Server load prediction device, server load prediction method and server load prediction program

Abstract

To provide a server load prediction apparatus, a server load prediction method, and a server load prediction program capable of accurately predicting a load of a server apparatus.SOLUTION: A server load prediction apparatus 100 includes: a content monitoring unit 110 that monitors a content provided on a web by a monitoring object person including an influencer, and acquires at least one of the number of reproduction of a video of the content, the number of retweets of the content, or a comment on the content; a data analysis unit 120 that calculates an evaluation function SCORE including a view score Views according to the number of reproduction of the video per unit time and the number of retweets per unit time, and a word store Words according to a rate of a predetermined word in the comment; and a server load prediction unit 130 that predicts a prediction value of a load of a server apparatus 200 that provides information on products or services on the web based on the evaluation function SCORE.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a server load prediction device, a server load prediction method, and a server load prediction program. [Background technology]

[0002] In recent years, there are many influencers on YouTube, Twitter, Instagram, etc. who have a large number of subscribers and whose comments and tweets have a great impact on society. There is a type of influencer marketing in which a company has its own products introduced by such influencers. One of the challenges is that it is difficult to predict in real time how much server load will be placed on a company's server due to the product information released by the influencer. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2019-215637 A [Patent Document 2] JP 2019-168920 A [Patent Document 3] JP 2019-028538 A [Patent Document 4] JP 2018-025944 A [Patent Document 5] JP 2017-151825 A [Patent Document 6] JP 2016-058005 A Summary of the Invention [Problem to be solved by the invention]

[0004] For example, there is a method for scaling by monitoring trending words in search engines or strings posted by random people on short message posting sites. However, such methods have difficulty in dealing with sudden server loads.

[0005] Trending features such as trending words from search engines and Twitter are not judged until a certain number of accesses are expected. This lag can delay the timing of scaling and cause the site to become overloaded.

[0006] Also, in influencer marketing, the URL of the product being introduced is directly pasted into the video or tweet, so there is a high possibility that people will not use the search terms to search, making it unrealistic to use search engines to determine this.

[0007] There is no resource control method for sudden requests due to product introductions, etc., and a resource control method that takes into account the characteristics of influencers is needed.

[0008] The object of the present disclosure has been made to solve such problems, and is to provide a server load prediction device, a server load prediction method, and a server load prediction program that are capable of predicting the load of a server device with high accuracy. [Means for solving the problem]

[0009] A server load prediction device according to one embodiment includes a content monitoring unit that monitors content provided on the web by monitored individuals, including influencers, and acquires at least one of the number of video views of the content, the number of retweets of the content, and comments on the content; a data analysis unit that calculates an evaluation function including a view score corresponding to the number of video views per unit time and the number of retweets per unit time, and a word score corresponding to a proportion of predetermined words in the comments; and a server load prediction unit that predicts a predicted value of the load of a server device that provides information on products or services on the web based on the evaluation function.

[0010] A server load prediction method according to one embodiment includes a content monitoring step of monitoring content provided on the web by monitored persons including influencers and acquiring at least one of the number of video views of the content, the number of retweets of the content, and comments on the content; a data analysis step of calculating an evaluation function including a view score corresponding to the number of video views per unit time and the number of retweets per unit time, and a word score corresponding to a proportion of predetermined words in the comments; and a server load prediction step of predicting a predicted value of the load of a server device that provides information on products or services on the web based on the evaluation function.

[0011] A server load prediction program in one embodiment causes a computer to execute a content monitoring procedure that monitors content provided on the web by monitored persons including influencers, and obtains at least one of the number of video views of the content, the number of retweets of the content, and comments on the content; a data analysis procedure that calculates an evaluation function that includes a view score corresponding to the number of video views per unit time and the number of retweets per unit time, and a word score corresponding to a proportion of predetermined words in the comments; and a server load prediction procedure that predicts a predicted value of the load of a server device that provides information on products or services on the web, based on the evaluation function. Effect of the Invention

[0012] According to the present disclosure, it is possible to provide a server load prediction device, a server load prediction method, and a server load prediction program capable of predicting the load of a server device with high accuracy. [Brief description of the drawings]

[0013] [Figure 1] 1 is a block diagram illustrating a server load prediction device according to an embodiment; [Diagram 2] 1 is a flowchart illustrating a server load prediction method using a server load prediction device according to an embodiment. [Diagram 3] 1 is a block diagram illustrating a server load prediction device according to a first embodiment. [Figure 4] 4 is a flowchart illustrating a server load prediction method using the server load prediction device according to the first embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, the embodiments will be described with reference to the drawings. For clarity of explanation, the following description and drawings are omitted and simplified as appropriate. In addition, in each drawing, the same elements are given the same reference numerals, and duplicate explanations are omitted as necessary.

[0015] (Overview of the embodiment) First, an overview of the embodiment will be described. Fig. 1 is a block diagram illustrating a server load prediction device according to the embodiment. As shown in Fig. 1, the server load prediction device 100 includes a content monitoring unit 110, a data analysis unit 120, and a server load prediction unit 130. The content monitoring unit 110, the data analysis unit 120, and the server load prediction unit 130 each have the functions of a content monitoring means, a data analysis means, and a server load prediction means.

[0016] The content monitoring unit 110 monitors content provided on the web by a person to be monitored. The person to be monitored includes an influencer. The content includes posting sites and distribution sites such as YouTube, Twitter, and Instagram. Thus, a person to be monitored such as an influencer provides content related to a product or service. The content monitoring unit 110 then acquires at least one of the number of video views of the content, the number of retweets of the content, and comments on the content.

[0017] The data analysis unit 120 calculates a view score according to the number of video views per unit time and the number of retweets per unit time. The data analysis unit 120 also calculates a word score according to the ratio of a predetermined word in the comments. The view score and the word score are collectively called an evaluation function. Therefore, the evaluation function includes the view score and the word score.

[0018] The server load prediction unit 130 predicts the load of a server device that provides information on a product or service on the web, based on the calculated evaluation function.

[0019] 2 is a flowchart illustrating a server load prediction method using the server load prediction device 100 according to the embodiment. As shown in FIG 2, the server load prediction method includes a content monitoring step (step S10), a data analysis step (step S20), and a server load prediction step (step S30).

[0020] As shown in step S10, in the content monitoring step, the content provided on the web by the monitoring target person including the influencer is monitored, and at least one of the number of video views of the content, the number of retweets of the content, and the comments on the content is acquired.

[0021] Next, as shown in step S20, in the data analysis step, a view score is calculated according to the number of video views per unit time and the number of retweets per unit time. Also, a word score is calculated according to the ratio of a predetermined word in the comments. In this way, an evaluation function including the view score and the word score is calculated.

[0022] Next, as shown in step S30, in a server load prediction step, the load of a server device that provides information on a product or service on the web is predicted based on the calculated evaluation function.

[0023] The server load prediction method may be implemented as a program and executed by a computer. According to this embodiment, the load of a server device can be predicted with high accuracy.

[0024] (Embodiment 1) Next, the server load prediction device 100 and the server load prediction method will be described in more detail. FIG. 3 is a block diagram illustrating the server load prediction device 100 according to the first embodiment. As shown in FIG. 3, the server load prediction device 100 is connected to the server device 200 in an information transferable state via a wireless or wired communication line. For example, the server load prediction device 100 is connected to the server device 200 via the Internet. For example, the server device 200 is on the cloud or on-premise. The server device 200 provides product or service information on the web. The server device 200 includes a processing device 220 that serves as a resource and a resource adjustment unit 210 that adjusts the resource. In addition, the server load prediction device 100 monitors the content 300 of a person to be monitored, such as an influencer, via the Internet. A web server device on which the person to be monitored provides the content 300 on the web is different from the server device 200.

[0025] The server load prediction device 100 includes a content monitoring unit 110, a data analysis unit 120, and a server load prediction unit 130, as well as a scaling command unit 140 and a learning unit 150. The server load prediction device 100 may further include a monitoring target memory unit 111, a product memory unit 112, a data memory unit 121, a server load prediction memory unit 131, a resource information monitoring unit 141, and a history memory unit 151.

[0026] The scaling command unit 140, the learning unit 150, the monitored person memory unit 111, the product memory unit 112, the data memory unit 121, the server load prediction memory unit 131, the resource information monitoring unit 141, and the history memory unit 151 each have the functions of a scaling command means, a learning means, a monitored person memory means, a product memory means, a data storage means, a server load prediction memory means, a resource information monitoring means, and a history memory means.

[0027] The server load prediction device 100 may be, for example, a computer including a CPU (Central Processing Unit), a microcomputer, etc. The server load prediction device 100 has a function as a calculation device that performs control processing, calculation processing, and the like.

[0028] The server load prediction device 100 may have a storage unit and an interface unit, not shown. The storage unit may have, for example, a storage device such as a memory or a hard disk. The storage device is, for example, a ROM (Read Only Memory) or a RAM (Random Access Memory). The storage unit has a function for storing control programs, calculation programs, etc. The storage unit also has a function for temporarily storing processing data, etc. The interface unit is, for example, a user interface. The interface unit has an input device such as a keyboard, a touch panel, or a mouse, and an output device such as a display or a speaker. The interface unit accepts data input operations by a user (operator, etc.), and outputs information to the user.

[0029] Each component of the server load prediction device 100 can be realized, for example, by executing a program. More specifically, each component can be realized by executing a program stored in a storage unit. Furthermore, each component may be realized by recording a necessary program on an arbitrary non-volatile recording medium and installing it as necessary. Furthermore, each component is not limited to being realized by software using a program, and may be realized by any combination of hardware, firmware, and software.

[0030] The content monitoring unit 110 acquires information on monitored persons such as influencers from the monitored person storage unit 111. The monitored person storage unit 111 stores the information on the monitored persons. Table 1 is a table illustrating information stored in the monitored person storage unit 111 of the server load prediction device 100 according to the first embodiment.

[0031] [Table 1]

[0032] As shown in Table 1, the monitoring target memory unit 111 stores the monitoring target, the number of subscribers, content 300, coefficient α, and coefficient β. The monitoring target is, for example, influencer A. The number of subscribers to influencer A's content 300 is 9 million. Furthermore, influencer A's content 300 is YouTube. For example, influencer A provides content related to a product or service. The coefficient α and coefficient β of influencer A are 0.3 and 0.7, respectively. The coefficient α and coefficient β will be described later.

[0033] The content monitoring unit 110 monitors the content 300 stored in the monitoring target storage unit 111. The content monitoring unit 110 monitors the content 300 provided by the monitoring target on the web via a network. The content monitoring unit 110 acquires the number of video plays of the content 300, the number of retweets of the content 300, and comments on the content 300. The content monitoring unit 110 also monitors the content 300 of the monitoring target at a specified time interval. For example, the content monitoring unit 110 acquires the number of video plays of the content 300 and the like at 400 second intervals. Note that the information acquired by the content monitoring unit 110 by monitoring the content 300 is not limited to the number of video plays, the number of retweets, and the content. Also, the time interval for acquisition is not limited to 400 second intervals.

[0034] Furthermore, the content monitoring unit 110 acquires information on products or services from the product storage unit 112. The product storage unit 112 stores information on products or services.

[0035] The data analysis unit 120 calculates the number of video plays of the content 300 per unit time, and the number of retweets of the content 300 per unit time.

[0036] The data analysis unit 120 also calculates the ratio of interest and concern in the comment section for the content 300. One method for calculating the ratio of interest and concern in the comment section is, for example, Bayesian classification. Bayesian classification is a classifier that classifies a certain sentence into several classes. For example, the data analysis unit 120 uses Bayesian classification when classifying comments into whether they have interest and concern. It is preferable to have the learning data learn beforehand. For example, the data analysis unit 120 classifies a comment "I would like to try this product A" as having interest and concern. On the other hand, the data analysis unit 120 classifies a comment "Please press like on this comment" as having no interest and concern. In this way, the data analysis unit 120 learns comments on the content 300. In other words, if a comment contains a predetermined word such as "product A" or "I would like to try it," it is classified as having interest and concern. On the other hand, if the word is not related to the content 300, it is classified as having low interest and concern. In this way, each word is given a feature value and classified. In this manner, the data analysis section 120 calculates the ratio of a predetermined word to all comments per unit time.

[0037] The data analysis unit 120 also calculates an evaluation function SCORE. The evaluation function SCORE includes a view score Views according to the number of video plays per unit time and the number of retweets per unit time, and a word score Words according to the proportion of predetermined words in comments.

[0038] Table 2 is a table illustrating the view scores Views and the word scores Words stored in the data storage unit 121 of the server load prediction device 100 according to the first embodiment.

[0039] [Table 2]

[0040] As shown in Table 2, the data storage unit 121 stores the correspondence between the number of video playbacks per unit time and the number of retweets per unit time, and the view score Views, as well as the correspondence between the ratio of a predetermined word of interest and the word score Words. The view score Views is 10, 11 to 30, 31 to 60, and 61 to 100 when the number of video playbacks per unit time (seconds) is 1 to 50, 51 to 100, 101 to 150, and more than 151, respectively. The same applies to the number of retweets. The word score Words is 10, 20 to 30, 40 to 70, and 71 to 100 when the ratio of the word of interest is 0.01 to 0.1, 0.11 to 0.4, 0.41 to 0.7, and more than 0.7, respectively.

[0041] The data analysis unit 120 calculates the evaluation function SCORE using the view score Views and word score Words stored in the data storage unit 121. The data analysis unit 120 also calculates the evaluation function SCORE using the coefficients α and β stored in the monitoring target storage unit 111. The data analysis unit 120 calculates the evaluation function SCORE using the following formula (1). Here, the evaluation function SCORE is CORE, the view score Views is Views, the word score Words is Words, the coefficient α is 0≦α≦1, and the coefficient β is 0≦β≦1.

[0042] SCORE = α × Views + β × Words (1)

[0043] Table 3 is a table illustrating the monitoring count, the elapsed time (seconds), the playback count (times), and the rate of words of interest stored in the data storage unit 121 of the server load prediction device 100 according to the first embodiment.

[0044] [Table 3]

[0045] As shown in Table 3, when the number of monitoring times is 1, the elapsed time (seconds), the number of playbacks (times), and the ratio of words of interest are, for example, 400, 10000, and 0.3. When the number of monitoring times is 2, the elapsed time (seconds), the number of playbacks (times), and the ratio of words of interest are, for example, 300, 50000, and 0.5. For example, when the number of monitoring times in Table 3 is 1, the number of playbacks per unit time is 25, so from Table 2, the view score is Views=10. Also, since the ratio of words of interest is 0.3, from Table 2, the word score is Words=30. Therefore, using the coefficients α and β in Table 1, the value of the evaluation function SCORE when the number of monitoring times in Table 3 is 1 is SCORE=0.3×10+0.7×30=24. In addition, in the case of the first monitoring, the elapsed time is the time from when the content is transmitted.

[0046] The server load prediction unit 130 predicts a predicted value of the load of the server device 200 that provides information on products or services on the web, based on the evaluation function SCORE. For example, the server load prediction unit 130 predicts the load of the server device 200 by comparing the calculated evaluation function SCORE with a predicted value stored in the server load prediction storage unit 131.

[0047] Table 4 is a table illustrating the predicted values ​​stored in the server load prediction storage unit 131 in the server load prediction device 100 according to the first embodiment.

[0048] [Table 4]

[0049] As shown in Table 4, the server load prediction storage unit 131 stores the correspondence between the evaluation function SCORE and the predicted value. For example, when the evaluation function SCORE value is 0 to 20, 21 to 80, 81 to 110, 111 to 140, or 141 or more, the predicted value 1 is 100, 120, 180, 240, and 300, and the predicted value 2 is 40, 100, 160, 220, and 260.

[0050] The server load prediction unit 130 predicts a predicted value by using the correspondence relationship stored in the server load prediction storage unit 131. For example, when the value of the evaluation function SCORE described above is 24, the server load prediction unit 130 predicts the predicted value of the load of the server device to be 120%.

[0051] The resource information monitor 141 monitors the current configuration of the processing device 220 serving as a resource of the server device 200 and the load on the server device 200 .

[0052] The scaling command unit 140 transmits a command to change the scale to the server device 200 based on the predicted value predicted by the server load prediction unit 130. For example, when the predicted value is 120%, the scaling command unit 140 transmits a command to the resource adjustment unit 210 of the server device 200 to scale up the resources by 20% or more than the current resources.

[0053] Specifically, if the predicted value indicates a load of 120%, the scaling command unit 140 sends a command to the resource adjustment unit 210 of the server device 200 to change the vertical or horizontal scale, such as scaling up a CPU from 3 GHz to 3.6 GHz.

[0054] Here, vertical scaling collectively refers to scale-up, which improves the performance of the server device 200, such as the CPU and memory, and scale-down, which reduces the performance of the server device 200. Thus, vertical scaling is a technique for manipulating the performance of the server device 200 itself. On the other hand, horizontal scaling is used when using cloud resources, etc., and collectively refers to scale-out, which increases the number of processing devices 220 for the purpose of distributing the processing load, and scale-in, which reduces the number of processing devices 220 for the purpose of reducing costs, etc. Thus, horizontal scaling refers to increasing processing performance by running the same Web service on a large number of processing devices 220, even if each one does not have high performance.

[0055] The resource information monitoring unit 141 obtains the current configuration of the server device 200 (such as the metrics and number of each processing device 220) in the cloud or on-premise after the scaling command unit 140 issues a command to change the scale and changes the configuration of the server device 200.

[0056] The learning unit 150 learns the coefficients α and β from the error between the predicted value predicted by the server load prediction unit 130 and the actual load. Then, the learning unit 150 updates the coefficients α and β stored in the monitored person memory unit 111. Specifically, the learning unit 150 learns the error between the predicted load of the server device 200 and the actual load of the server device 200 within a specified time interval. Then, the learning unit 150 updates the values ​​of the coefficients α and β from the learned error. This is because the load of the server device 200 differs depending on the characteristics of the monitored person, and therefore the load of the server device 200 is predicted by incorporating the characteristics of the monitored person.

[0057] For example, depending on the influencer to be monitored, there may be a case where the number of views is large, but many viewers are not interested. In this case, the load on the server device 200 is small. In this way, when the load on the server device 200 is smaller than expected, the coefficient β is weighted.

[0058] In addition, there may be cases where the number of comments showing interest and the number of views are large, but the server load is smaller than expected. In the case of an influencer with such characteristics, the weights of both the coefficient α and the coefficient β are made small.

[0059] For example, the predicted value of the load of the server device 200 described above is 120%. However, if the actual result is 90%, the coefficient α and the coefficient β are updated. For example, assume that the current coefficients are coefficient α=0.4 and coefficient β=0.6. Then, it is determined that the weights of the interest and concern of the comment are too high based on the values ​​of Views=10 and Words=30. Therefore, the coefficient α=0.4 and the coefficient β=0.6 are changed to coefficient α=0.6 and the coefficient β=0.4. As a result, the predictor value of the load of the server device 200 becomes 0.6×10+0.4×30=18. Therefore, the predicted value can be reduced to 100%. In addition, such a learning result of the learning unit 150 may be reflected in the coefficient α and the coefficient β stored in the monitoring target memory unit 111. The learning unit 150 stores the learning result in the history memory unit 151.

[0060] The history storage unit 151 stores the history of each parameter including the predicted value predicted by the server load prediction unit 130, the actual load, the error between the predicted value and the actual load, the results of learning by the learning unit 150, the coefficients α and β at that time, etc. When the learning unit 150 updates the coefficients α and β, the learning unit 150 refers to the prediction value, the actual load, the error, and the coefficients α and β at that time stored in the history storage unit 151, and changes the coefficients α and β to appropriate ones.

[0061] Next, the operation of the server load prediction device will be described. Fig. 4 is a flow chart illustrating a server load prediction method using the server load prediction device 100 according to the first embodiment. As shown in Fig. 4, the server load prediction method includes a monitoring target storing step (step S5), a content monitoring step (step S10), a data storing step (step S15), a data analyzing step (step S20), a server load prediction storing step (step S25), a server load prediction step (step S30), a scaling step (step S35), and a learning step (step S40).

[0062] The order of each step is not limited to this. For example, the scaling step (step S35) and the learning step (step S40) may be performed either before or at the same time as long as they are performed after the server load prediction step (step S30). Also, the server load prediction storage step (step S25) may be performed before or at the same time as the data storage step (step S15) as long as it is performed before the server load prediction step (step S30).

[0063] As shown in step S5, in a monitoring target storage step, the content 300 of the monitoring target including the influencer, the number of registered users, the coefficient α and the coefficient β, etc. are stored in the monitoring target storage unit 111.

[0064] Next, as shown in step S10, in a content monitoring step, the content 300 provided on the web by the person to be monitored is monitored. Here, the content 300 is the content 300 stored in the person to be monitored storage unit 111. In addition, the number of video plays of the content 300, the number of retweets of the content 300, and comments on the content 300 are acquired.

[0065] Next, as shown in step S15, in a data storage step, the data storage unit 121 stores the correspondence between the number of video plays and retweets and the view score, as well as the correspondence between the proportion of a predetermined word and the word score.

[0066] Next, as shown in step S20, in a data analysis step, an evaluation function SCORE is calculated, which includes a view score Views according to the number of video views per unit time and the number of retweets per unit time, and a word score Words according to the proportion of a predetermined word in the comments. For example, the evaluation function SCORE is calculated by the above-mentioned formula (1). In the data analysis step, the evaluation function SCORE may be calculated using the view score Views and the word score Words stored in the data storage unit 121. In addition, in the data analysis step, the evaluation function SCORE may be calculated using the coefficient α and coefficient β stored in the monitored person storage unit 111.

[0067] Next, as shown in step S25, in a server load prediction storage step, the correspondence between the evaluation function SCORE and the predicted value is stored in the server load prediction storage unit 131.

[0068] Next, as shown in step S30, in a server load prediction step, a predicted value of the load of the server device 200 that provides information on products or services on the web is predicted based on the evaluation function SCORE. In the server load prediction step, the predicted value of the load of the server device 200 may be predicted using the correspondence between the evaluation function SCORE stored in the server load prediction storage unit 131 and the predicted value.

[0069] Next, as shown in step S35, in a scaling step, a command is issued to the server device 200 to change the scale based on the predicted value. For example, when the predicted value is 120%, a command is sent to the resource adjustment unit 210 of the server device 200 to scale up the current resources by 20% or more.

[0070] Next, as shown in step S40, in a learning step, the coefficients α and β are learned from the error between the predicted value and the actual load. Then, the coefficients α and β stored in the monitoring target storage unit 111 are updated. In this manner, the load of the server device 200 can be predicted.

[0071] Next, the effects of this embodiment will be described. The server load prediction device 100 of this embodiment monitors the content 300 of a monitoring target such as an influencer who is a cause of sudden requests. Therefore, it is possible to predict the load of the server device 200, which changes suddenly in real time, with high accuracy.

[0072] In addition, by using the evaluation function SCORE that utilizes the number of video views, the number of retweets, and the interest of comments, it is possible to predict the load on the server device 200 according to the characteristics of the influencer, etc. Furthermore, since the coefficients α and β used in calculating the evaluation function SCORE are trained according to the characteristics of the influencer, etc., the prediction accuracy of the server device 200 can be improved.

[0073] (Embodiment 2) Next, a server load prediction device according to a second embodiment will be described. The server load prediction device of this embodiment can be applied to the load of the server device 200, which is difficult to predict. For example, in the first embodiment, the background is influencer marketing, and the target monitor is determined. However, the server load prediction device of this embodiment periodically monitors the contents of all influencers related to the company's products or services in the world. This makes it possible to deal with the access load of the company's server device 200 and the down of the server device 200 caused by the introduction of products, etc. without the company's involvement.

[0074] Another application is to monitor influencers in the world and find influencers who are compatible with one's company's products or services based on the value of the evaluation function SCORE. This utilizes the fact that the value of the evaluation function SCORE indicates the influence of the influencer's product PR (whether it leads to purchases, etc.). For example, a product related to a product that one's company wants to introduce is targeted, and the content of multiple influencers who are promoting the product is monitored. Specifically, the content monitoring unit 110 monitors the content of multiple influencers, and the data analysis unit 120 calculates the evaluation function of each influencer. Then, by calculating each evaluation function SCORE and finding influencers with high evaluation function SCORE values, it is possible to determine which influencers are effective in promoting one's company's products.

[0075] The present invention is not limited to the above-described embodiment, and can be modified as appropriate without departing from the spirit and scope of the present invention. For example, an embodiment that combines the configurations of the first and second embodiments is also included in the scope of the technical idea.

[0076] A part or all of the above-described embodiments can be described as, but is not limited to, the following supplementary notes.

[0077] (Appendix 1) A content monitoring step of monitoring content provided on the web by a person to be monitored, including an influencer, and acquiring at least one of the number of video views of the content, the number of retweets of the content, and comments on the content; A data analysis step of calculating an evaluation function including a view score according to the number of video views per unit time and the number of retweets per unit time, and a word score according to a ratio of a predetermined word in the comments; a server load prediction step of predicting a load of a server device that provides information on a product or service on the web based on the evaluation function; A server load prediction method comprising: (Appendix 2) In the data analysis step, The evaluation function is calculated by the following formula (A), where the evaluation function is SCORE, the view score is Views, the word score is Words, the first coefficient is α with 0≦α≦1, and the second coefficient is β with 0≦β≦1: SCORE = α × Views + β × Words (A) 2. A server load prediction method according to claim 1. (Appendix 3) A data storage step of storing the correspondence between the number of video views and the number of retweets and the view score, and the correspondence between the percentage of the predetermined word and the word score in a data storage unit, In the data analysis step, Calculating the evaluation function using the view scores and the word scores stored in the data storage unit. 3. A server load prediction method according to claim 2. (Appendix 4) a server load prediction storage step of storing the correspondence between the evaluation function and the predicted value in a server load prediction storage unit, In the server load prediction step, predicting the predicted value using the correspondence relationship stored in the server load prediction storage unit; 4. A server load prediction method according to claim 2 or 3. (Appendix 5) The content of the influencer, the first coefficient, and the second coefficient are stored in a monitoring target storage unit, In the content monitoring step, monitor the content stored in the monitoring target storage unit; In the data analysis step, Calculating the evaluation function using the first coefficient and the second coefficient stored in the monitoring target storage unit. A server load prediction method according to any one of appendix 2 to 4. (Appendix 6) a learning step of learning the first coefficient and the second coefficient from an error between the predicted value predicted in the server load prediction step and an actual load, In the learning step, updating the first coefficient and the second coefficient stored in the monitoring target storage unit; 6. A server load prediction method according to claim 5. (Appendix 7) The method further includes a scaling instruction step of instructing the server device to change a scale based on the predicted value predicted in the server load prediction step. 7. A server load prediction method according to any one of claims 1 to 6. (Appendix 8) The influencer provides the content related to the product or service; A server load prediction method according to any one of appendix 1 to 7. (Appendix 9) In the content monitoring step, monitoring said content of a plurality of said influencers; In the data analysis step, Calculating the evaluation function for each influencer; A server load prediction method according to any one of appendix 1 to 8. (Appendix 10) a web server device on which the monitored person provides the content on the web is different from the server device on which the information on the product or service is provided on the web; A server load prediction method according to any one of appendixes 1 to 9. (Appendix 11) A content monitoring step of monitoring content provided on the web by a person to be monitored, including an influencer, and acquiring at least one of the number of video views of the content, the number of retweets of the content, and comments on the content; A data analysis procedure for calculating an evaluation function including a view score according to the number of video views per unit time and the number of retweets per unit time, and a word score according to a ratio of a predetermined word in the comment; a server load prediction step of predicting a load of a server device that provides information on a product or service on the web based on the evaluation function; A server load prediction program that causes a computer to execute the above. (Appendix 12) In the data analysis step, The evaluation function is calculated by the following formula (A), where the evaluation function is SCORE, the view score is Views, the word score is Words, the first coefficient is α with 0≦α≦1, and the second coefficient is β with 0≦β≦1: SCORE = α × Views + β × Words (A) Make the computer do that, 12. A server load prediction program according to claim 11. (Appendix 13) The method further includes a data storage step of storing, in a data storage unit, a correspondence relationship between the number of video views and the number of retweets, and the view score, and a correspondence relationship between the percentage of the predetermined word and the word score; In the data analysis step, Calculating the evaluation function using the view scores and the word scores stored in the data storage unit. Make the computer do that, 13. A server load prediction program according to claim 12. (Appendix 14) a server load prediction storage step of storing the correspondence between the evaluation function and the predicted value in a server load prediction storage unit; In the server load prediction procedure, predicting the predicted value using the correspondence relationship stored in the server load prediction storage unit; Make the computer do that, 14. A server load prediction program according to claim 12 or 13. (Appendix 15) The content of the influencer, the first coefficient, and the second coefficient are stored in a monitoring target storage unit, In the content monitoring step, monitor the content stored in the monitoring target storage unit; In the data analysis step, Calculating the evaluation function using the first coefficient and the second coefficient stored in the monitoring target storage unit. Make the computer do that, 15. A server load prediction program according to any one of appendixes 12 to 14. (Appendix 16) a learning step of learning the first coefficient and the second coefficient from an error between the predicted value predicted in the server load prediction step and an actual load, In the learning procedure, updating the first coefficient and the second coefficient stored in the monitoring target storage unit; Make the computer do that, 16. A server load prediction program according to claim 15. (Appendix 17) a scaling instruction step of instructing the computer to change a scale of the server device based on the predicted value predicted in the server load prediction step; 17. A server load prediction program according to any one of claims 11 to 16. (Appendix 18) The influencer provides the content related to the product or service; 18. A server load prediction program according to any one of claims 11 to 17. (Appendix 19) In the content monitoring step, monitoring said content of a plurality of said influencers; In the data analysis step, Calculating the evaluation function for each influencer; Make the computer do that, 19. A server load prediction program according to any one of appendixes 11 to 18. (Appendix 20) a web server device on which the monitored person provides the content on the web is different from the server device on which the information on the product or service is provided on the web; 20. A server load prediction program according to any one of appendixes 11 to 19.

[0078] The program includes a set of instructions (or software code) that, when loaded into a computer, causes the computer to perform one or more functions described in the embodiments. The program may be stored in a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, the computer-readable medium or tangible storage medium includes random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technology, CD-ROM, digital versatile disc (DVD), Blu-ray® disk or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The program may be transmitted on a transitory computer-readable medium or communication medium. By way of example and not limitation, the transitory computer-readable medium or communication medium includes electrical, optical, acoustic, or other forms of propagated signals. [Explanation of symbols]

[0079] 100 Server Load Prediction Device 110 Content Monitoring Department 111 Surveillance target memory unit 112 Product storage section 120 Data Analysis Department 121 Data storage unit 130 Server Load Prediction Department 131 Server load prediction memory unit 140 Scaling Command Section 141 Resource Information Monitoring Unit 150 Learning Department 151 History memory unit 200 Server device 210 Resource Coordination Department 220 Processing equipment 300 Contents

Claims

1. A content monitoring unit that monitors content provided on the web by a monitoring target person, including an influencer, and acquires at least one of the number of video views of the content, the number of retweets of the content, and comments on the content; A data analysis unit that calculates an evaluation function including a view score according to the number of video views per unit time and the number of retweets per unit time, and a word score according to a ratio of a predetermined word in the comment; a server load prediction unit that predicts a load prediction value of a server device that provides information on a product or service on the web based on the evaluation function; A server load prediction device comprising:

2. When the evaluation function is SCORE, the view score is Views, the word score is Words, a first coefficient is α with 0≦α≦1, and a second coefficient is β with 0≦β≦1, the data analysis unit calculates the evaluation function by the following formula (A): SCORE=α×Views+β×Words (A) The server load prediction device according to claim 1 .

3. A data storage unit stores a correspondence between the number of video views and the number of retweets and the view score, and a correspondence between a ratio of the predetermined word and the word score, The data analysis unit calculates the evaluation function by using the view scores and the word scores stored in the data storage unit. The server load prediction device according to claim 2.

4. a server load prediction storage unit that stores a correspondence relationship between the evaluation function and the predicted value; the server load prediction unit predicts the predicted value by using the correspondence relationship stored in the server load prediction storage unit.

4. The server load prediction device according to claim 2.

5. A monitoring target storage unit that stores the content of the influencer, the first coefficient, and the second coefficient, The content monitoring unit monitors the content stored in the monitoring target storage unit, The data analysis unit calculates the evaluation function by using the first coefficient and the second coefficient stored in the monitoring target storage unit. The server load prediction device according to any one of claims 2 to 4.

6. a learning unit that learns the first coefficient and the second coefficient from an error between the predicted value predicted by the server load prediction unit and an actual load, The learning unit updates the first coefficient and the second coefficient stored in the monitoring target storage unit. The server load prediction device according to claim 5.

7. The server load prediction unit further includes a scaling command unit that commands the server device to change a scale based on the predicted value predicted by the server load prediction unit. The server load prediction device according to any one of claims 1 to 6.

8. The influencer provides the content related to the product or service; The server load prediction device according to any one of claims 1 to 7.

9. A content monitoring step of monitoring content provided on the web by a person to be monitored, including an influencer, and acquiring at least one of the number of video views of the content, the number of retweets of the content, and comments on the content; A data analysis step of calculating an evaluation function including a view score according to the number of video views per unit time and the number of retweets per unit time, and a word score according to a ratio of a predetermined word in the comments; a server load prediction step of predicting a load of a server device that provides information on a product or service on the web based on the evaluation function; A server load prediction method comprising:

10. A content monitoring step of monitoring content provided on the web by a person to be monitored, including an influencer, and acquiring at least one of the number of video views of the content, the number of retweets of the content, and comments on the content; A data analysis procedure for calculating an evaluation function including a view score according to the number of video views per unit time and the number of retweets per unit time, and a word score according to a ratio of a predetermined word in the comments; a server load prediction step of predicting a load of a server device that provides information on a product or service on the web based on the evaluation function; A server load prediction program that causes a computer to execute the above.

Images