Short video supply and demand diagnosis method and device

Abstract

The invention relates to the technical field of video recommendation, in particular to a short video supply and demand diagnosis method and device which are applied to short video recommendation, a short video replenishment model is constructed through the per capita short video playing duration and the video playing amount of short videos in a short video platform, and the short video replenishment model is constructed as follows: S1, supply and demand data are collected on the short video platform, calculating a supply-demand ratio based on the acquired supply-demand data, judging a content value gap according to the supply-demand ratio, and taking out-of-stock duration as measurement of the content value gap; s2, the marginal distribution value is used for calculating the video production value, and the content value measure is that when one piece of short video content is introduced or produced every time, duration increment can be brought to a short video large disk; and S3, obtaining an optimal short video replenishment scheme according to the stockout duration and the video production value. Compared with the prior art, reference can be provided for content acquisition and buying of the platform, the overall ROI of content acquisition and buying is improved, and the short video platform achieves supply and demand balance.

Description

Technical Field

[0001] The present invention relates to the technical field of video recommendation, and specifically provides a short video supply and demand diagnosis method and device. Background Art

[0002] Currently, most short video businesses do not have an effective and systematic solution to determine whether there is a problem with the current supply and demand situation. A small number of short video businesses will use manual experience to determine whether traffic is too concentrated on specific categories. Some individual businesses will conduct a large number of random experiments and explore the supply and demand situation of different categories by giving different categories promotion and demotion rights.

[0003] On the one hand, the method of judging by manual experience has the problem of inaccurate manual judgment. When judging whether the content volume of a single category is too large or too small, it is mainly judged subjectively by short video deep users or experts, but it is usually impossible to accurately describe or quantify the judgment standard, and the standard cannot be consistent in actual implementation. On the other hand, manual experience is applicable to limited scenarios. When the problem is very obvious, manual judgment may be made on whether a single category is too much or too little. For example, when the game category accounts for 80% of the traffic, this ecology is obviously unhealthy; but when the problem is not so obvious, such as when all categories evenly divide the traffic, manual judgment cannot give specific suggestions on whether the current state is healthy, or in which direction the ecology should be adjusted, and what category should continue to be introduced.

[0004] If a large number of random experiments are conducted, there will be problems such as large traffic usage and high experimental costs. It is indeed possible to try to explore whether a specific category is currently out of stock or redundant by increasing or decreasing the distribution rights of a specific category, but there are many categories, and there is actually no prior knowledge of whether to increase or decrease the rights; if you want to obtain relatively stable results, you need to conduct many experiments, and these experiments do not always have positive profit expectations, which is a huge waste of traffic.

[0005] There is also the problem of being unable to judge whether it is the global optimality, because based on a certain experiment, good returns were indeed obtained, but this can only verify whether the experimental strategy at that time is better than the control group. There is no way to know whether there is further room for improvement, which makes it impossible to provide systematic and stable guidance to the business.

[0006] Therefore, how to determine which categories of content need to be supplemented so that the platform can achieve a balance between supply and demand while minimizing the replenishment cost is an urgent problem to be solved in this application. Summary of the invention

[0007] The technical task of the present invention is to provide a short video supply and demand diagnosis method and device to solve the problem of determining which categories of content currently need to be supplemented, so that the platform can achieve supply and demand balance while minimizing the replenishment cost.

[0008] The technical task of the present invention is achieved in the following way: a short video supply and demand diagnosis method is applied to short video recommendation, and a short video replenishment model is constructed by the average short video playback time and the video playback volume of the short video in the short video platform. The short video replenishment model is constructed as follows: S1. Collect supply and demand duration data on the short video platform, calculate the supply-demand ratio based on the collected supply and demand duration data, determine the content value gap based on the supply-demand ratio, and use the out-of-stock duration as a measure of the content value gap; S2. Calculate the video production value using marginal distribution value, where the content value is measured by the incremental time that can be brought to the short video market for each additional short video content introduced or produced; S3. Obtain the optimal short video replenishment plan based on the out-of-stock duration and video production value.

[0009] Furthermore, in step S1, the supply-demand measurement indicator is the supply-demand ratio, and the supply-demand ratio is:

[0010] The supply-demand ratio is a scalar between 0 and 1. The closer it is to 1, the more balanced the supply and demand of the short video platform is; the closer it is to 0, the more serious the supply problem of the short video platform is. Among them, the supply duration is equal to the actual consumption duration, that is, the duration that the short video platform actually satisfies the current users under the current content supply and content distribution conditions; The demand duration is equal to the maximum consumable duration, that is, the upper limit of the duration that the current user may consume on the short video platform, considering neither any restrictions on content supply nor any restrictions on content distribution.

[0011] Furthermore, the out-of-stock duration is to migrate the user group with a low consumption duration to a user group with a higher quantile by adjusting the supply and demand structure, wherein: Out-of-stock duration = (average duration per capita of the high percentile population – average consumption duration per capita) * user size below the high percentile population.

[0012] Furthermore, in step S2, the video production value is P, the marginal distribution value is S, P = (Δ (duration) | Δ (number of replenishment items)) = Δ (number of replenishment items) * * *S; where S = (Δ (duration) | do (number of distributions)); Said For the pool funnel, the To recommend distribution preferences, Δ (the number of replenished contents) is the change in content production, and S is the incremental duration that each additional short video distribution can bring to the short video market.

[0013] Furthermore, the pool entry funnel is that when creators create short video content, the short video content is reviewed manually or by machine, and only high-quality short video content that meets the conditions will enter the recommendation pool; The recommendation distribution preference is to expose different contents based on the estimation of user preferences under the premise of determining the recommendation pool; The pool entry funnel and recommendation distribution preference are considered constants.

[0014] Furthermore, when calculating the marginal distribution value S, the order of video playback needs to be considered. Let the intervention variable T be the number of video playbacks of effective short videos in a certain category, the result variable Y be the sum of the subsequent video playback time brought about by effective playback, and the control variable X be the covariate that needs to be controlled; When the result variable Y is running, first, the short videos played for a certain period of time are defined as the video playback volume of effective short videos, and the short video playback behaviors that the users quickly decline due to lack of interest are eliminated. Then, the playback time of each short video is evenly distributed to the video playback volume of the previous effective short videos, and the sum of the time brought by the video playback volume of each effective short video is the result variable Y.

[0015] Furthermore, in step S3, the optimal short video replenishment solution is: ; Said is the replenishment cost measure, It is a measure of the difference between the category distribution of short video replenishment and the distribution of historical consumption habits of users. The replenishment cost measure is multiplied by the difference between the category distribution of short video replenishment and the distribution of historical consumption habits of users as the optimization target. The smaller the optimization target, the better.

[0016] A short video supply and demand diagnosis device, comprising: at least one memory and at least one processor; The at least one memory is used to store a machine-readable program; The at least one processor is used to call the machine-readable program to execute a short video supply and demand diagnosis method.

[0017] The short video supply and demand diagnosis method and device of the present invention have the following advantages: The present invention not only provides a measure of the current supply and demand situation, but also calculates the duration gap of the short video market, content value measurement, and a cost-effective short video replenishment plan. It is expected to provide a reference for content procurement on short video platforms, improve the overall ROI of content procurement, and enable short video platforms to achieve a balance between supply and demand. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Attached Figure 1 It is a flowchart of a short video supply and demand diagnosis method; Attached Figure 2 Schematic diagram of calculating marginal distribution value in a short video supply and demand diagnosis method. DETAILED DESCRIPTION

[0020] The retail terminal human-computer interaction method, system, device and medium based on speech recognition of the present invention are described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Embodiment 1: like Figure 1 As shown, this embodiment provides a short video supply and demand diagnosis method, the method is specifically as follows: Applied to short video recommendations, a short video replenishment model is constructed through the average short video playback time and the video playback volume of short videos on the short video platform. The short video replenishment model is used to diagnose the supply and demand of short videos and provide the optimal short video replenishment plan.

[0022] The short video replenishment model is constructed as follows: S1. Collect supply and demand duration data on the short video platform, calculate the supply-demand ratio based on the collected supply and demand duration data, determine the content value gap based on the supply-demand ratio, and use the out-of-stock duration as a measure of the content value gap; Among them, the supply and demand measurement indicator is the supply-demand ratio, which is:

[0023] The supply-demand ratio is a scalar between 0 and 1. The closer it is to 1, the more balanced the supply and demand of the short video platform is; the closer it is to 0, the more serious the supply problem of the short video platform is. Supply duration is equal to actual consumption duration, that is, the duration that the short video platform actually satisfies current users under the current content supply and content distribution conditions; The demand duration is equal to the maximum consumable duration, that is, the upper limit of the duration that the current user may consume on the short video platform, considering neither any restrictions on content supply nor any restrictions on content distribution.

[0024] To increase the supply time, you can do so in two ways: (1) When the platform's content supply is stable, improving the distribution mechanism and increasing distribution efficiency will allow the platform's existing content to reach users who are interested in it, thereby increasing user consumption and directly improving the supply-demand ratio indicator.

[0025] (2) When content distribution is consistent, if new content replenishment is of interest to users, it can also increase user consumption and improve the supply-demand ratio indicator.

[0026] like: By circling any supply unit and demand unit pair, we can calculate the supply-demand ratio of this supply-demand pair.

[0027] Supply unit: Videos of the same category can be regarded as a supply unit. On this basis, more features can be added to further refine the supply unit. For example, content of the same category with a video length of less than 5 minutes and an author with more than 100,000 fans can also be regarded as the same supply unit. The core assumption of the supply unit is that the content within the same supply unit is homogeneous, which can help us abstract hundreds of millions of different short videos into dozens or hundreds of supply units.

[0028] Demand unit: Users of the same age, gender, and region can be considered as a demand unit. Similarly, restrictions such as user activity, interest preferences, etc. can be added to further refine the demand unit. The core assumption of the demand unit is that users in the same demand unit are homogeneous.

[0029] Supply-demand ratio calculation: In the supply-demand ratio calculation, the supply time of the short video platform can be directly obtained through historical data statistics, and the demand time of the short video platform needs to be estimated in an appropriate way. We believe that relatively active users (for example, the average time per person is in the top 90 percentile) can be considered as the upper limit of the demand time of a single user. By improving the content ecology, the activity level of relatively inactive users can be raised to the 90th percentile. The specific percentile selection here can be further adjusted according to the business development stage.

[0030] With the supply-demand ratio, we further calculate the supply time and out-of-stock time to migrate the user group with low consumption time to the user group with higher consumption time by adjusting the supply structure. Out-of-stock duration = (average duration per capita of the high percentile population – average consumption duration per capita) * user size below the high percentile population.

[0031] If the high percentile population is the 90th percentile, then: Out-of-stock duration = (average duration of people above the 90th percentile - average consumption duration) * user size below the 90th percentile.

[0032] S2. Calculate the video production value using marginal distribution value, where the content value is measured by the amount of time added to the short video market for each additional short video content introduced or produced; Assume that the video production value is P and the marginal distribution value is S, P = (Δ (duration) | Δ (number of replenishment items)) = Δ (number of replenishment items) * * *S; where S = (Δ (duration) | do (number of distributions)); in, For the pool funnel, is the recommended distribution preference, Δ (the number of replenished content) is the change in content production, S is the time when the platform distributes one more content. After seeing the corresponding content, users will feel satisfied and will browse more content, thus bringing more consumption time. The marginal distribution value refers to how much time increment can be brought to the market by the playback of each additional short video.

[0033] The entry funnel is when creators create short video content, the short video content is reviewed manually or by machine, and only the high-quality short video content that meets the requirements will enter the recommendation pool; Recommended distribution preference is to expose different content to different people based on the estimated user preferences under the premise of determining the recommendation pool; Among them, the pool entry funnel and recommendation distribution preference are difficult to influence from the perspective of replenishment and can be regarded as constants; however, replenishment will indirectly cause the platform to distribute more short video views, thereby affecting the marginal distribution value of the content, which needs to be solved.

[0034] Users will continue to watch short videos only after they like the content. That is, the number of views of the previous short video will affect the subsequent duration. Therefore, the order of video playback needs to be considered in the modeling process.

[0035] Assume that the intervention variable T is the number of video playbacks of effective short videos in a certain category; the result variable Y is the sum of subsequent video playback time brought about by effective playback; the control variable X is the covariate that needs to be controlled, such as the user's age, gender, region, activity level, historical consumption time of a certain category, video playback volume of short videos and other characteristics.

[0036] In this embodiment, the causal inference model Double Machine Learning is used to solve the problem. In this model, the covariates need to be controlled to remain unchanged, and the result variable Y is obtained by intervening the variable T.

[0037] Among them, the result variable Y is: because the user will continue to scroll down after seeing a certain video and feel satisfied, the subsequent video playback behavior can be considered as the result of the combined effect of each previous video playback behavior, and the playback time of each video is evenly distributed to the video playback volume of the previous effective short video. In this embodiment, the video playback volume of the effective short video is defined as the video that has been played for more than three seconds, thereby eliminating the video playback behavior of the user that quickly declines due to lack of interest. The sum of the video playback time brought by each effective short video is the result variable Y.

[0038] like Figure 2 As shown, the examples are 5 short videos in sequence. The video playback duration of the first short video is 10S. Since the above definition states that videos played for more than three seconds are valid short video playbacks, the first short video is a valid short video playback, and the video playback duration it brings is 10S / 1; The second short video has a playback duration of 2 seconds, which is less than 3 seconds. Therefore, it is an invalid short video playback and does not bring any additional subsequent duration. Its own playback duration is brought by the first short video, so the 2S playback volume is added to the first short video playback, expressed as 2S / 1; The video playback time of the third short video is 13S, which is longer than 3S. Therefore, it is a valid short video playback. It is brought by the playback of the first and third short videos. The 13S played is added to the first and third short video playbacks, which is expressed as 13S / 2. The video playback time of the 4th short video is 15S, which is longer than 3S. Therefore, it is a valid short video playback, which is brought about by the first, third and fourth short video playbacks. Therefore, 15S is added to the first, third and fourth short video playbacks respectively, which is expressed as 13S / 3; The video playback time of the 5th short video is 2S, which is less than 3S. Therefore, it is an invalid short video playback. It is caused by the first, third and fourth short video playbacks. Therefore, 2S is added to the first, third and fourth short video playbacks respectively, which is expressed as 2S / 3; Therefore, the resulting variable Y in the example is 10S / 1+2S / 1+13S / 2+15S / 3+2S / 3.

[0039] S3. Obtain the optimal short video replenishment plan based on the out-of-stock duration and video production value.

[0040] The optimal short video replenishment solution is: ; is the replenishment cost measure, The replenishment cost measure is multiplied by the difference measure between the category distribution of short video replenishment and the distribution difference measure of the user's historical consumption habits as the optimization target. The smaller the optimization target, the better, to prevent drastic changes in content distribution due to replenishment, thereby affecting the user experience.

[0041] The following steps can be used as a reference for the implementation of the above short video supply and demand diagnosis and short video replenishment model: Collect the consumption and browsing time of all users on short video products in one day through the client report operated by the user, and store the corresponding data in big data storage engines such as hive, mongodb, redis, etc., storing original user behavior logs (such as consumption time, click-through rate), user portraits (age, gender, region) and content metadata; The clients may include smartphones, desktop computers, tablet computers, laptops, smart devices, etc.

[0042] Utilize the big data operating system within the short video platform, and use SQL or PySpark to process relevant data from the database, including data cleaning, filtering invalid data (such as abnormal short-term clicks), and anonymizing and encrypting user privacy data (such as region, age). In this way, we can obtain supply and demand duration data, and build the supply-demand ratio and out-of-stock duration based on the supply and demand duration data.

[0043] And through SQL or Python language, the user's age, gender, region, consumption and activity in the past 30 days and other data are processed from the database into modeling features, and modeled and solved by Python on Windows or macOS system to obtain the marginal distribution value S of the content. At the same time, the actual indicator values ​​of parameters such as pool entry funnel and recommendation preference are counted from the database to obtain the video production value P. Then, operations planning is solved through Python to obtain the content replenishment plan; The business operator will purchase and replenish content based on the replenishment plan, input the purchased and replenished content into the storage of the short video business backend at one time, use Ab testing to compare user satisfaction before and after replenishment, dynamically adjust model parameters, and finally present it to the client through links such as the pool entry funnel and recommendation preferences.

[0044] Embodiment 2: This embodiment also provides a short video supply and demand diagnosis device, including: at least one memory and at least one processor; The memory is used to store machine-readable programs; The processor is used to call the machine-readable program to execute a short video supply and demand diagnosis method.

[0045] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor may be a microprocessor or any conventional processor, etc.

[0046] The memory can be used to store computer programs and / or modules. The processor realizes various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the terminal, etc. In addition, the memory can also include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash memory card, at least one disk storage period, a flash memory device, or other volatile solid-state storage devices.

[0047] The above-mentioned specific implementations are only specific cases of the present invention. The patent protection scope of the present invention includes but is not limited to the above-mentioned specific implementations. Any technical solutions that conform to the above-mentioned specific implementations of the present invention and any appropriate changes or substitutions made by ordinary technicians in the relevant technical field shall fall within the patent protection scope of the present invention.

[0048] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A short video supply and demand diagnosis method, characterized in that: Applied to short video recommendation, a short video replenishment model is constructed based on the average short video playback time and the video playback volume of the short video on the short video platform. The short video replenishment model is constructed as follows: S1. Collect supply and demand duration data on the short video platform, calculate the supply-demand ratio based on the collected supply and demand duration data, determine the content value gap based on the supply-demand ratio, and use the out-of-stock duration as a measure of the content value gap; S2. Calculate the video production value using marginal distribution value, where the content value is measured by the incremental time that can be brought to the short video market for each additional short video content introduced or produced; S3. Obtain the optimal short video replenishment plan based on the out-of-stock duration and video production value.

2. A short video supply and demand diagnosis method according to claim 1, characterized in that: In step S1, the supply-demand ratio is: ； The supply-demand ratio is a scalar between 0 and 1. The closer it is to 1, the more balanced the supply and demand of the short video platform is; the closer it is to 0, the more serious the supply problem of the short video platform is. Among them, the supply duration is equal to the actual consumption duration, that is, the duration that the short video platform actually satisfies the current users under the current content supply and content distribution conditions; The demand duration is equal to the maximum consumable duration, that is, the upper limit of the duration that the current user may consume on the short video platform, considering neither any restrictions on content supply nor any restrictions on content distribution.

3. A short video supply and demand diagnosis method according to claim 2, characterized in that: The out-of-stock duration is to migrate the user group with low consumption duration to the user group with higher consumption duration by adjusting the supply and demand structure, wherein: Out-of-stock duration = (average duration per capita of the high percentile population – average consumption duration per capita) * user size below the high percentile population.

4. A short video supply and demand diagnosis method according to claim 3, characterized in that: In step S2, the video production value is P, the marginal distribution value is S, P = (Δ (duration) | Δ (number of replenishment items)) = Δ (number of replenishment items) * * *S; where S = (Δ (duration) | do (number of distributions)); Said For the pool funnel, the To recommend distribution preferences, Δ (the number of replenished contents) is the change in content production, and S is the incremental duration that each additional short video distribution can bring to the short video market.

5. A short video supply and demand diagnosis method according to claim 4, characterized in that: The pool entry funnel is when creators create short video content, the short video content is reviewed manually or by machine, and only high-quality short video content that meets the conditions will enter the recommendation pool; The recommendation distribution preference is to expose different contents based on the estimation of user preferences under the premise of determining the recommendation pool; The pool entry funnel and recommendation distribution preference are considered constants.

6. A short video supply and demand diagnosis method according to claim 5, characterized in that: When calculating the marginal distribution value S, the order of video playback needs to be considered. Let the intervention variable T be the number of video playbacks of effective short videos in a certain category, the result variable Y be the sum of the subsequent video playback time brought about by effective playback, and the control variable X be the covariate that needs to be controlled; When the result variable Y is running, first, the short videos played for a certain period of time are defined as the video playback volume of effective short videos, and the short video playback behaviors that the users quickly decline due to lack of interest are eliminated. Then, the playback time of each short video is evenly distributed to the video playback volume of the previous effective short videos, and the sum of the time brought by the video playback volume of each effective short video is the result variable Y.

7. A short video supply and demand diagnosis method according to claim 6, characterized in that: In step S3, the optimal short video replenishment solution is: ； Said is the replenishment cost measure, It is a measure of the difference between the category distribution of short video replenishment and the distribution of historical consumption habits of users. The replenishment cost measure is multiplied by the difference between the category distribution of short video replenishment and the distribution of historical consumption habits of users as the optimization target. The smaller the optimization target, the better.

8. A short video supply and demand diagnosis device, characterized in that: include: at least one memory and at least one processor; The at least one memory is used to store a machine-readable program; The at least one processor is configured to call the machine-readable program to execute the method according to any one of claims 1 to 7.

Images