may not be actually watching the video. At this point, it can return to execute Figure 1 The illustrated step S101 is to re-collect the user's behavior data and the device status data of the mobile device and perform pre-processing when it is detected that the mobile device is playing a video.

[0101] Step S602: If yes, determine whether there is any operation data indicating the user's operation on the video in the preprocessed behavior data. If there is any operation data indicating the user's operation on the video in the preprocessed behavior data, execute step S603; if there is no operation data indicating the user's operation on the video in the preprocessed behavior data, execute step S604.

[0102] In the specific implementation of step S602, if it is determined that the screen of the mobile device continues to face the user, it is determined whether there is data indicating the user's operation on the video in the pre-processed behavior data.

[0103] It should be noted that the data used to indicate the user's operation on the video includes but is not limited to liking the video, sharing the video, etc.

[0104] Step S603: If yes, it is determined that there is behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during movement.

[0105] In the process of implementing step S603, if the screen of the mobile device continues to face the user, and the preprocessed behavior data contains behavior data indicating the user's operation on the video, it is determined that the preprocessed behavior data contains behavior data indicating that the user is in a state of watching the video during the movement.

[0106] In some specific embodiments, after determining that the screen of the mobile device is continuously facing the user and the preprocessed behavior data contains data indicating the user's operation on the video, the preprocessed behavior data and the preprocessed device status data are labeled, for example, with a "real viewing" label.

[0107] Step S604: If not, it is determined that there is no behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during movement.

[0108] In the specific implementation of step S604, if the preprocessed behavior data does not contain any behavior data indicating the user's operation on the video, it is determined that the preprocessed behavior data does not contain any behavior data indicating that the user is in a state of watching the video during movement.

[0109] Step S102: If the preprocessed behavior data contains behavior data indicating that the user is in a video watching state during movement, a feature vector set is constructed based on the preprocessed behavior data and the preprocessed device state data, and the feature vector set includes a behavior feature vector and a device state feature vector.

[0110] In the process of specifically implementing step S102, when it is determined that the preprocessed behavior data contains behavior data indicating that the user is in a state of watching a video while on the move, a feature vector set is constructed based on the preprocessed behavior data and the preprocessed device state data, wherein the feature vector set includes a behavior feature vector and a device state feature vector, that is, the feature vector set can be used to represent the user's behavior data in a mobile state.

[0111] It should be noted that the feature vector set is constructed based on the user behavior characteristics, video playback status, user operation behavior, screen direction and position change data in the preprocessed behavior data and preprocessed device status data.

[0112] It is understandable that, combined with Figure 7 The characteristic vector diagram shown in the figure, the characteristic vector set (such as Figure 7 The "C") shown in FIG. 1 includes a behavior feature vector and a device state feature vector. The behavior feature vector includes walking feature, operation feature and position feature. The device state feature vector includes playback feature and screen direction feature.

[0113] It can be further understood that walking characteristics include step frequency, stride, direction change, etc.; operation characteristics include operation frequency, operation type, etc.; position characteristics include position change amplitude and frequency, etc.; playback characteristics include playback duration, number of pauses, etc.; screen direction characteristics include light sensor occlusion, screen direction change, etc.

[0114] It should be noted that the data such as step frequency, stride, and direction change in walking characteristics can be determined through the pre-processed accelerometer data, gyroscope data, and GPS location data in the behavior data. The data such as operation frequency and operation type in the operation characteristics can be determined through the pre-processed user operation behavior data on the video in the behavior data. The data such as the amplitude and frequency of position change in the location characteristics can be determined through the pre-processed GPS location data in the behavior data. The data such as the duration of playback and the number of pauses in the playback characteristics can be determined through the pre-processed video playback status data in the device status data. The data such as the screen direction change data and light sensor data in the device status data can be determined through the screen direction characteristics such as the light sensor occlusion and the screen direction change.

[0115] It should be noted that constructing a feature vector set based on the preprocessed behavior data and the preprocessed device status data helps to more accurately capture the user's real viewing behavior, and conduct further analysis and judgment on this basis.

[0116] Step S103: According to the preset weights of the feature vectors, the scores of the behavior feature vector and the device state feature vector are calculated and summed to obtain a comprehensive score.

[0117] It should be noted that the weight of the feature vector includes the preset weight of the behavior feature vector and the preset weight of the device state feature vector. The preset weight of the behavior feature vector includes the preset weight of the walking feature, the preset weight of the operation feature, and the preset weight of the position feature. The preset weight of the device state feature vector includes the preset weight of the playback feature and the preset weight of the screen direction feature.

[0118] It should be noted that the historical behavior data and historical device status data in the database are analyzed in advance by a preset method, and the preset weights of each feature vector are set according to the analysis results. The preset method may be empirical knowledge and domain expert opinions, statistical analysis and data mining technology, or historical data analysis and pattern recognition.

[0119] It can be understood that the weight of each feature is assigned according to its importance, among which the preset weight of the walking feature is relatively high, because the walking feature can directly reflect the user's movement status; the preset weight of the operation feature is relatively high, because the operation feature can directly reflect the user's interaction with the video content; the preset weight of the position feature is relatively low, because the position feature is mainly used to assist in judging the user's environmental status; the preset weight of the playback feature is medium, because the playback feature can reflect the user's viewing status; the preset weight of the screen direction feature is relatively high, because the screen direction feature can directly indicate whether the user is watching the video.

[0120] In the specific implementation of step S103, the weight of the pre-set feature vector is obtained to obtain the preset weight of the behavior feature vector and the preset weight of the device state feature vector. Based on the preset weight of the behavior feature vector and the preset weight of the device state feature vector, the scores of the behavior feature vector and the device state feature vector are calculated and summed to obtain a comprehensive score.

[0121] It can be understood that the specific process of calculating the comprehensive score is as follows (process B1-process B4):

[0122] Process B1: Calculate the eigenvalue corresponding to the behavior feature vector according to the behavior feature vector, and calculate the eigenvalue corresponding to the device state feature vector according to the device state feature vector.

[0123] It should be noted that historical behavior data and historical device status data of the user under different circumstances are obtained in advance, and the historical minimum value and historical maximum value corresponding to each feature vector are determined based on the historical behavior data and historical device status data.

[0124] In the specific implementation process B1, for each behavior feature vector, the current value corresponding to the behavior feature vector, the historical minimum value corresponding to the behavior feature vector, and the historical maximum value corresponding to the behavior feature vector are used to calculate and obtain the feature value corresponding to the current behavior feature vector. For each device state feature vector, the current value corresponding to the device state feature vector, the historical minimum value corresponding to the device state feature vector, and the historical maximum value corresponding to the device state feature vector are used to calculate and obtain the feature value corresponding to the current device state feature vector.

[0125] For example, if the current value of the position feature vector is 5, the historical minimum value of the position feature vector is 2, and the historical maximum value of the position feature vector is 10, (5-2) / (10-2)=0.375 is calculated, and it is determined that the eigenvalue of the position feature vector is 0.375.

[0126] Process B2: Calculate the product of the eigenvalue corresponding to the behavior feature vector and the preset weight of the behavior feature vector to obtain the score of the behavior feature vector.

[0127] In the specific implementation process B2, for each behavior feature vector, a product operation is performed on the corresponding feature value of the behavior feature vector and the preset weight of the behavior feature vector to obtain a score of the behavior feature vector.

[0128] For example, for the cadence feature in the walking feature, if the eigenvalue corresponding to the cadence feature vector is 3, and the corresponding preset weight of the cadence feature vector is 0.3, 3*0.3=0.9 is calculated, and the score of the cadence feature vector is determined to be 0.9.

[0129] For example, for the operation frequency feature in the operation feature, if the corresponding eigenvalue of the operation frequency feature vector is 2, the corresponding preset weight of the operation frequency feature vector is 0.2, and 2*0.2=0.4 is calculated, and the score of the operation frequency feature vector is determined to be 0.4.

[0130] For another example: for the position change amplitude feature in the position feature, if the corresponding eigenvalue of the position change amplitude feature vector is 1, and the corresponding preset weight of the position change amplitude feature vector is 0.1, 1*0.1=0.1 is calculated, and the score of the position change amplitude feature vector is determined to be 0.1.

[0131] Process B3: Calculate the product of the eigenvalue corresponding to the device state feature vector and the preset weight of the device state feature vector to obtain the score of the device state feature vector.

[0132] In the specific implementation process B3, for each device state feature vector, a product operation is performed on the corresponding feature value of the device state feature vector and the preset weight of the device state feature vector to obtain a score of the device state feature vector.

[0133] For example: for the video playback duration feature in the playback feature, if the corresponding eigenvalue of the video playback duration feature vector is 2.5, and the corresponding preset weight of the video playback duration feature vector is 0.25, 2.5*0.25=0.625 is calculated, and the score of the video playback duration feature vector is determined to be 0.625.

[0134] For another example: for the screen direction change feature in the screen direction feature, if the corresponding eigenvalue of the screen direction change feature vector is 1.5, and the corresponding preset weight of the screen direction change feature vector is 0.15, 1.5*0.15=0.225 is calculated, and the score of the screen direction change feature vector is determined to be 0.225.

[0135] Process B4: Add the score of the behavior feature vector and the score of the device state feature vector to obtain a comprehensive score.

[0136] For example, the comprehensive score is equal to the sum of the scores of each behavior feature vector and the scores of each device state feature vector, such as 3×0.3+2.5×0.25+2×0.2+1.5×0.15+1×0.1=2.25.

[0137] Step S104: If the comprehensive score is higher than or equal to the threshold, recommend video content of the same type as the currently playing video to the user.

[0138] In the specific implementation of step S104, the comprehensive score and the threshold are compared. When the comprehensive score is higher than or equal to the threshold (for example, the threshold is 7.0), it indicates that the user is highly interested in the currently playing video. Based on the type of the currently playing video, video content of the same type as the currently playing video is obtained and recommended to the user.

[0139] Step S105: If the comprehensive score is lower than the threshold, recommend video content different from the type of the currently playing video to the user.

[0140] In the specific implementation of step S105, the comprehensive score is compared with the threshold, and when the comprehensive score is lower than the threshold, it indicates that the user is not interested in the currently playing video or is less interested in it. Therefore, according to the type of the currently playing video, video content different from the type of the currently playing video can be obtained and recommended to the user.

[0141] Combined with the above content, in the actual application process, assuming that the mobile device is detected to be in a moving state through the sensor, it is determined that the user is in a moving state. At the same time, the video is determined to be in a playing state through the device status data, and it is detected that the user has performed several fast-forward operations on the video, then it can be determined that the user is in a video watching state during the movement. In addition, the light sensor indicates that the screen is always facing the user, and the GPS data indicates that the user's position is constantly changing but there is no significant movement, indicating that the user may be moving indoors. If the comprehensive score of the currently playing video is higher than or equal to the threshold, the video content of the same type as the currently playing video is recommended to the user.

[0142] In some specific embodiments, the comprehensive score and the feature vector set are further used to classify the video resources stored in the database, and a video recommendation list that meets the comprehensive score is constructed, and the video recommendation list is used as the first recommended video list for the next video playback.

[0143] It should be noted that the classification model is also trained using the comprehensive score and the feature vector set to improve the classification accuracy of the classification model.

[0144] In actual application, the video content in the video recommendation list is adjusted by utilizing the detected real-time location and environment (such as walking, running, riding, etc.) of the user. For example, when it is detected that the user is watching a video while moving, video content that is suitable for short-term viewing or has strong visual appeal is recommended to improve the user's viewing experience.

[0145] In an embodiment of the present invention, when a mobile device is detected to play a video, behavior data and device status data are collected and preprocessed; if there is behavior data in the preprocessed behavior data indicating that the user is in a video watching state during the mobile process, a feature vector set is constructed based on the preprocessed behavior data and device status data, and the feature vector set includes a behavior feature vector and a device status feature vector; according to the weight of the pre-set feature vector, the scores of the behavior feature vector and the device status feature vector are calculated and summed to obtain a comprehensive score; if the comprehensive score is higher than or equal to the threshold, the user is recommended to be recommended video content of the same type as the currently playing video; if the comprehensive score is lower than the threshold, the user is recommended to be recommended video content of a different type from the currently playing video. The present invention collects the user's behavior data and device status data in real time, calculates the comprehensive score of the currently playing video based on the behavior data and the device status data, and dynamically adjusts the recommended content based on the comprehensive score, significantly improving the accuracy of video recommendations, improving user experience, reducing user churn, and increasing user stickiness. Further, the user's viewing time on the platform is increased, thereby increasing the click-through rate of advertisements and platform revenue, and ultimately improving the market competitiveness of the platform.

[0146] Corresponding to the video content recommendation method proposed in the above embodiment of the present invention, see Figure 8 , shows a structural block diagram of a video content recommendation device proposed in an embodiment of the present invention.

[0147] The recommendation device includes: a collection unit 801 , a construction unit 802 , a calculation unit 803 , a first recommendation unit 804 and a second recommendation unit 805 .

[0148] The collection unit 801 is used to collect user behavior data and device status data of the mobile device and perform pre-processing when it is detected that the mobile device plays a video.

[0149] The construction unit 802 is used to construct a feature vector set based on the preprocessed behavior data and the preprocessed device status data if there is behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during movement. The feature vector set includes a behavior feature vector and a device status feature vector.

[0150] The calculation unit 803 is used to calculate the scores of the behavior feature vector and the device state feature vector according to the preset feature vector weights and sum them up to obtain a comprehensive score. If the comprehensive score is higher than or equal to the threshold, the first recommendation unit 804 is executed; if the comprehensive score is lower than the threshold, the second recommendation unit 805 is executed.

[0151] The first recommendation unit 804 is configured to recommend video content of the same type as the currently playing video to the user if the comprehensive score is higher than or equal to a threshold.

[0152] The second recommendation unit 805 is configured to recommend video content different from the type of the currently playing video to the user if the comprehensive score is lower than a threshold.

[0153] In the embodiment of the present invention, the user's behavior data and device status data are collected in real time, the comprehensive score of the currently playing video is calculated based on the behavior data and device status data, and the recommended content is dynamically adjusted based on the comprehensive score, which significantly improves the accuracy of video recommendations, improves user experience, reduces user churn rate, and increases user stickiness. Furthermore, the user's viewing time on the platform is increased, thereby increasing the ad click-through rate and platform revenue, and ultimately improving the platform's market competitiveness.

[0154] Combination Figure 8 The content shown, the recommendation device also includes: a list construction unit, which is used to classify the video resources stored in the database using the comprehensive score and the feature vector set, construct a video recommendation list that meets the comprehensive score, and use the video recommendation list as the first recommended video list for the next video playback.

[0155] Combination Figure 8 As shown in the content, the collection unit 801 is specifically used for: collecting the user's movement data and the user's operation data on the video to obtain behavior data; collecting the screen direction change data of the mobile device when playing the video and the video playback status data to obtain the device status data of the mobile device; denoising and smoothing the behavior data and the device status data to obtain pre-processed behavior data and device status data.

[0156] Combination Figure 8 The content shown, the recommendation device also includes: a first judgment unit, a second judgment unit, a first confirmation unit and a second determination unit.

[0157] The first judgment unit is used to judge whether the video is in a continuous playback state according to the preprocessed device status data.

[0158] The second judgment unit is used to judge whether the user is in a continuous moving state according to the pre-processed behavior data if the video is in a continuous playing state.

[0159] The first confirmation unit is used to determine that if yes, there is behavior data in the preprocessed behavior data indicating that the user is in a video watching state during the movement.

[0160] The second determining unit is configured to, if not, determine that there is no behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during movement.

[0161] Combination Figure 8According to the content shown, the recommendation device also includes: a third judgment unit, a fourth judgment unit, a third determination unit and a fourth determination unit.

[0162] The third judgment unit is used to judge whether the video is in a continuous playback state according to the pre-processed device status data.

[0163] The fourth determination unit is configured to determine whether, if the video is in a continuous playback state, the preprocessed behavior data contains behavior data for instructing the user to adjust the playback state of the video.

[0164] The third determining unit is configured to determine, if any, whether there is behavior data in the preprocessed behavior data indicating that the user is in a state of watching a video during movement.

[0165] The fourth determining unit is used to determine that, if not present, the preprocessed behavior data does not contain behavior data indicating that the user is in a state of watching a video during movement.

[0166] Combination Figure 8 According to the content shown, the recommendation device also includes: a fifth judgment unit, a sixth judgment unit, a fifth determination unit and a sixth determination unit.

[0167] The fifth judgment unit is used to judge whether the screen of the mobile device is facing the user according to the preprocessed device status data.

[0168] The sixth judgment unit is used to judge whether there is data indicating the user's operation on the video in the preprocessed behavior data.

[0169] The fifth determining unit is used to determine whether there is behavior data indicating that the user is in a state of watching a video during movement in the preprocessed behavior data, if any.

[0170] The sixth determining unit is used to determine that, if not present, the preprocessed behavior data does not contain behavior data indicating that the user is in a state of watching a video during movement.

[0171] Combination Figure 8 The content shown, the calculation unit 803, includes: a first calculation module, a second calculation module, a third calculation module and a summation module.

[0172] The first calculation module is used to calculate the eigenvalue corresponding to the behavior feature vector according to the behavior feature vector, and to calculate the eigenvalue corresponding to the device state feature vector according to the device state feature vector.

[0173] The second calculation module is used to calculate the product of the eigenvalue corresponding to the behavior feature vector and the preset weight of the behavior feature vector to obtain the score of the behavior feature vector.

[0174] The third calculation module is used to calculate the product of the eigenvalue corresponding to the device state feature vector and the preset weight of the device state feature vector to obtain the score of the device state feature vector.

[0175] The summation module is used to add the score of the behavior feature vector and the score of the device state feature vector to obtain a comprehensive score.

[0176] In summary, the embodiments of the present invention provide a method and device for recommending video content, which collects user behavior data and device status data in real time, calculates the comprehensive score of the currently playing video based on the behavior data and device status data, and dynamically adjusts the recommended content based on the comprehensive score, thereby significantly improving the accuracy of video recommendations, improving user experience, reducing user churn, and increasing user stickiness. Furthermore, the viewing time of users on the platform is increased, thereby increasing the click-through rate of advertisements and platform revenue, and ultimately improving the market competitiveness of the platform.

[0177] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.

[0178] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0179] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for recommending video content, characterized in that: The method comprises: When it is detected that a mobile device is playing a video, the user's behavior data and the device status data of the mobile device are collected and pre-processed; If there is behavior data in the preprocessed behavior data indicating that the user is in a video watching state during movement, constructing a feature vector set based on the preprocessed behavior data and the preprocessed device state data, the feature vector set including a behavior feature vector and a device state feature vector; According to the preset weights of the feature vectors, the scores of the behavior feature vector and the device state feature vector are calculated and summed to obtain a comprehensive score; If the comprehensive score is higher than or equal to the threshold, recommending video content of the same type as the currently playing video to the user; If the comprehensive score is lower than a threshold, video content different from the type of the currently playing video is recommended to the user.

2. The method according to claim 1, characterized in that The method further comprises: The video resources stored in the database are classified using the comprehensive score and the feature vector set, a video recommendation list that meets the comprehensive score is constructed, and the video recommendation list is used as the first recommended video list for starting video playback next time.

3. The method according to claim 1, characterized in that The collecting of user behavior data and device status data of the mobile device and preprocessing thereof include: Collecting the user's movement data and the user's operation data on the video to obtain behavior data; Collecting screen direction change data of the mobile device and playback status data of the video when playing the video, to obtain device status data of the mobile device; De-noising and smoothing are performed on the behavior data and the device status data to obtain the pre-processed behavior data and the device status data.

4. The method according to claim 1, characterized in that After the preprocessing, the method further comprises: Determining whether the video is in a continuous playback state according to the preprocessed device status data; If the video is in a continuous playback state, determining whether the user is in a continuous moving state based on the pre-processed behavior data; If yes, determining that there is behavior data in the preprocessed behavior data indicating that the user is in a video watching state during the moving process; If not, it is determined that there is no behavior data in the preprocessed behavior data indicating that the user is in a video watching state during the movement.

5. The method according to claim 1, characterized in that After the preprocessing, the method further comprises: Determining whether the video is in a continuous playback state according to the preprocessed device status data; If the video is in a continuous playback state, determining whether there is behavior data in the preprocessed behavior data for instructing the user to adjust the playback state of the video; If so, determining that the preprocessed behavior data contains behavior data indicating that the user is in a video watching state during the moving process; If not, it is determined that there is no behavior data in the preprocessed behavior data for indicating that the user is in a state of watching a video during movement.

6. The method according to claim 1, characterized in that After the preprocessing, the method further comprises: Determining whether the screen of the mobile device is facing the user according to the preprocessed device status data; If yes, determining whether there is data indicating the user's operation on the video in the preprocessed behavior data; If so, determining that the preprocessed behavior data contains behavior data indicating that the user is in a video watching state during the moving process; If not, it is determined that there is no behavior data in the preprocessed behavior data for indicating that the user is in a state of watching a video during movement.

7. The method according to any one of claims 1 to 6, characterized in that The step of calculating the scores of the behavior feature vector and the device state feature vector according to the preset feature vector weights and summing them up to obtain a comprehensive score includes: Calculating a corresponding eigenvalue of the behavior feature vector according to the behavior feature vector, and calculating a corresponding eigenvalue of the device state feature vector according to the device state feature vector; Calculating the product of the characteristic value corresponding to the behavior characteristic vector and the preset weight of the behavior characteristic vector to obtain the score of the behavior characteristic vector; Calculating the product of the eigenvalue corresponding to the device state feature vector and the preset weight of the device state feature vector to obtain a score of the device state feature vector; The score of the behavior feature vector and the score of the device state feature vector are added together to obtain a comprehensive score.

8. A video content recommendation device, characterized in that: The device comprises: A collection unit, for collecting user behavior data and device status data of the mobile device and performing preprocessing when detecting that the mobile device plays a video; a constructing unit, configured to construct a feature vector set based on the preprocessed behavior data and the preprocessed device state data if there is behavior data indicating that the user is in a video watching state during movement, wherein the feature vector set includes a behavior feature vector and a device state feature vector; A calculation unit, configured to calculate scores of the behavior feature vector and the device state feature vector according to preset feature vector weights and sum them up to obtain a comprehensive score; A first recommendation unit, configured to recommend video content of the same type as the currently played video to the user if the comprehensive score is higher than or equal to a threshold; The second recommendation unit is configured to recommend video content different from the type of the currently played video to the user if the comprehensive score is lower than a threshold.

9. The device according to claim 8, characterized in that The device also includes: A list building unit is used to classify the video resources stored in the database using the comprehensive score and the feature vector set, build a video recommendation list that meets the comprehensive score, and use the video recommendation list as the first recommended video list for starting video playback next time.

10. The device according to claim 8, characterized in that The acquisition unit is specifically used for: Collect the user's movement data and the user's operation data on the video to obtain behavior data; collect the screen direction change data of the mobile device and the playback status data of the video when playing the video to obtain the device status data of the mobile device; denoise and smooth the behavior data and the device status data to obtain the pre-processed behavior data and the device status data.

Images