Video heat prediction method and device, storage medium and electronic device
The pre-constructed random forest model uses video tags and playback volume training decision trees in video data samples, and solves the problem of predicting new video popularity and improves user stickiness and experience.
Patent Information
- Application Number
- CN202210615382.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-06-01
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2042-06-01
AI Technical Summary
The existing technology is difficult to effectively predict the popularity of newly launched videos, affecting user stickiness and user experience.
The pre-constructed random forest model is used to train the decision tree through the video tag and playback volume in the video data sample to calculate the popularity value of the newly launched video, thereby achieving popularity prediction.
It has achieved popularity predictions for newly launched videos, increasing user stickiness and improving user experience.
Smart Images

Figure CN114969436B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of machine learning technology, and in particular to a video heat prediction method and device, a storage medium, and an electronic device. Background Art
[0002] With the development of Internet technology and video services, predicting the popularity of newly launched videos can bring many benefits, such as increasing user stickiness and improving user experience.
[0003] Therefore, how to provide a technical solution for predicting the popularity of newly released videos is a technical problem that technical personnel in this field urgently need to solve. Summary of the invention
[0004] The present application provides a video popularity prediction method and device, a storage medium and an electronic device, the purpose of which is to realize popularity prediction of newly released videos.
[0005] In order to achieve the above objectives, this application provides the following technical solutions:
[0006] A video heat prediction method, comprising:
[0007] Get the video tag of each newly launched video;
[0008] For each newly launched video, the video label of the newly launched video is respectively input into a pre-built random forest model to obtain the heat value corresponding to the newly launched video decided by each decision tree included in the random forest model; wherein the random forest model is pre-trained based on the video data sample included in the video data sample set, the video data sample includes the video label and video playback volume of the historical video to which it belongs, each leaf node of each decision tree included in the random forest model is used to store the label dimension attribute and the heat value corresponding to the label dimension attribute, and the heat value corresponding to the video dimension attribute is determined based on the video playback volume included in the video data sample corresponding to the label dimension attribute;
[0009] For each newly released video, the popularity value of the newly released video determined by each decision tree is calculated to obtain a popularity prediction result of the newly released video.
[0010] In the above method, optionally, the heat value of the newly released video determined by each decision tree is calculated to obtain the heat prediction result of the newly released video, including:
[0011] The heat values of the newly released videos determined by each decision tree are summed up to obtain the heat prediction result of the newly released videos.
[0012] The above method may, optionally, after obtaining the popularity prediction result of the newly released video, further include:
[0013] Based on the popularity prediction results of each newly launched video, determine the target video from each newly launched video;
[0014] Generate a new online video playlist based on the video information of each target video;
[0015] The newly launched video playlist is displayed using a video player.
[0016] In the above method, optionally, the process of constructing the random forest model includes:
[0017] Obtain a video data sample set;
[0018] Randomly select N video data samples from each video data sample included in the video data sample set; N is a positive integer, and N is not greater than the number of video data samples included in the video data sample set;
[0019] Selecting a label dimension attribute subset from each label dimension attribute subset included in the label dimension attribute set as a target set;
[0020] Based on the label dimension attributes of each level included in the target set, a decision tree corresponding to the target set is constructed; the number of levels of the target view set is the same as the number of levels of the decision tree, and each layer of nodes included in the decision tree is used to store the label dimension attributes of the corresponding level;
[0021] For each leaf node included in the constructed decision tree, determine the target video data sample from the video data sample to which the video label corresponding to the label dimension attribute stored in the leaf node belongs among the selected N video data samples, calculate the playback volume included in each target video data sample, obtain the heat value, and store the heat value in the leaf node;
[0022] Determine whether the label dimension attribute set is an empty set;
[0023] If the label dimension attribute set is not an empty set, returning to execute the step of randomly selecting N samples from each video data sample included in the video data sample set;
[0024] If the label dimension attribute set is an empty set, each decision tree is combined into a random forest model.
[0025] In the above method, optionally, the randomly selecting N video data samples from each video data sample included in the video data sample set includes:
[0026] Start the counter to start initial counting;
[0027] Randomly selecting a video data sample from each video data sample included in the video data sample set;
[0028] Adding one to the counting result of the counter, and putting the video data sample back into the video data sample set;
[0029] Determine whether the current counting result of the counter is equal to N;
[0030] If the current counting result of the counter is not N, the process returns to executing the step of randomly selecting a video data sample from each video data sample included in the video data sample set until the current counting result of the counter is N.
[0031] In the above method, optionally, obtaining the video tag of each newly released video includes:
[0032] Get the video content of each newly launched video;
[0033] Use the pre-built video tag recognition system to identify video tags for each newly released video.
[0034] For each newly released video, the video tag identified by the video tag identification system is used as the video tag of the newly released video.
[0035] The above method may optionally include, before using the video tag identified by the video tag identification system as the video tag of the newly-released video, the following steps:
[0036] The video tags identified by the video tag identification system are sent to the review terminal so that the reviewer can modify the video tags identified by the video tag identification system.
[0037] A video heat prediction device, comprising:
[0038] An acquisition unit, used to acquire the video tag of each newly released video;
[0039] An input unit is used to input the video tags of each newly launched video into a pre-built random forest model, respectively, to obtain the heat value corresponding to the newly launched video decided by each decision tree included in the random forest model; wherein the random forest model is pre-trained based on the video data samples included in the video data sample set, the video data samples include the video tags and video playback volume of the historical videos to which they belong, each leaf node of each decision tree included in the random forest model is used to store the label dimension attribute and the heat value corresponding to the label dimension attribute, and the heat value corresponding to the video dimension attribute is determined based on the video playback volume included in the video data sample corresponding to the label dimension attribute;
[0040] The calculation unit is used to calculate the heat value of each newly launched video determined by each decision tree for each newly launched video, so as to obtain the heat prediction result of the newly launched video.
[0041] A storage medium stores an instruction set, wherein the instruction set, when executed by a processor, implements the video heat prediction method as described above.
[0042] An electronic device, comprising:
[0043] A memory for storing at least one set of instructions;
[0044] The processor is used to execute the instruction set stored in the memory, and implement the video heat prediction method as mentioned above by executing the instruction set.
[0045] Compared with the prior art, this application has the following advantages:
[0046] The present application provides a method and device for predicting the popularity of a video, a storage medium, and an electronic device. The method pre-constructs a random forest model, the random forest model includes multiple decision trees, each leaf node of each decision tree is used to store a label dimension attribute and a heat value corresponding to the label dimension attribute, and the video label of the newly launched video is input into the random forest model, so that each decision tree included in the random forest model can decide to obtain the heat value corresponding to the newly launched video, and then the heat value of the newly launched video decided by each decision tree is calculated to obtain the heat prediction result of the newly launched video. The heat prediction of the newly launched video is realized, thereby increasing user stickiness and improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] In order to more clearly illustrate the embodiments of the present application 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 application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0048] Figure 1 A method flow chart of a video heat prediction method provided in this application;
[0049] Figure 2 Another method flow chart of a video heat prediction method provided by the present application;
[0050] Figure 3 Another method flow chart of a video heat prediction method provided by the present application;
[0051] Figure 4 Another method flow chart of a video heat prediction method provided by the present application;
[0052] Figure 5 Another method flow chart of a video heat prediction method provided by the present application;
[0053] Figure 6 A schematic diagram of the structure of a video heat prediction device provided in this application;
[0054] Figure 7 A schematic diagram of the structure of an electronic device provided in this application. DETAILED DESCRIPTION
[0055] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0056] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.
[0057] It should be noted that the concepts such as "first" and "second" mentioned in the present application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.
[0058] It should be noted that the modifications of "one" and "plurality" mentioned in the disclosure of this application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".
[0059] The present application embodiment provides a method for predicting video popularity. The flowchart of the method is as follows: Figure 1 As shown, specifically including:
[0060] S101: Obtain the video tag of each newly released video.
[0061] In this embodiment, the video tag of each newly released video is obtained, specifically, according to a preset time period, the video tag of each newly released video in the current time period is obtained. Preferably, the preset time period is 24 hours.
[0062] For details, see Figure 2 ,The process of obtaining the video tag of each newly launched video includes the following steps:
[0063] S201, obtaining the video content of each newly released video.
[0064] In this embodiment, the video content of each newly released video is obtained.
[0065] S202: Using a pre-built video tag recognition system, perform video tag recognition on the video content of each newly released video.
[0066] In this example, a video tag recognition system is pre-built, and the video content of each newly launched video is input into the video tag recognition system. The video tag recognition system recognizes the video tag of the video content and outputs the recognition result, wherein the recognition result includes the video tag of the newly launched video.
[0067] S203: For each newly released video, the video tag identified by the video tag identification system is used as the video tag of the newly released video.
[0068] In this embodiment, for each newly released video, the video tag identified by the video tag identification system is used as the video tag of the newly released video.
[0069] Optionally, before step S203, the following steps may also be included:
[0070] The video tags identified by the video tag identification system are sent to the review terminal so that the reviewer can correct the video tags identified by the video tag identification system.
[0071] In this embodiment, before the video tag identified by the video tag identification system is used as the video tag of the newly-released video, the video tag identified by the video tag identification system can also be sent to the review system so that the reviewer can correct the video tag identified by the video tag identification system, thereby determining the corrected video tag as the video tag of the newly-released video.
[0072] S102: For each newly released video, the video tags of the newly released video are respectively input into a pre-built random forest model to obtain a heat value corresponding to the newly released video determined by each decision tree included in the random forest model.
[0073] In this embodiment, a random forest model is pre-constructed. The random forest model is pre-trained based on video data samples included in a video data sample set. The video data samples include video tags and video playback volume of the historical videos to which they belong. Each leaf node of each decision tree included in the random forest model is used to store label dimension attributes and heat values corresponding to the label dimension attributes. The heat value corresponding to the video label is determined based on the video playback volume included in the video data samples corresponding to the label dimension attributes.
[0074] See also Figure 3 ,The construction process of the random forest model includes the following steps:
[0075] S301: Obtain a set of video data samples.
[0076] In this embodiment, a video data sample set is obtained, wherein the video data sample set includes video data samples corresponding to multiple historical videos, that is, the video data sample set includes multiple video data samples.
[0077] The video data sample includes all video tags and video playback volume of the historical video to which the video data sample belongs; it should be noted that the video data sample includes the video playback volume and at least one video tag.
[0078] S302: Randomly select N video data samples from the video data samples included in the video data sample set.
[0079] In this embodiment, N video data samples are randomly selected from each video data sample included in the video data sample set each time, where N is a positive integer and is not greater than the number of video data samples included in the video data sample set.
[0080] In this embodiment, refer to Figure 4 The process of randomly selecting N video data samples from each video data sample included in the video data set specifically includes the following steps:
[0081] S401, start the counter to start initial counting.
[0082] In this embodiment, before randomly selecting N video data samples each time, a counter is started and the initial count is set to 0.
[0083] S402: Randomly select a video data sample from each video data sample included in the video data sample set.
[0084] In this embodiment, a video data sample is randomly selected from each video data sample included in the video data sample set.
[0085] S403: Add one to the counting result of the counter, and put the video data sample back into the video data sample set.
[0086] In this embodiment, after a video data sample is selected, the counting result of the counter is incremented by one, and the video data sample is returned to the video data sample set, thereby realizing selection with replacement.
[0087] S404, determine whether the current count result of the counter is equal to N, if not, return to execute S402, if so, determine that the selection of N video data samples is completed, and end.
[0088] In this embodiment, it is determined whether the current counting result of the counter is equal to N, that is, it is determined whether the selection of N video data samples is completed.
[0089] In this embodiment, if the current counting result of the counter is not equal to N, that is, the current counting result of the counter is less than N, it means that the selection of N video data samples has not been completed, and returns to execute step S402. If the current counting result of the counter is equal to N, it means that the selection of N video data samples has been completed.
[0090] S303: Select a label dimension attribute subset from each label dimension attribute subset included in the label dimension attribute set as a target set.
[0091] In this embodiment, a label dimension attribute set is pre-constructed, and the label dimension attribute set includes multiple label dimension attribute subsets, and the label dimension attribute subsets include label dimension attributes of multiple levels. For example, the label dimension set includes a label dimension attribute of the first level: subject matter, a label dimension attribute of the second level: sports, and a label dimension attribute of the third level: ball games and track and field. It should be noted that there is a genus-species relationship between label dimension attributes of adjacent levels. For example, ball games are a type of sports, and track and field is also a type of sports.
[0092] It should be noted that the number of label dimension attributes of the first level included in the label dimension attribute subset is one, and the number of label dimension attributes of other levels except the first level in the label dimension attribute subset is at least one.
[0093] In this embodiment, one label dimension attribute subset is selected from each label dimension attribute subset included in the label dimension attribute set as the target set each time.
[0094] It should be noted that the selection of the label dimension attribute subset from each label dimension attribute subset included in the view label dimension attribute set is a selection without replacement.
[0095] S304: Construct a decision tree corresponding to the target set based on the label dimension attributes of each level included in the target set.
[0096] In this embodiment, a decision tree corresponding to the target set is constructed based on the label dimension attributes of each level included in the target set.
[0097] It should be noted that the label dimension attributes of the first level included in the target set are stored in the root node of the constructed decision tree. The number of levels of the target view set is the same as the number of levels of the decision tree. Each layer of nodes included in the decision tree is used to store the label dimension attributes of the corresponding level.
[0098] S305. For each leaf node included in the constructed decision tree, determine the target video data sample from the selected N video data samples, the video data sample to which the video label corresponding to the label dimension attribute stored in the leaf node belongs, and calculate the playback volume included in each target video data sample to obtain the heat value, and store the heat value in the leaf node.
[0099] In this embodiment, for each leaf node included in the constructed decision tree, the target video data sample is determined by the video data sample to which the video label corresponding to the label dimension attribute stored in the leaf node belongs among the N selected video data samples. In other words, the target video data sample corresponding to each leaf node included in the decision tree is determined, and the video label possessed by the target video data sample corresponding to each leaf node contains the video label corresponding to the label dimension attribute stored in the leaf node.
[0100] In this embodiment, for each leaf node, the playback volume included in the target video data sample corresponding to the leaf node is calculated. Specifically, the corresponding relationship between the preset playback volume interval and the heat value is determined to determine the playback volume interval in which the playback volume included in each target video data sample is located, so that the heat value corresponding to the playback volume is determined as the heat value corresponding to the target video data sample, and the heat value corresponding to each target video data sample is summed to obtain the heat value corresponding to the leaf node, and the heat value corresponding to the leaf node is stored in the leaf node.
[0101] The process of calculating the playback volume of each target video data sample and obtaining the heat value mentioned above is illustrated as follows:
[0102] If the target video data sample 1 includes 10W video playbacks, and the target video sample 2 includes 4W video playbacks, the corresponding relationship between the preset playback volume interval and the heat value is: the heat value in the playback volume interval [1, 2W) is 1, the heat value in the playback volume interval [2W, 5W) is 2, and the heat value in the playback volume interval [5W, 20W) is 3. Therefore, the heat value corresponding to the target video data sample 1 is determined to be 3, the heat value corresponding to the target video data sample 2 is 2, and the total heat value is 5, that is, the playback volume included in each target video data sample is calculated, and the heat value is 5.
[0103] S306. Determine whether the label dimension attribute is an empty set. If so, execute S307. If not, return to execute S302.
[0104] In this embodiment, it is determined whether the label dimension attribute set is an empty set. If the label dimension attribute set is an empty set, step S307 is executed. If the label dimension attribute set is not an empty set, the process returns to step S302.
[0105] S307, composing each decision tree into a random forest model.
[0106] In this embodiment, if the label dimension attribute set is an empty set, the constructed decision trees are combined into a random forest model.
[0107] It should be noted that the constructed random forest model can be updated according to a preset update cycle.
[0108] In this embodiment, for each newly launched video, the video label of the newly launched video is input into the random forest model respectively. Each decision tree in the random forest model decides the video label corresponding to the label dimension attribute stored in its own leaf node. If the video label corresponding to its own leaf is decided, the heat value stored in the leaf node is output.
[0109] S103: For each newly released video, the popularity value of the newly released video determined by each decision tree is calculated to obtain a popularity prediction result of the newly released video.
[0110] In this embodiment, for each newly launched video, the heat value of the newly launched video decided by each decision tree is calculated. Specifically, the heat value of the newly launched video decided by each decision tree is summed up to obtain the heat prediction result of the newly launched video.
[0111] For example, if the video tags of a newly released video are football and ridicule, the heat value obtained by the decision tree for the football tag is 15, and the heat value obtained by the decision tree for the ridicule tag is 12, then the predicted heat value of the newly released video is 27.
[0112] The video popularity prediction method provided in the embodiment of the present application pre-builds a random forest model, which includes multiple decision trees. Each leaf node of each decision tree is used to store the label dimension attribute and the heat value corresponding to the label dimension attribute. The video label of the newly launched video is input into the random forest model, so that each decision tree included in the random forest model can decide to obtain the heat value corresponding to the newly launched video, and then the heat value of the newly launched video decided by each decision tree is calculated to obtain the heat prediction result of the newly launched video. The heat prediction of the newly launched video is realized, thereby increasing user stickiness and improving user experience.
[0113] See also Figure 5 The video popularity prediction method provided in the embodiment of the present application may further include the following steps after step S103:
[0114] S501. Determine a target video from among the newly released videos based on the popularity prediction results of the newly released videos.
[0115] In this embodiment, after obtaining the heat prediction result of each newly launched video, a target video is determined from each newly launched video based on the heat prediction result of each newly launched video. Specifically, based on the heat prediction result of each newly launched video, each newly launched video is sorted in the order of heat prediction results from large to small or from small to large, and a preset number of newly launched videos are selected from the sorted new online videos in order from front to back or from back to front as target videos. It should be noted that if the new online videos are sorted in order from large to small according to the heat prediction results, then correspondingly, the selection of the preset number of new online videos is selected in order from front to back; if the new online videos are sorted in order from small to large according to the heat prediction results, then correspondingly, the selection of the preset number of new online videos is selected in order from back to front.
[0116] S502: Generate a new online video playlist based on the video information of each target video.
[0117] Based on the video information of each target video, a new online video playlist is generated, wherein the video information includes but is not less than the video name.
[0118] It should be noted that the target videos in the newly launched video playlist are arranged in descending order according to the popularity prediction results.
[0119] S503: Use a video player to display the newly released video playlist.
[0120] In this embodiment, after the newly launched video playlist is generated, the newly launched video playlist can be displayed using a video player, so that users can quickly find the newly launched videos of interest through the newly launched video playlist, thereby improving user experience and user stickiness.
[0121] In the video popularity prediction method provided in the embodiment of the present application, after obtaining the popularity prediction result of each newly online video, a target video is determined from each newly online video based on the popularity prediction result of each newly online video, and a newly online video playlist is generated based on the video information of each target video. The newly online video playlist is displayed using a video player, so that users can quickly find the newly online videos of interest through the newly online video playlist, thereby improving user experience and user stickiness.
[0122] It should be noted that although the instructions are depicted in a particular order, this should not be understood as requiring that the instructions be executed in the particular order shown or in a sequential order. In certain circumstances, multitasking and parallel processing may be advantageous.
[0123] It should be understood that the various steps described in the method embodiments disclosed in the present application can be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present application is not limited in this respect.
[0124] and Figure 1 Corresponding to the method described above, the embodiment of the present application also provides a video heat prediction device for predicting the heat of a video. Figure 1 The specific implementation of the method is shown in the following diagram: Figure 6 As shown, specifically including:
[0125] An acquisition unit 601 is used to acquire a video tag of each newly released video;
[0126] An input unit 602 is used to input the video tags of each newly launched video into a pre-built random forest model, and obtain the heat value corresponding to the newly launched video decided by each decision tree included in the random forest model; wherein the random forest model is pre-trained based on the video data samples included in the video data sample set, and the video data samples include the video tags and video playback volume of the historical videos to which they belong, and each leaf node of each decision tree included in the random forest model is used to store the label dimension attribute and the heat value corresponding to the label dimension attribute, and the heat value corresponding to the video dimension attribute is determined based on the video playback volume included in the video data sample corresponding to the label dimension attribute;
[0127] The calculation unit 603 is used to calculate the popularity value of each newly launched video determined by each decision tree, and obtain the popularity prediction result of the newly launched video.
[0128] The video popularity prediction device provided by the embodiment of the present application pre-builds a random forest model, which includes multiple decision trees. Each leaf node of each decision tree is used to store the label dimension attribute and the popularity value corresponding to the label dimension attribute. The video label of the newly launched video is input into the random forest model, so that each decision tree included in the random forest model can decide to obtain the popularity value corresponding to the newly launched video, and then the popularity value of the newly launched video decided by each decision tree is calculated to obtain the popularity prediction result of the newly launched video. The popularity prediction of the newly launched video is realized, thereby increasing user stickiness and improving user experience.
[0129] In one embodiment of the present application, based on the above solution, the calculation unit 603 is specifically used to:
[0130] The heat values of the newly released videos determined by each decision tree are summed up to obtain the heat prediction result of the newly released videos.
[0131] In one embodiment of the present application, based on the above solution, it can also be configured as follows:
[0132] A determination unit, configured to determine a target video from each of the newly released videos based on the popularity prediction results of each of the newly released videos;
[0133] A generating unit, used for generating a new online video playlist based on the video information of each target video;
[0134] The display unit is used to display the newly launched video playlist using a video player.
[0135] In one embodiment of the present application, based on the above solution, the input unit 602 is specifically used in the process of constructing the random forest model:
[0136] Obtain a video data sample set;
[0137] Randomly select N video data samples from each video data sample included in the video data sample set; N is a positive integer, and N is not greater than the number of video data samples included in the video data sample set;
[0138] Selecting a label dimension attribute subset from each label dimension attribute subset included in the label dimension attribute set as a target set;
[0139] Based on the label dimension attributes of each level included in the target set, a decision tree corresponding to the target set is constructed; the number of levels of the target view set is the same as the number of levels of the decision tree, and each layer of nodes included in the decision tree is used to store the label dimension attributes of the corresponding level;
[0140] For each leaf node included in the constructed decision tree, determine the target video data sample from the video data sample to which the video label corresponding to the label dimension attribute stored in the leaf node belongs among the selected N video data samples, calculate the playback volume included in each target video data sample, obtain the heat value, and store the heat value in the leaf node;
[0141] Determine whether the label dimension attribute set is an empty set;
[0142] If the label dimension attribute set is not an empty set, returning to execute the step of randomly selecting N samples from each video data sample included in the video data sample set;
[0143] If the label dimension attribute set is an empty set, each decision tree is combined into a random forest model.
[0144] In one embodiment of the present application, based on the above solution, when the input unit 602 randomly selects N video data samples from each video data sample included in the video data sample set, it is specifically used to:
[0145] Start the counter to start initial counting;
[0146] Randomly selecting a video data sample from each video data sample included in the video data sample set;
[0147] Adding one to the counting result of the counter, and putting the video data sample back into the video data sample set;
[0148] Determine whether the current counting result of the counter is equal to N;
[0149] If the current counting result of the counter is not N, the process returns to executing the step of randomly selecting a video data sample from each video data sample included in the video data sample set until the current counting result of the counter is N.
[0150] In one embodiment of the present application, based on the above solution, the acquisition unit 501 is specifically used for:
[0151] Get the video content of each newly launched video;
[0152] Use the pre-built video tag recognition system to identify video tags for each newly released video.
[0153] For each newly released video, the video tag identified by the video tag identification system is used as the video tag of the newly released video.
[0154] In one embodiment of the present application, based on the above solution, the acquisition unit 601 is further used to:
[0155] The video tags identified by the video tag identification system are sent to the review terminal so that the reviewer can modify the video tags identified by the video tag identification system.
[0156] An embodiment of the present application further provides a storage medium storing an instruction set, wherein when the instruction set is executed, a video heat prediction method as disclosed in any of the above embodiments is executed.
[0157] The present application also provides an electronic device, whose structural diagram is as follows: Figure 7 As shown, it specifically includes a memory 701 for storing at least one set of instruction sets; a processor 702 for executing the instruction set stored in the memory by executing the video heat prediction method disclosed in any of the above embodiments.
[0158] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.
[0159] Although several specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope disclosed in the present application. Certain features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination.
[0160] The above description is only a preferred embodiment disclosed in this application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the disclosure of this application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the disclosure of this application (but not limited to) to form a technical solution.
Claims
1. A video heat prediction method, characterized in that: include: Get the video tag of each newly launched video; For each newly launched video, the video label of the newly launched video is respectively input into a pre-built random forest model to obtain the heat value corresponding to the newly launched video decided by each decision tree included in the random forest model; wherein the random forest model is pre-trained based on the video data sample included in the video data sample set, the video data sample includes the video label and video playback volume of the historical video to which it belongs, each leaf node of each decision tree included in the random forest model is used to store the label dimension attribute and the heat value corresponding to the label dimension attribute, and the heat value corresponding to the label dimension attribute is determined based on the video playback volume included in the video data sample corresponding to the label dimension attribute; For each newly released video, the popularity value of the newly released video determined by each decision tree is calculated to obtain a popularity prediction result of the newly released video; The construction process of the random forest model includes: Obtain a video data sample set; Randomly select N video data samples from each video data sample included in the video data sample set; N is a positive integer, and N is not greater than the number of video data samples included in the video data sample set; Selecting a label dimension attribute subset from each label dimension attribute subset included in the label dimension attribute set as a target set; Based on the label dimension attributes of each level included in the target set, a decision tree corresponding to the target set is constructed; the number of levels of the target set is the same as the number of levels of the decision tree, and each layer of nodes included in the decision tree is used to store the label dimension attributes of the corresponding level; For each leaf node included in the constructed decision tree, determine the target video data sample from the video data sample to which the video label corresponding to the label dimension attribute stored in the leaf node belongs among the selected N video data samples, calculate the playback volume included in each target video data sample, obtain the heat value, and store the heat value in the leaf node; Determine whether the label dimension attribute set is an empty set; If the label dimension attribute set is not an empty set, returning to execute the step of randomly selecting N samples from each video data sample included in the video data sample set; If the label dimension attribute set is an empty set, each decision tree is combined into a random forest model.
2. The method according to claim 1, characterized in that The calculating the popularity value of the newly released video determined by each decision tree to obtain the popularity prediction result of the newly released video includes: The heat values of the newly released videos determined by each decision tree are summed up to obtain the heat prediction result of the newly released videos.
3. The method according to claim 1 or 2, characterized in that: After obtaining the popularity prediction result of the newly released video, the method further includes: Based on the popularity prediction results of each newly launched video, determine the target video from each newly launched video; Generate a new online video playlist based on the video information of each target video; The newly launched video playlist is displayed using a video player.
4. The method according to claim 1, characterized in that: The randomly selecting N video data samples from the video data samples included in the video data sample set includes: Start the counter to start initial counting; Randomly selecting a video data sample from each video data sample included in the video data sample set; Adding one to the counting result of the counter, and putting the video data sample back into the video data sample set; Determine whether the current counting result of the counter is equal to N; If the current counting result of the counter is not N, the process returns to executing the step of randomly selecting a video data sample from each video data sample included in the video data sample set until the current counting result of the counter is N.
5. The method according to claim 1, characterized in that The step of obtaining the video tag of each newly released video includes: Get the video content of each newly launched video; Use the pre-built video tag recognition system to identify video tags for each newly released video. For each newly released video, the video tag identified by the video tag identification system is used as the video tag of the newly released video.
6. The method according to claim 5, characterized in that Before using the video tag identified by the video tag identification system as the video tag of the newly-released video, the method further includes: The video tags identified by the video tag identification system are sent to the review terminal so that the reviewer can modify the video tags identified by the video tag identification system.
7. A video heat prediction device, characterized in that: include: An acquisition unit, used to acquire the video tag of each newly released video; An input unit is used to input the video tags of each newly launched video into a pre-built random forest model, respectively, to obtain the heat value corresponding to the newly launched video decided by each decision tree included in the random forest model; wherein the random forest model is pre-trained based on the video data samples included in the video data sample set, the video data samples include the video tags and video playback volume of the historical videos to which they belong, each leaf node of each decision tree included in the random forest model is used to store the label dimension attribute, and the heat value corresponding to the label dimension attribute, and the heat value corresponding to the label dimension attribute is determined based on the video playback volume included in the video data sample corresponding to the label dimension attribute; A calculation unit, configured to calculate, for each newly released video, the popularity value of the newly released video determined by each decision tree, to obtain a popularity prediction result of the newly released video; During the construction process of the random forest model, the input unit is specifically used to: Obtain a video data sample set; Randomly select N video data samples from each video data sample included in the video data sample set; N is a positive integer, and N is not greater than the number of video data samples included in the video data sample set; Selecting a label dimension attribute subset from each label dimension attribute subset included in the label dimension attribute set as a target set; Based on the label dimension attributes of each level included in the target set, a decision tree corresponding to the target set is constructed; the number of levels of the target set is the same as the number of levels of the decision tree, and each layer of nodes included in the decision tree is used to store the label dimension attributes of the corresponding level; For each leaf node included in the constructed decision tree, determine the target video data sample from the video data sample to which the video label corresponding to the label dimension attribute stored in the leaf node belongs among the selected N video data samples, calculate the playback volume included in each target video data sample, obtain the heat value, and store the heat value in the leaf node; Determine whether the label dimension attribute set is an empty set; If the label dimension attribute set is not an empty set, returning to execute the step of randomly selecting N samples from each video data sample included in the video data sample set; If the label dimension attribute set is an empty set, each decision tree is combined into a random forest model.
8. A storage medium, characterized in that: The storage medium stores an instruction set, wherein the instruction set, when executed by a processor, implements the video heat prediction method according to any one of claims 1 to 6.
9. An electronic device, characterized in that: include: A memory for storing at least one set of instructions; A processor is used to execute an instruction set stored in the memory, and implement the video heat prediction method as described in any one of claims 1-6 by executing the instruction set.
