On-demand video caching method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an on-demand video caching method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a user on-demand behavior log which comprises a video identifier; constructing a popularity feature vector based on the user on-demand behavior log; determining a video popularity feature value corresponding to the video identifier based on the popularity feature vector; and under the condition that the video popularity characteristic value is a preset popularity characteristic value, caching video data corresponding to the video identifier. Through the embodiment of the invention, the request processing pressure of the source station can be reduced, the request response time is reduced, and the user experience is improved.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology, and in particular to a video-on-demand caching method, a video-on-demand caching device, an electronic device, and a computer-readable storage medium. Background Technology

[0002] In related technologies, on-demand precaching is based on access time. The precaching strategy is determined according to the most recent access time of the video content, and the content that has not been accessed for the longest time is prioritized for elimination. However, this method cannot cope with sudden access. When a large number of users simultaneously access newly uploaded "new content" or "long-unaccessed" old content, it will cause a sudden increase in back-to-origin requests, resulting in excessive pressure on the origin server. Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention are proposed to provide a video-on-demand caching method, a video-on-demand caching device, an electronic device, and a computer-readable storage medium that overcome or at least partially solve the above problems.

[0004] To address the above problems, this invention discloses a method for caching on-demand videos, including: Obtain user on-demand behavior logs, wherein the user on-demand behavior logs include video identifiers; Based on the user on-demand behavior logs, a popularity feature vector is constructed; The video popularity feature value corresponding to the video identifier is determined based on the popularity feature vector. If the video popularity feature value is a preset popularity feature value, the video data corresponding to the video identifier is cached to a preset edge node.

[0005] Optionally, the step of constructing a popularity feature vector based on the user's on-demand behavior log includes: Analyze the user on-demand behavior logs to determine the video identifier, timestamp, total number of requests, number of requesting users, and effective playback duration; The video identifier, the timestamp, the total number of requests, the number of requesting users, and the effective playback duration are combined to generate a popularity feature vector.

[0006] Optionally, the step of determining the video popularity feature value corresponding to the video identifier based on the popularity feature vector includes: Based on a preset time-series popularity prediction model, the popularity feature vector is analyzed for popularity trend to determine the video popularity feature value corresponding to the video identifier.

[0007] Optionally, the preset time-series heat prediction model is trained in the following manner: Determine the stable heat index and the sudden growth index corresponding to the heat feature vector; Based on the stable heat index and the sudden growth index, a reference heat characteristic value is determined; The heat feature vector is labeled based on the reference heat feature value to generate a historical behavior sequence; Supervised learning is employed to train a preset initial neural network model using the historical behavior sequence, thereby obtaining a preset time-series popularity prediction model; the preset time-series popularity prediction model is used to output video popularity feature values ​​based on the popularity feature vector.

[0008] Optionally, the step of determining the stable heat index and the sudden growth index corresponding to the heat feature vector includes: A stable popularity index is determined based on the number of requesting users; Based on the changing trend of the stable heat index, the sudden growth index is determined.

[0009] Optionally, determining the reference heat characteristic value based on the stable heat index and the sudden growth index includes: When the stable popularity index is greater than the preset stable popularity threshold, or the sudden growth index is greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the first feature value, and the first feature value represents the popular video. When the stable popularity index is not greater than the preset stable popularity threshold and the sudden growth index is not greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the second feature value, and the second feature value represents the non-popular video.

[0010] Optionally, the step of caching the video data corresponding to the video identifier to a preset edge node when the video popularity feature value is a preset popularity feature value includes: If the video popularity feature value is a preset popularity feature value, a list to be cached is generated based on the video identifier; A caching task is triggered based on the list of items to be cached; The cache task is then sent to the preset edge node.

[0011] In a second aspect, embodiments of the present invention disclose a video-on-demand caching device, comprising: The acquisition module is used to acquire user on-demand behavior logs, which include video identifiers; The construction module is used to construct a popularity feature vector based on the user's on-demand behavior logs; The popularity module is used to determine the video popularity feature value corresponding to the video identifier based on the popularity feature vector; The caching module is used to cache the video data corresponding to the video identifier to a preset edge node when the video popularity feature value is a preset popularity feature value.

[0012] In a third aspect, an embodiment of the present invention discloses an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the on-demand video caching method as described above.

[0013] In a fourth aspect, embodiments of the present invention disclose a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the on-demand video caching method described above.

[0014] The embodiments of the present invention have the following advantages: This invention, in its embodiments, acquires user on-demand behavior logs, including video identifiers; constructs a popularity feature vector based on these logs; determines a video popularity feature value corresponding to each video identifier based on the popularity feature vector; and caches the video data corresponding to each video identifier to a preset edge node when the video popularity feature value is a preset value. By collecting user on-demand behavior logs to construct a popularity feature vector, analyzing video popularity, and dynamically adjusting pre-cached content based on video popularity feature values, the invention reduces the request processing pressure on the origin server; and prioritizes storing popular content to edge nodes, reducing user request response time and improving user experience. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of an embodiment of the video-on-demand caching method of the present invention; Figure 2 This is a flowchart illustrating the steps of another embodiment of the on-demand video caching method of the present invention; Figure 3 This is a schematic diagram illustrating the architecture of an example of a video-on-demand caching method according to the present invention; Figure 4 This is a flowchart illustrating the steps of an example of a video-on-demand caching method according to the present invention; Figure 5 This is a structural block diagram of an embodiment of a video-on-demand caching device according to the present invention; Figure 6 This is a schematic diagram of an embodiment of an electronic device according to the present invention; Figure 7 This is a schematic diagram of an embodiment of a computer-readable storage medium according to the present invention. Detailed Implementation

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] To ensure a good viewing experience for users, such as IPTV (Internet Protocol version 1) Television (interactive network television) typically caches audio and video content in advance at edge nodes to reduce the waiting time for users to request content in real time. Unlike live content, which is real-time and has a short storage time (typically seven days for replay content), video-on-demand content has a large amount of existing content, and user access is asynchronous, random, and diverse. Therefore, it is necessary to capture user behavior for caching optimization. Among related technologies, video-on-demand pre-caching methods are based on access time. The pre-caching strategy is determined according to the most recent access time of the video content, prioritizing the elimination of the least accessed content. However, this method cannot handle sudden access. When a large number of users simultaneously access newly added "new content" or "long-unaccessed" old content, it will cause a surge in back-to-origin requests, leading to excessive pressure on the origin server. Static pre-caching based on user access frequency is based on historical statistics such as clicks and visits. This method cannot capture sudden traffic surges in real time, and sudden surges in access to popular content may cause stuttering. Traditional solutions all have common problems such as lack of dynamic adaptability and the lack of historical data for new content, which leads to poor pre-caching effectiveness.

[0018] Reference Figure 1 The diagram illustrates a flowchart of an embodiment of a video-on-demand caching method according to the present invention. The video-on-demand caching method may specifically include the following steps: Step 101: Obtain user on-demand behavior logs, wherein the user on-demand behavior logs include video identifiers; User video-on-demand behavior logs can be obtained from either the server or the client. These logs record user video-on-demand behavior data. By analyzing these logs, we can understand a user's video-on-demand habits and the specific content they request. The user video-on-demand behavior logs include video identifiers.

[0019] For example, user on-demand behavior logs can be obtained from IPTV servers or clients.

[0020] Step 102: Construct a popularity feature vector based on the user on-demand behavior log; The popularity feature vector for each video can be analyzed from user on-demand behavior logs. The popularity feature vector represents the on-demand popularity index of the video.

[0021] Step 103: Determine the video popularity feature value corresponding to the video identifier based on the popularity feature vector; Based on the popularity of each video in the popularity feature vector, the video popularity feature value corresponding to the video identifier can be determined. The video identifier corresponds to the video data; that is, the on-demand popularity of the video data can be represented by the video popularity feature value.

[0022] Step 104: If the video popularity feature value is a preset popularity feature value, cache the video data corresponding to the video identifier to a preset edge node.

[0023] If the video popularity feature value is a preset popularity feature value, it means that the video corresponding to the video popularity feature value is relatively popular. The video data corresponding to the video identifier can be cached in the preset edge node, so that when the user requests a video, the preset edge node can respond and play the corresponding video data in a timely manner, thereby improving the user experience.

[0024] This invention, in its embodiments, acquires user on-demand behavior logs, including video identifiers; constructs a popularity feature vector based on these logs; determines a video popularity feature value corresponding to each video identifier based on the popularity feature vector; and caches the video data corresponding to each video identifier to a preset edge node when the video popularity feature value is a preset value. By collecting user on-demand behavior logs to construct a popularity feature vector, analyzing video popularity, and dynamically adjusting pre-cached content based on video popularity feature values, the invention reduces the request processing pressure on the origin server; and prioritizes storing popular content to edge nodes, reducing user request response time and improving user experience.

[0025] Reference Figure 2 The diagram illustrates a flowchart of another embodiment of the on-demand video caching method of the present invention, which specifically includes the following steps: Step 201: Obtain user on-demand behavior logs, wherein the user on-demand behavior logs include video identifiers; It can obtain user on-demand behavior logs, such as collecting user on-demand behavior logs from IPTV edge gateways. The content of the user on-demand behavior logs may include, but is not limited to, user access addresses, video identifiers, viewing durations, and other user behavior log data.

[0026] Furthermore, user on-demand behavior logs can be cleaned to avoid errors. Abnormal hang-up requests in service termination reasons can be filtered out. Core fields of user access logs for different request protocols can be extracted. For example, for RTSP (Real Time Streaming Protocol) user on-demand behavior logs, multiple operations by the same user can be associated using sessionID (user identifier), and the user access address, content number, and request start and end time of each request can be extracted. For HLS (HTTP Live Streaming) user on-demand behavior logs, the user access address, video identifier, and segment name of short connections can be statistically analyzed. The user on-demand behavior log data is normalized, and invalid data with a duration less than or equal to the threshold is cleaned according to business thresholds. For example, records with a viewing time of less than 10 seconds are judged as "accidental clicks." Simultaneously, for HLS protocol short connection requests, the actual viewing time is calculated by aggregating the request times of consecutive segments. The cleaned user on-demand behavior logs are then used for subsequent processing.

[0027] Step 202: Construct a popularity feature vector based on the user on-demand behavior log; Content can be extracted from user on-demand behavior logs. By constructing time features from historical user on-demand behavior log data, a popularity feature vector can be generated for each video identifier, representing the basic historical popularity data of the video. Access logs from two transmission protocols can be mapped to a unified format popularity feature vector and integrated along the statistical period dimension.

[0028] In an optional embodiment of the present invention, the step of constructing a popularity feature vector based on the user on-demand behavior log includes: Sub-step S2021: parse the user on-demand behavior log to determine the video identifier, timestamp, total number of requests, number of requesting users, and effective playback duration; It can parse the content of user on-demand behavior logs to determine the video identifier, timestamp, total number of requests, number of requesting users, and effective playback duration. The video identifier is the identity identifier corresponding to the video data. The timestamp is the time of the user's action. The total number of requests is the number of times the video data was viewed. The number of requesting users is the total number of users who viewed the video data. The effective playback duration is the valid playback duration of the video data.

[0029] Sub-step S2021: Combine the video identifier, the timestamp, the total number of requests, the number of requesting users, and the effective playback duration to generate a popularity feature vector.

[0030] For each video data, the video identifier, timestamp, total number of requests, number of requesting users, and effective playback duration can be combined according to a preset format to form the popularity feature vector corresponding to that video data.

[0031] For example, the following popularity feature vector can be combined: ['Video ID', 'Timestamp', 'Total Requests', 'Number of Requesting Users', 'Effective Playback Duration']. Based on this basic data, feature vectors are generated, including peak period identifier and time dimension feature generation: the timestamp is encoded to generate time feature encoding for the statistical period dimension, while adding evening peak time period identifiers and weekday identifiers; historical visit volume feature generation: original visit volume feature, total visit volume, average visit volume, maximum visit volume, minimum visit volume; playback duration feature generation: effective playback time feature; the time dimension feature, historical visit volume feature, and playback duration feature constitute a complete feature vector. Among them, the original visit volume and playback duration have the technical characteristic of a large range of values, so they need to be standardized to make their weights more balanced.

[0032] Step 203: Perform a popularity trend analysis on the popularity feature vector based on a preset time-series popularity prediction model to determine the video popularity feature value corresponding to the video identifier; Because IPTV content access has time-series characteristics, user viewing behavior has time-time attributes, such as weekends, holidays, and morning and evening peak hours. At the same time, the viewing of variety shows and drama series on-demand content has continuity. Since the time-series model captures time dependence and dynamic changes by integrating time information, it can perform popularity trend analysis on the popularity feature vector based on the preset time-series popularity prediction model, accurately identify the popularity of the video, and determine the video popularity feature value corresponding to the video identifier.

[0033] In an optional embodiment of the present invention, the preset time-series popularity prediction model is trained as follows: determining the stable popularity index and the sudden growth index corresponding to the popularity feature vector; determining the reference popularity feature value based on the stable popularity index and the sudden growth index; labeling the popularity feature vector based on the reference popularity feature value to generate a historical behavior sequence; using supervised learning, training the preset initial neural network model with the historical behavior sequence to obtain the preset time-series popularity prediction model; the preset time-series popularity prediction model is used to output video popularity feature values ​​based on the popularity feature vector.

[0034] In the training process of the time-series popularity prediction model, the stable popularity index and the burst growth index corresponding to the popularity feature vector can be calculated first. The stable popularity index represents the trend of sustained popularity of video data. The burst growth index represents the trend of burst popularity of video data. Combining the trends represented by the stable popularity index and the burst growth index, the reference popularity feature value of the video data is determined. The popularity feature vector can be labeled based on the reference popularity feature value, thereby annotating the samples corresponding to the popularity feature vector, which is convenient for use in model training. The labeled labels and time series are combined to generate a historical behavior sequence. Using a supervised learning training method, the preset initial neural network model is trained using the above historical behavior sequence to obtain the preset time-series popularity prediction model. The input data of the preset time-series popularity prediction model is the popularity feature vector, and the output data is the video popularity feature value.

[0035] In practical applications, the steps for determining the stable heat index and the sudden growth index corresponding to the heat feature vector include: Sub-step S1: Determine a stable popularity index based on the number of requesting users; By analyzing access volume in edge cache logs, the number of deduplicated user requests is used as a stable popularity indicator. The calculation formula is as follows:

[0036] Among them, SUI is the stable popularity index; This represents a set of playback sessions for a particular user on this content; Indicates the playback duration of a single session, in seconds; This indicates the preset playback duration threshold, such as 5 seconds to exclude accidental clicks; the stable popularity metric filters based on the effective playback duration and deduplicates users, counting the number of users who effectively watched a video within the corresponding time period under a given video identifier.

[0037] Sub-step S2: Based on the changing trend of the stable heat index, determine the sudden growth index.

[0038] A sudden surge in popularity can be measured by the trend of stable metrics, such as the month-on-month increase in the number of requests. The formula for calculating the sudden surge metric is:

[0039] Among them, BGI is a burst growth indicator; Indicates the current statistical period SUI That is, the number of effective viewers. Indicates the previous statistical period SUI ; This represents the base threshold. If the number of visits in the current period does not exceed the preset popularity visit volume, then no month-on-month growth statistics will be performed to prevent misjudgment of small base requests, such as the case where the number of visits increases from 1 to 20. This represents the smoothing coefficient to prevent division by zero; it is usually set to 0.1.

[0040] In one example, determining the reference popularity feature value based on the stable popularity index and the sudden growth index includes: when the stable popularity index is greater than a preset stable popularity threshold, or when the sudden growth index is greater than a preset sudden growth threshold, determining the reference popularity feature value as a first feature value, where the first feature value represents a popular video; when the stable popularity index is not greater than the preset stable popularity threshold, and the sudden growth index is not greater than the preset sudden growth threshold, determining the reference popularity feature value as a second feature value, where the second feature value represents a non-popular video.

[0041] When the stable popularity index exceeds the preset stable popularity threshold, or the sudden growth index exceeds the preset sudden growth threshold, indicating that the video data has significant popularity and is considered a popular video, the reference popularity feature value can be determined as the first feature value. This first feature value represents popular videos. When the stable popularity index does not exceed the preset stable popularity threshold, and the sudden growth index does not exceed the preset sudden growth threshold, indicating that the video data is not a popular video, the reference popularity feature value can be determined as the second feature value. This second feature value represents non-popular videos. This can be expressed using a formula:

[0042] in, This is a reference heat characteristic value. This is a preset stable popularity threshold set according to business requirements; This represents a preset threshold for sudden growth based on business needs.

[0043] Furthermore, for model deployment, the trained lightweight model is deployed to the central node, while the edge nodes receive user request logs in real time. Based on the real-time request data, the log data written locally to disk is also sent back. The central node is equipped with a sliding window mechanism to incrementally train the pre-trained model using new data and fine-tune the model using new data, maintaining the model's awareness and adaptability to changes in user behavior. At the same time, the cache hit rate of the edge nodes is calculated and compared with the time-series popularity prediction results. The model parameters are dynamically adjusted based on the cache hit rate feedback. Since the content center adds new content regularly, the update cycle needs to be divided according to the amount of new content, and the model is retrained regularly according to the cycle to solve the problem of data distribution drift.

[0044] Step 204: If the video popularity feature value is a preset popularity feature value, generate a list to be cached based on the video identifier; When the video popularity feature value is a preset popularity feature value, such as 1 in the example above, if the video popularity feature value is 1, the video data corresponding to that video identifier can be identified as a popular video and needs to be cached. The video identifier can be added to the cacheable list. The cacheable list can contain video identifiers corresponding to multiple popular videos that need to be cached.

[0045] Furthermore, if the video popularity feature value is not the preset popularity feature value, continuing with the example above, when the video popularity feature value is 0, it means that the video data corresponding to the video identifier is determined to be a non-popular video, and there is no need to cache it, thereby reducing the consumption of caching and improving caching efficiency.

[0046] Furthermore, based on the heat prediction results, the content can be divided into multiple levels and stored in different cache media to further optimize the utilization of edge cache resources.

[0047] Step 205: Trigger a caching task based on the list of items to be cached; A caching task can be triggered based on the list of videos to be cached. The caching task is used to control the preset edge nodes to cache the video data corresponding to the video identifiers in the list of videos to be cached.

[0048] For example, we can count the output bandwidth and cached outflow bandwidth of popular content during peak periods, calculate the storage space according to the size of popular content, and determine the access threshold of popular content; based on the output of the prediction model, content exceeding the threshold is marked as popular content and maintained in a database table, and a pre-caching task is triggered based on the video identifier in the table.

[0049] Step 206: Send the cache task to the preset edge node.

[0050] The caching task is distributed to a preset edge node, which then executes the caching task to achieve caching of popular data.

[0051] In addition, the model performance evaluation metrics include the percentage improvement in edge cache hit rate, user request response time, and the quality rate data. A traditional pre-caching strategy using the number of user requests as a metric for popularity is used as the control group, while a pre-caching strategy using time-series model prediction is used as the experimental group. The edge cache hit rate, average response time for user on-demand requests, and the quality rate data for on-demand playback and loading are compared. The optimized model is gradually scaled up through A / B testing, and the stability of core metrics is monitored. Considering different resource conditions, the selection of the time-series model needs to be adjusted based on the computing resources of the central node. If computing resources are insufficient, a lightweight model can be prioritized. The edge server collects and reports access logs to disk. The central server performs normalization processing and feature vector extraction, and inputs the logs into the time-series prediction model deployed at the central server for popularity prediction. Based on the prediction results, the central server issues pre-caching tasks to the edge, triggering content pre-caching to ensure timely response to user access to popular content.

[0052] This invention comprehensively considers both stable high-popularity data and sudden surges in popularity. It also introduces a time-series model to learn content popularity trends, maintaining the model's awareness and adaptability to user behavior through pre-training and online real-time fine-tuning. The model does not process data according to classification tasks; instead, its output popularity predictions facilitate flexible application across various architectures. Based on the prediction model's output, a pre-caching task is triggered in real-time, prioritizing the storage of popular content on edge nodes closest to users to reduce user request response time. Simultaneously, edge cache resource utilization is optimized, reducing invalid cache usage and optimizing edge node storage space. Furthermore, a unified collection and preprocessing mechanism is designed for log data generated by RTSP and HLS protocols in IPTV services. User access data from both protocols is normalized to a unified format, and feature extraction is used to fuse multi-source heterogeneous data, providing a standardized, high-quality data foundation for model training and improving recognition accuracy.

[0053] To enable those skilled in the art to clearly understand the implementation process of the embodiments of the present invention, the following examples are used for illustration: You can refer to Figure 3 This system can be applied to server clusters, including central and edge nodes. The central node collects logs; inputs them into a model for popularity prediction; writes the model's output to a hot content data table; and periodically scans the data table to identify newly added popular content. Then, a pre-caching task is triggered and distributed to the edge nodes. The edge nodes retrieve hot content slices from the origin server and feed them back to the central node. The central node retrieves the content, slices it, and returns the content slices to the edge nodes. When a user initiates a hot content access request through a terminal, the edge nodes retrieve the user's request slice from the origin server, and if the edge cache is hit, the edge hit is returned to the terminal.

[0054] For specific execution procedures, please refer to... Figure 4 The system collects edge access logs; constructs an input sequence using a sliding window; performs popularity prediction using a time-series model; the central module makes a judgment: sets the prediction threshold to a business popularity threshold; when the popularity is high, it writes the data to a popularity data table; it then performs a pre-caching judgment: checks whether the data table has been updated; if there is an update, it triggers the pre-caching task; the edge gateway pre-caches the content to edge storage; and it does not make any adjustments or trigger any tasks when the popularity is low or there are no updates.

[0055] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0056] Reference Figure 5 The diagram illustrates a structural block diagram of an embodiment of a video-on-demand caching device according to the present invention. The video-on-demand caching device may specifically include the following modules: The acquisition module 501 is used to acquire user on-demand behavior logs, wherein the user on-demand behavior logs include video identifiers; Construction module 502 is used to construct a popularity feature vector based on the user on-demand behavior log; The popularity module 503 is used to determine the video popularity feature value corresponding to the video identifier based on the popularity feature vector; The caching module 504 is used to cache the video data corresponding to the video identifier to a preset edge node when the video popularity feature value is a preset popularity feature value. In an optional embodiment of the present invention, the construction module 502 includes: The parsing submodule is used to parse the user on-demand behavior logs and determine the video identifier, timestamp, total number of requests, number of requesting users, and effective playback duration. The combination submodule is used to combine the video identifier, the timestamp, the total number of requests, the number of requesting users, and the effective playback duration to generate a popularity feature vector.

[0057] In an optional embodiment of the present invention, the heat module 503 includes: The popularity submodule is used to perform popularity trend analysis on the popularity feature vector based on a preset time-series popularity prediction model, and determine the video popularity feature value corresponding to the video identifier.

[0058] In an optional embodiment of the present invention, the preset time-series heat prediction model is trained in the following manner: Determine the stable heat index and the sudden growth index corresponding to the heat feature vector; Based on the stable heat index and the sudden growth index, a reference heat characteristic value is determined; The heat feature vector is labeled based on the reference heat feature value to generate a historical behavior sequence; Supervised learning is employed to train a preset initial neural network model using the historical behavior sequence, thereby obtaining a preset time-series popularity prediction model; the preset time-series popularity prediction model is used to output video popularity feature values ​​based on the popularity feature vector.

[0059] In an optional embodiment of the present invention, the step of determining the stable heat index and the sudden growth index corresponding to the heat feature vector includes: A stable popularity index is determined based on the number of requesting users; Based on the changing trend of the stable heat index, the sudden growth index is determined.

[0060] In an optional embodiment of the present invention, determining the reference heat characteristic value based on the stable heat index and the sudden growth index includes: When the stable popularity index is greater than the preset stable popularity threshold, or the sudden growth index is greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the first feature value, and the first feature value represents the popular video. When the stable popularity index is not greater than the preset stable popularity threshold and the sudden growth index is not greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the second feature value, and the second feature value represents the non-popular video.

[0061] In an optional embodiment of the present invention, the cache module 504 includes: The list submodule is used to generate a list to be cached based on the video identifier when the video popularity feature value is a preset popularity feature value; The triggering submodule is used to trigger a caching task based on the list of items to be cached; The dispatch submodule is used to dispatch the cached task to the preset edge node.

[0062] This invention, in its embodiments, acquires user on-demand behavior logs, including video identifiers; constructs a popularity feature vector based on these logs; determines a video popularity feature value corresponding to each video identifier based on the popularity feature vector; and caches the video data corresponding to each video identifier to a preset edge node when the video popularity feature value is a preset value. By collecting user on-demand behavior logs to construct a popularity feature vector, analyzing video popularity, and dynamically adjusting pre-cached content based on video popularity feature values, the invention reduces the request processing pressure on the origin server; and prioritizes storing popular content to edge nodes, reducing user request response time and improving user experience.

[0063] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0064] Reference Figure 6 The present invention also provides an electronic device, comprising: A processor 601 and a memory 602 are provided. The memory 602 stores a computer program executable by the processor 601. When the electronic device is running, the processor 601 executes the computer program to perform the on-demand video caching method as described in any embodiment of the present invention. The on-demand video caching method includes: Obtain user on-demand behavior logs, wherein the user on-demand behavior logs include video identifiers; Based on the user on-demand behavior logs, a popularity feature vector is constructed; The video popularity feature value corresponding to the video identifier is determined based on the popularity feature vector. If the video popularity feature value is a preset popularity feature value, the video data corresponding to the video identifier is cached to a preset edge node.

[0065] Optionally, the step of constructing a popularity feature vector based on the user's on-demand behavior log includes: Analyze the user on-demand behavior logs to determine the video identifier, timestamp, total number of requests, number of requesting users, and effective playback duration; The video identifier, the timestamp, the total number of requests, the number of requesting users, and the effective playback duration are combined to generate a popularity feature vector.

[0066] Optionally, the step of determining the video popularity feature value corresponding to the video identifier based on the popularity feature vector includes: Based on a preset time-series popularity prediction model, the popularity feature vector is analyzed for popularity trend to determine the video popularity feature value corresponding to the video identifier.

[0067] Optionally, the preset time-series heat prediction model is trained in the following manner: Determine the stable heat index and the sudden growth index corresponding to the heat feature vector; Based on the stable heat index and the sudden growth index, a reference heat characteristic value is determined; The heat feature vector is labeled based on the reference heat feature value to generate a historical behavior sequence; Supervised learning is employed to train a preset initial neural network model using the historical behavior sequence, thereby obtaining a preset time-series popularity prediction model; the preset time-series popularity prediction model is used to output video popularity feature values ​​based on the popularity feature vector.

[0068] Optionally, the step of determining the stable heat index and the sudden growth index corresponding to the heat feature vector includes: A stable popularity index is determined based on the number of requesting users; Based on the changing trend of the stable heat index, the sudden growth index is determined.

[0069] Optionally, determining the reference heat characteristic value based on the stable heat index and the sudden growth index includes: When the stable popularity index is greater than the preset stable popularity threshold, or the sudden growth index is greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the first feature value, and the first feature value represents the popular video. When the stable popularity index is not greater than the preset stable popularity threshold and the sudden growth index is not greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the second feature value, and the second feature value represents the non-popular video.

[0070] Optionally, the step of caching the video data corresponding to the video identifier to a preset edge node when the video popularity feature value is a preset popularity feature value includes: If the video popularity feature value is a preset popularity feature value, a list to be cached is generated based on the video identifier; A caching task is triggered based on the list of items to be cached; The cache task is then sent to the preset edge node.

[0071] This invention, in its embodiments, acquires user on-demand behavior logs, including video identifiers; constructs a popularity feature vector based on these logs; determines a video popularity feature value corresponding to each video identifier based on the popularity feature vector; and caches the video data corresponding to each video identifier to a preset edge node when the video popularity feature value is a preset value. By collecting user on-demand behavior logs to construct a popularity feature vector, analyzing video popularity, and dynamically adjusting pre-cached content based on video popularity feature values, the invention reduces the request processing pressure on the origin server; and prioritizes storing popular content to edge nodes, reducing user request response time and improving user experience.

[0072] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0073] Reference Figure 7 This invention also provides a computer-readable storage medium 701, on which a computer program is stored. When the computer program is executed by a processor, it performs the on-demand video caching method as described in any one of the embodiments of this invention. The on-demand video caching method includes: Obtain user on-demand behavior logs, wherein the user on-demand behavior logs include video identifiers; Based on the user on-demand behavior logs, a popularity feature vector is constructed; The video popularity feature value corresponding to the video identifier is determined based on the popularity feature vector. If the video popularity feature value is a preset popularity feature value, the video data corresponding to the video identifier is cached to a preset edge node.

[0074] Optionally, the step of constructing a popularity feature vector based on the user's on-demand behavior log includes: Analyze the user on-demand behavior logs to determine the video identifier, timestamp, total number of requests, number of requesting users, and effective playback duration; The video identifier, the timestamp, the total number of requests, the number of requesting users, and the effective playback duration are combined to generate a popularity feature vector.

[0075] Optionally, the step of determining the video popularity feature value corresponding to the video identifier based on the popularity feature vector includes: Based on a preset time-series popularity prediction model, the popularity feature vector is analyzed for popularity trend to determine the video popularity feature value corresponding to the video identifier.

[0076] Optionally, the preset time-series heat prediction model is trained in the following manner: Determine the stable heat index and the sudden growth index corresponding to the heat feature vector; Based on the stable heat index and the sudden growth index, a reference heat characteristic value is determined; The heat feature vector is labeled based on the reference heat feature value to generate a historical behavior sequence; Supervised learning is employed to train a preset initial neural network model using the historical behavior sequence, thereby obtaining a preset time-series popularity prediction model; the preset time-series popularity prediction model is used to output video popularity feature values ​​based on the popularity feature vector.

[0077] Optionally, the step of determining the stable heat index and the sudden growth index corresponding to the heat feature vector includes: A stable popularity index is determined based on the number of requesting users; Based on the changing trend of the stable heat index, the sudden growth index is determined.

[0078] Optionally, determining the reference heat characteristic value based on the stable heat index and the sudden growth index includes: When the stable popularity index is greater than the preset stable popularity threshold, or the sudden growth index is greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the first feature value, and the first feature value represents the popular video. When the stable popularity index is not greater than the preset stable popularity threshold and the sudden growth index is not greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the second feature value, and the second feature value represents the non-popular video.

[0079] Optionally, the step of caching the video data corresponding to the video identifier to a preset edge node when the video popularity feature value is a preset popularity feature value includes: If the video popularity feature value is a preset popularity feature value, a list to be cached is generated based on the video identifier; A caching task is triggered based on the list of items to be cached; The cache task is then sent to the preset edge node.

[0080] This invention, in its embodiments, acquires user on-demand behavior logs, including video identifiers; constructs a popularity feature vector based on these logs; determines a video popularity feature value corresponding to each video identifier based on the popularity feature vector; and caches the video data corresponding to each video identifier to a preset edge node when the video popularity feature value is a preset value. By collecting user on-demand behavior logs to construct a popularity feature vector, analyzing video popularity, and dynamically adjusting pre-cached content based on video popularity feature values, the invention reduces the request processing pressure on the origin server; and prioritizes storing popular content to edge nodes, reducing user request response time and improving user experience.

[0081] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0082] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0086] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0087] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0088] The foregoing has provided a detailed description of a video-on-demand caching method, a video-on-demand caching device, an electronic device, and a computer-readable storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for caching on-demand videos, characterized in that, include: Obtain user on-demand behavior logs, wherein the user on-demand behavior logs include video identifiers; Based on the user on-demand behavior logs, a popularity feature vector is constructed; The video popularity feature value corresponding to the video identifier is determined based on the popularity feature vector. If the video popularity feature value is a preset popularity feature value, the video data corresponding to the video identifier is cached to a preset edge node.

2. The method according to claim 1, characterized in that, The step of constructing a popularity feature vector based on the user's on-demand behavior log includes: Analyze the user on-demand behavior logs to determine the video identifier, timestamp, total number of requests, number of requesting users, and effective playback duration; The video identifier, the timestamp, the total number of requests, the number of requesting users, and the effective playback duration are combined to generate a popularity feature vector.

3. The method according to claim 2, characterized in that, The step of determining the video popularity feature value corresponding to the video identifier based on the popularity feature vector includes: Based on a preset time-series popularity prediction model, the popularity feature vector is analyzed for popularity trend to determine the video popularity feature value corresponding to the video identifier.

4. The method according to claim 3, characterized in that, The preset time-series heat prediction model is trained in the following manner: Determine the stable heat index and the sudden growth index corresponding to the heat feature vector; Based on the stable heat index and the sudden growth index, a reference heat characteristic value is determined; The heat feature vector is labeled based on the reference heat feature value to generate a historical behavior sequence; Supervised learning is employed to train a preset initial neural network model using the historical behavior sequence, thereby obtaining a preset time-series popularity prediction model; the preset time-series popularity prediction model is used to output video popularity feature values ​​based on the popularity feature vector.

5. The method according to claim 4, characterized in that, The steps for determining the stable heat index and the sudden growth index corresponding to the heat feature vector include: A stable popularity index is determined based on the number of requesting users; Based on the changing trend of the stable heat index, the sudden growth index is determined.

6. The method according to claim 4, characterized in that, The determination of reference heat characteristic values ​​based on the stable heat index and the sudden growth index includes: When the stable popularity index is greater than the preset stable popularity threshold, or the sudden growth index is greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the first feature value, and the first feature value represents the popular video. When the stable popularity index is not greater than the preset stable popularity threshold and the sudden growth index is not greater than the preset sudden growth threshold, the reference popularity feature value is determined to be the second feature value, and the second feature value represents the non-popular video.

7. The method according to claim 1, characterized in that, The step of caching the video data corresponding to the video identifier to a preset edge node when the video popularity feature value is a preset popularity feature value includes: If the video popularity feature value is a preset popularity feature value, a list to be cached is generated based on the video identifier; A caching task is triggered based on the list of items to be cached; The cache task is then sent to the preset edge node.

8. A video-on-demand caching device, characterized in that, include: The acquisition module is used to acquire user on-demand behavior logs, which include video identifiers; The construction module is used to construct a popularity feature vector based on the user's on-demand behavior logs; The popularity module is used to determine the video popularity feature value corresponding to the video identifier based on the popularity feature vector; The caching module is used to cache the video data corresponding to the video identifier to a preset edge node when the video popularity feature value is a preset popularity feature value.

9. An electronic device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the on-demand video caching method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the on-demand video caching method as described in any one of claims 1 to 7.

Images