Prediction device, prediction method, and prediction program

Abstract

Provided are a prediction device, a prediction method, and a prediction program with which it is possible to predict an audience rating with high accuracy. A prediction device according to one embodiment of the present invention is characterized by being provided with: a storage unit for storing a reference feature amount for the audience rating of a previous program; a setting unit that sets a request period for at least one of the target programs; an extraction unit that extracts a feature amount of the request period; and an acquisition unit that acquires, on the basis of the reference feature amount and the feature amount of the request period, a predicted feature amount pertaining to the prediction of the audience rating of the request period.

Description

Technical Field

[0001] The present disclosure relates to a prediction device, a prediction method, and a prediction program. Background Art

[0002] The present applicant has applied for a patent for a prediction device for predicting broadcast contact information with high accuracy based on conditions related to broadcast information, and has obtained the patent right (Patent Document 1).

[0003] The prediction device of Patent Document 1 includes a storage unit, a calculation unit, and a prediction unit. The storage unit stores broadcast contact information indicating the degree of contact from the receiver to each of a plurality of broadcast information broadcasted in the broadcast media. The calculation unit calculates at least two or more statistical values ​​among a first statistical value, a second statistical value, and a third statistical value. The prediction unit obtains predicted contact information indicating the degree of contact from the receiver to the broadcast information scheduled to be broadcasted based on at least two or more statistical values. The first statistical value indicates the moving average of the broadcast contact information of each of the broadcast information broadcasted on the multiple days closest to the scheduled broadcast date, among the multiple broadcast information broadcasted by the broadcast media. The second statistical value indicates the moving average of the broadcast contact information of each of the multiple broadcast information broadcasted on the multiple days closest to the scheduled broadcast date, among the multiple broadcast information broadcasted by the broadcast media. The third statistical value indicates the moving average of the broadcast contact information of each of the multiple broadcast information broadcasted on the multiple days closest to the scheduled broadcast date, among the multiple broadcast information broadcasted by the broadcast media, corresponding to the same day of the week as the scheduled broadcast date.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent No. 6902636 Summary of the invention

[0007] Problems to be solved by the invention

[0008] However, in terms of accurately predicting audience ratings, the prior art including Patent Document 1 has room for improvement.

[0009] Therefore, one of the objects of the present disclosure is to provide a prediction device, a prediction method, and a prediction program that can predict audience ratings with high accuracy.

[0010] Technical solutions to solve problems

[0011] A prediction device according to one embodiment of the present invention is characterized in that it comprises: a storage unit that stores baseline feature quantities of audience ratings of previous programs; a setting unit that sets a request time period for at least one target program; an extraction unit that extracts feature quantities of the request time period; and an acquisition unit that acquires predicted feature quantities related to the prediction of the audience rating of the request time period based on the baseline feature quantities and the feature quantities of the request time period.

[0012] Effects of the Invention

[0013] According to one embodiment of the present disclosure, it is possible to provide a prediction device, a prediction method, and a prediction program capable of predicting audience ratings with high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a diagram showing an example of a prediction calculation method in conventional long-term trend prediction.

[0015] Figure 2 This is a diagram showing an example of the relationship between the past audience ratings and the target period, which are explanatory variables in conventional long-term trend prediction.

[0016] Figure 3 This is a diagram showing a first example of the storage process of the reference feature quantity of the audience rating of the past program.

[0017] Figure 4 This is a diagram showing a second example of the storage process of the reference feature quantity of the audience rating of past programs.

[0018] Figure 5 This is a diagram showing an example of obtaining a predicted feature value of a request time period based on a reference feature value and a feature value of a request time period.

[0019] Figure 6 This is a diagram showing an example of a schematic configuration of a system including a prediction device.

[0020] Figure 7 This is a functional block diagram showing an example of the internal structure of the prediction device.

[0021] Figure 8 This is a diagram showing an example of the functional structure of a cost calculation device, a broadcast distribution device, and a prediction device.

[0022] Fig. 9 This is a diagram showing an example of the hardware configuration of a cost calculation device, a broadcast distribution device, and a prediction device. DETAILED DESCRIPTION

[0023] As a structure (model) for predicting the audience rating of a target program, there are, for example, recent trend prediction (SHAREST_RT, Recent Trend Prediction Version) and long-term trend prediction (SHAREST_LT, Long-term Trend Prediction Version).

[0024] Recent trend prediction is a process of predicting the audience rating of a target program that is close to its scheduled broadcast date. For example, it is a process of predicting the audience rating with high accuracy based on detailed data such as program composition, genre, and Internet reviews. The prediction target date (broadcast date of the target program) in recent trend prediction can be, for example, 1 day to 7 days from now (OA 7 days ago to 1 day ago).

[0025] Long-term trend prediction is a process of predicting the ratings of a target program that is far from the scheduled broadcast date. For example, it is a prediction process of medium- and long-term prediction that takes into account the passage of time and seasonality (seasonal changes) of the ratings. The prediction target date (broadcast date of the target program) in the long-term trend prediction can also be, for example, 8 days to 70 days from now (OA70 days ago to 8 days ago).

[0026] In long-term trend prediction, the audience rating of the target program is predicted based on the real-time audience rating (average audience rating) every 10 minutes from 8 days to 70 days from now (OA 70 days ago to 8 days ago). At this time, a specified period, such as weekly learning of the past 2 years (starting at 19:00 every Wednesday) can also be used as learning data. In addition, as the prediction target period, for example, daily predictions (starting at 3:00 every day) can also be used. In this case, the prediction target period can also be from the latest date of the quick report data to 70 days ago. The prediction target period can also be extended.

[0027] Figure 1 This is a diagram showing an example of a prediction calculation method in conventional long-term trend prediction. Figure 1 In the model, a recurrent neural network (RNN) is used to predict the daily ratings of all targets (for example, 31 layers) × all TV stations. Figure 1In the example, when you want to use the measured values ​​up to July 20 of a certain year to predict the audience rating on July 21, you learn the measured values ​​according to the time series and output the predicted values. Specifically, in the time period of 5:00-5:10, based on the measured values ​​up to July 20 (the measured values ​​from July 18 to July 20 in the figure), the predicted value for July 21, which is 0.3, is output. Similarly, in the time period of 5:10-5:20, based on the measured values ​​up to July 20 (the measured values ​​from July 18 to July 20 in the figure), the predicted value for July 21, which is 0.2, is output. In addition, in the time period of 5:20-5:30, based on the measured values ​​up to July 20 (the measured values ​​from July 18 to July 20 in the figure), the predicted value for July 21, which is 0.6, is output. In addition, in the time period of 5:30-5:40, the predicted value for July 21, 0.7, is output based on the actual measured values ​​up to July 20 (the measured values ​​in the figure are from July 18 to July 20). In addition, in the time period of 28:50-29:00, the predicted value for July 21, 0.1, is output based on the actual measured values ​​up to July 20 (the measured values ​​in the figure are from July 18 to July 20).

[0028] Thus, in the previous long-term trend prediction, since the model is used to learn and predict the average ratings every 10 minutes in the past, there is a tendency for the ratings prediction accuracy to deteriorate (there is a divergence between the predicted ratings and the actual ratings) (technical issue).

[0029] Figure 2 This is a diagram showing an example of the relationship between the past ratings and the target period as explanatory variables in the past long-term trend forecast. Figure 2In the example, February 18th of a certain year is set as the prediction base date, and the average ratings in the past up to n days ago are calculated and defined as explanatory variables. Here, it is assumed that the average of n days ago is calculated from the day before the prediction base date, n=28, that is, the average of the previous four weeks (the average of each day regardless of the day of the week) is used for calculation. Specifically, the average ratings in the past from January 21st to February 17th of a certain year are calculated, and defined as explanatory variables for ratings prediction on February 18th. In this way, there is data from the day before the prediction base date based on daily collaboration, and by making use of the most recent data as much as possible, the average ratings in the past are obtained based on the latest data completed in the collaboration at the time of the prediction base date. In addition, in order to take into account recent trends, the previous ratings are calculated starting from the day before the prediction base date. The previous ratings are used as variables for accurately predicting the ratings on the prediction base date. Furthermore, when the prediction object is set to February 19th to April 29th of a certain year, the predicted value from February 18th, which is the prediction base date, to April 29th, 70 days later, can also use the average ratings of the past n days starting from February 17th as an explanatory variable.

[0030] In the previous long-term trend prediction, for example, since part of the OA 60-day-ahead prediction value is wrong, even if it is corrected to a correct value, the prediction accuracy will lose to (be worse than) the average of the previous 8 weeks. In addition, when focusing on the accuracy comparison of the number of minutes of broadcasting of each program, there are cases where the prediction accuracy deteriorates (gets worse) depending on the broadcasting time (for example, in a mini-program of about 6 minutes) compared to the average of the previous 8 weeks. In particular, there is a tendency for the audience rating to be predicted to be higher than the actual rating for the mini-program included in the A time, which is considered to be the best time, such as 19:00 to 23:00 on weekdays and 18:00 to 23:00 on Saturdays and Sundays. It is speculated that this is because the previous long-term trend prediction is a model that learns and predicts the average audience rating every 10 minutes in the past. For the same reason, even in the 56-minute program adjacent to (affected by) the mini-program inserted in the A time, the prediction accuracy of the audience rating tends to deteriorate. In this way, even if the accuracy of the audience rating prediction of the long-term trend prediction is on the rise as a whole, it sometimes cannot keep up with the actual increase. That is, even if a trend is captured, there is a possibility that it is unpredictable in the case of a sudden increase with few previous examples.

[0031] The present inventors have identified the above-mentioned problems as important technical issues, and have conceived a prediction device, a prediction method, and a prediction program having the following components in order to predict the audience rating in particular in long-term trend prediction with high accuracy.

[0032] In order to improve the recurrent neural network (RNN), the weekly time series changes are changed to a model structure that is easy to learn, and the time series trend is changed to a target variable that is easy to learn. In addition, the program time period that is likely to affect the ratings prediction, that is, the request period, is set, and the prediction value corresponding to the request period is corrected. Specifically, the mini program period where the ratings prediction is likely to deteriorate (likely to be estimated to be high) is set as the request period, and the prediction value in the long-term trend prediction is constructed (newly added) based on the feature quantity transformation output of the request period.

[0033] <Storage Processing (Storage Step)>

[0034] The reference feature quantity of the audience rating of the past program is stored. The reference feature quantity may be the audience rating of the past program itself, or may be any parameter related to the audience rating of the past program.

[0035] Figure 3 1 is a diagram showing a first example of a storage process (storage step) of a reference feature value of the audience rating of a previous program. Figure 3 In the model, a recurrent neural network (RNN) is used to predict the daily ratings of all targets (for example, 31 layers) × all stations.

[0036] like Figure 3As shown, in the time periods of 5:00-5:10, 5:10-5:20, and 5:20-5:30 every 10 minutes on Monday of a certain broadcasting station (● broadcasting), the measured values ​​(measured ratings, measured feature quantities) from 20 weeks ago to 1 week ago of the prediction base date (prediction target date) are stored. In addition, in the time periods of 5:00-5:10, 5:10-5:20, and 5:20-5:30 every 10 minutes on Tuesday of a certain broadcasting station (● broadcasting), the measured values ​​(measured ratings, measured feature quantities) from 20 weeks ago to 1 week ago of the prediction base date are stored. Similarly, for Wednesday to Sunday of a certain broadcasting station (● broadcasting), the measured values ​​(measured ratings, measured feature quantities) from 20 weeks ago to 1 week ago of the prediction base date are stored in the time periods of 10 minutes every 10 minutes. Since the prediction base date applies to any day of the week from Monday to Sunday, the time series ratings for the same day of the week as the prediction base date are stored as the benchmark feature values ​​while storing the feature values ​​of the ratings for a specified period in a plurality of past programs. That is, when storing the benchmark feature values, the time series changes for each week are changed to a structure that is easier to learn. In addition, as described in detail later, by multiplying the benchmark feature values ​​by a correction based on the feature values ​​of at least one set request time period for the target program, a prediction feature value related to the prediction of the ratings for the request time period can be obtained. The prediction feature value can be the predicted ratings for the request time period itself, or it can be any parameter associated with the predicted ratings for the request time period. In addition, in Figure 3 In the second week or later, the predicted feature quantity may be added to the feature quantity used for parameter correction to perform recursive prediction (updating of the predicted ratings).

[0037] Figure 4 2 is a diagram showing a second example of the storage process (storage step) of the reference feature quantity of the audience rating of the past program. Figure 4 In the model, a recurrent neural network (RNN) is used to predict the daily ratings of all targets (for example, 31 layers) × all stations.

[0038] like Figure 4As shown, in the time periods of 5:00-5:10, 5:10-5:20, and 5:20-5:30 every 10 minutes on Monday of a certain broadcasting station (● broadcast), the measured values ​​(measured ratings, measured feature quantities) from 20 weeks ago to 1 week ago of the prediction base date (prediction target date) are stored. Similarly, for Tuesday to Sunday of a certain broadcasting station (● broadcast), the measured values ​​(measured ratings, measured feature quantities) from 20 weeks ago to 1 week ago of the prediction base date (prediction target date) are also stored in the time periods of 10 minutes every 10 minutes. These measured values ​​can be set to include the difference between the weekly ratings of the prediction base date and the 24-week exponential moving average ratings of the prediction base date. In addition, the standard value of the difference in the measured values ​​is not limited to the 24-week exponential moving average ratings, but can also be the 4-week exponential moving average ratings, the 8-week exponential moving average ratings, the 12-week exponential moving average ratings, and the 48-week exponential moving average ratings. In this way, feature quantities of ratings for a specified period covering multiple past programs are stored, and the difference between the ratings for the week including the prediction base date and the exponential moving average ratings for the specified period before the prediction base date is stored as a baseline feature quantity. That is, the trend of the time series is changed to a target variable that is easy to learn. In addition, as described in detail later, by multiplying the baseline feature quantity by a correction based on the feature quantity of at least one set request time period for the target program, a predicted feature quantity related to the prediction of the ratings for the request time period can be obtained. The predicted feature quantity can be the predicted ratings for the request time period itself, or it can be any parameter associated with the predicted ratings for the request time period. In addition, in Figure 4 In the second week or later, the predicted feature quantity may be added to the feature quantity used for parameter correction to perform recursive prediction (updating of the predicted ratings).

[0039] Here, the past ratings, which are an example of the reference feature quantity of the ratings of past programs, include, for example, the average ratings of the program before a predetermined period, the ratings at the end of a predetermined period, the average ratings on the same day of the previous year, and the average ratings for the same time period of the previous year. The past ratings also include the latest moving average ratings, the moving average ratings on the same day of the week, the missing flag 1, the moving average ratings on the same day of the week and the same time period, and the missing flag 2. The latest moving average ratings, the moving average ratings on the same day of the week, and the moving average ratings on the same day of the week and the same time period are examples of statistical values ​​of broadcast contact information.

[0040] The average ratings of programs before the specified period indicate the average ratings of programs in the main time slot, the next time slot, the previous time slot, and the previous and next time slots that were broadcast before the specified period on the forecast target day. The ratings at the end of the specified period before the specified period indicate the ratings at the end of the programs in the main time slot that were broadcast before the specified period on the forecast target day. The average ratings on the same day of the previous year indicate the average ratings on the same day of the previous year. The average ratings for the same time slot of the previous year indicate the average ratings for the same time slot of the previous year.

[0041] The most recent moving average ratings represent the moving average of the average ratings of the respective programs played in the most recent N times (N is an integer greater than or equal to 2) on the prediction target day. The N times of playing are the playing of programs in the main time period, and are extracted according to the playing interval of each program such as daily programs, weekly programs, annual programs, or irregular programs. The most recent N times on the prediction target day are sometimes also referred to as the previous N times. It is also possible to calculate the average ratings of the respective programs played in the previous N times, and use the average ratings of the programs played N times to calculate the most recent moving average ratings.

[0042] The moving average ratings on the same day of the week represent the moving average ratings of the previous N times corresponding to the same day of the week as the prediction target day. Missing flag 1 is a flag indicating whether there is a program broadcast on the same day of the week as the prediction target day. The moving average ratings on the same day of the week and the same time period represent the moving average ratings of the previous N times corresponding to the same day of the week as the prediction target day and the same time period. Missing flag 2 is a flag indicating whether there is a program broadcast on the same day of the week and the same time period as the prediction target day.

[0043] As the moving average, for example, a simple moving average or a weighted moving average can be used. As the weighted moving average, for example, a linear weighted moving average or an exponential weighted moving average can be used. The weight of the weighted moving average is set so that the later the broadcast day, the greater the weight. The moving average audience rating calculated by the weighted moving average is an example of a statistical value in which the contribution of the broadcast contact information of the second broadcast information broadcast after the first broadcast information is greater than the contribution of the broadcast contact information of the first broadcast information.

[0044] By using the weighted moving average, since the contribution of broadcasts on later broadcast days to the moving average ratings is greater than the contribution of broadcasts on earlier broadcast days to the moving average ratings, it is possible to obtain a moving average rating that largely reflects the recent viewing trends. As a result, the recent viewing trends are easily reflected in the predicted ratings, and the prediction accuracy is improved. It is also possible to use multiple moving average ratings with different N values, such as the moving average ratings of the previous 4 times, the moving average ratings of the previous 8 times, and the moving average ratings of the previous 10 times.

[0045] The previous N broadcasts that are the subject of the moving average are not necessarily the same programs as the program to be predicted, but may be various programs belonging to the same program category as the program to be predicted. In addition, the program description information of each program included in the previous N broadcasts may also include a predetermined keyword or a synonym thereof included in the program description information of the program to be predicted. The synonym of the predetermined keyword may be determined by a synonym dictionary.

[0046] <Setting process (setting procedure)>

[0047] A request period is set for at least one of the target programs that are the target of the audience rating prediction. The setting method of the request period has a degree of freedom, and various design changes can be made, but as an example, it is preferable to set the sub-program (mini-program) with a relatively short broadcasting number of minutes adjacent to the main program with a relatively long broadcasting number of minutes in the target program as the request period. This is because, as described above, when focusing on the accuracy comparison of the broadcasting number of minutes of each program, there is a situation where the prediction accuracy of the mini-program of about 6 minutes deteriorates depending on the broadcasting time. In particular, there is a tendency to predict the audience rating of the mini-program included in the A time such as 19:00 to 23:00 on weekdays and 18:00 to 23:00 on Saturdays and Sundays, which is considered to have the best audience ratings, higher than the actual. Alternatively, a 56-minute program adjacent to (affected by) the mini-program, which also has a tendency to deteriorate the prediction accuracy of the audience rating, can be set as the request period.

[0048] <Extraction Process (Extraction Step)>

[0049] Extracting the feature amount of the set requested time period, for example, the mini-program time period. Examples of the feature amount of the requested time period include the following (1) to (9).

[0050] (1) Number of minutes played during the requested time period

[0051] (2) The day of the week of the requested period (the day of the week when it starts and the day of the week when it ends)

[0052] (3) Time period of the requested period (start hour, end hour, start minute, end minute)

[0053] (4) The radio and television station of the requested time slot

[0054] (5) Saturday and Sunday flags of the requested period (at the beginning and end)

[0055] (6) Holiday flag of the requested period (start, end)

[0056] (7) Holiday flag of the working day of the requested period (at the beginning and end)

[0057] (8) The time span flag of the requested period (whether the start hour and the end hour are the same)

[0058] (9) The cross-day flag of the requested period (whether it crosses 29:00)

[0059] The purpose of adopting the above (1) as the characteristic quantity (variable) of the requested time period is that in the case of mini-time period programs with a short broadcast time, the actual ratings are likely to be lower than the basic 10-minute average predicted ratings, and this tendency is learned (reflected) in the ratings prediction.

[0060] As the purpose of using the above (2) to (4) as characteristic quantities (variables) of the requested time period, it can be cited that the actual ratings have different tendencies to increase, decrease, or remain unchanged compared to the basic 10-minute average predicted ratings depending on the specific broadcasting station, day of the week, and time period, and this difference is learned (reflected) in the ratings prediction.

[0061] The purpose of using (5) to (7) as characteristic quantities (variables) of the request time period is to assume that the audience's behavior is more likely to change on Saturdays and Sundays, holidays, and weekdays than on weekdays, and to learn (reflect) such a tendency in ratings prediction.

[0062] As the purpose of adopting the above (8) as the characteristic quantity (variable) of the request time period, it can be cited that there is a tendency for viewers to switch channels based on time (● hours), and there is a lot of difference between the basic 10-minute average predicted ratings and the actual ratings of the program that starts exactly at 0 minutes and the program that started a few minutes ago, and such a tendency is learned (reflected) in the ratings prediction.

[0063] As the purpose of adopting the above (9) as the characteristic quantity (variable) of the requested time period, it can be cited that the ratings trend across time is different in morning programs (especially across 4:00-5:00) from that of normal programs, and such differences are learned (reflected) in ratings prediction.

[0064] Based on the above, when extracting the characteristic amount of the set request time period, such as the mini program time period, it is preferred to extract the number of minutes of broadcasting in the request time period, the holiday mark of the weekday, the cross-time mark, and the cross-day mark. These characteristic parameters are extremely useful for eliminating the drawbacks caused by the model of learning and predicting the average ratings every 10 minutes in the past, and performing high-precision ratings prediction. Alternatively, when extracting the characteristic amount of the set request time period, such as the mini program time period, it is also possible to extract the number of minutes of broadcasting in the request time period, the day of the week, the time period, the radio and television station, the Saturday and Sunday mark, the holiday mark, the holiday mark of the weekday, the cross-time mark, and the cross-day mark. In this case, it is also possible to eliminate the drawbacks caused by the model of learning and predicting the average ratings every 10 minutes in the past, and perform high-precision ratings prediction.

[0065] <Acquisition Processing (Acquisition Steps)>

[0066] Based on the stored reference feature and the extracted feature of the request period, a predicted feature related to the prediction of the audience rating of the request period is obtained. In other words, the predicted feature related to the prediction of the audience rating of the request period is obtained by correcting the stored reference feature based on the extracted feature of the request period. Here, the predicted feature may be the predicted audience rating of the request period itself, or may be any parameter associated with the predicted audience rating of the request period.

[0067] Figure 5 FIG. 1 is a diagram showing an example of obtaining a predicted feature quantity of a request period based on a reference feature quantity and a feature quantity of a request period. Figure 5As shown, the reference feature of the ratings of the previous programs based on the 10-minute average predicted ratings and the feature of the request period set for the target program are input. As for the reference feature, since the information of the program period cannot be grasped at the time of prediction, the 10-minute average predicted ratings are calculated without using the information of the program period, and the reference feature is calculated based on this. In addition, the time series ratings of the same day of the week as the prediction reference day are used, or the 24-week moving average difference (recent trend) is added to calculate the prediction value taking the trend into consideration. As described above, the feature of the request period may include at least one of the broadcast minutes, day of the week, time period, broadcasting station, Saturday and Sunday flag, holiday flag, holiday flag of weekdays, cross-time flag, and cross-day flag of the request period. Then, based on the input reference feature and the feature of the request period, the prediction feature related to the prediction of the rating of the request period is obtained (the prediction feature is obtained by correcting the reference feature based on the feature of the request period). In this way, a model (prediction value conversion model of the request period) for correcting the prediction feature based on the feature of the request period (broadcast time, etc.) is added. For example, by reading feature quantities such as time period and span of program information, the program of short minutes such as mini program period can be corrected to achieve high accuracy of ratings prediction. More specifically, considering that the predicted ratings of mini program period are easily overestimated, the predicted ratings of mini program period can be corrected to be low and close to the actual measured value.

[0068] Figure 6 1 is a diagram showing an example of a schematic structure of a system including a prediction device (audience rating prediction device). System 1 includes a cost calculation device 10, a broadcast distribution device 20, and a prediction device (audience rating prediction device) 30. Cost calculation device 10, broadcast distribution device 20, and prediction device 30 are connected to each other via a network 40 so as to be communicable.

[0069] The cost calculation device 10 calculates the cost of a television commercial, for example. The cost calculation device 10 calculates the cost of a television commercial based on a calculation formula including the audience rating (predicted audience rating) of a target program in which the television commercial is inserted.

[0070] The cost calculation device 10, for example, obtains the cost of the delivery pattern of commercial advertisements in each time period within a specified period (e.g., a week) of a certain broadcasting and television station (also read as a channel). Here, the cost in the present disclosure may be a rating point cost (Cost Per Rating) representing the cost of 1GRP (1%) per rating. The delivery pattern may include existing (arbitrary) delivery patterns (e.g., inverted L-shaped, all-day type, ヨ-shaped, U-shaped, straight-shaped, late-night type, all-day daytime type, etc.). In addition, the cost calculation device 10 may also obtain the cost of each delivery pattern, the time period constituting the delivery pattern, etc. based on input from the outside (e.g., the settings of the broadcasting and television station).

[0071] The cost calculation device 10 may also obtain (e.g., receive) information used for cost calculation. For example, the cost calculation device 10 may obtain (receive) information on the ratings related to each advertising position from the prediction device 30 (e.g., the ratings obtained by adding the household ratings, personal ratings, personal overall ratings, real-time ratings, and time-shifted ratings up to the 7th day after the broadcast (may also be referred to as "All&P+C7 (Program+Commercial7)", etc.). For example, the cost calculation device 10 may also derive the cost of one position at time A based on the setting of one unit price at time A of the broadcasting station and the ratings information obtained from the prediction device 30.

[0072] The cost calculation device 10 may be a portable terminal (mobile communication terminal) such as a mobile phone, a smart phone, or a tablet terminal, or may be a fixed communication terminal such as a personal computer (PC) or a server. That is, the cost calculation device 10 in the present disclosure may be read as a communication device. In addition, the cost calculation device 10 may communicate with a network (such as the Internet) via wired and / or wireless (for example, LTE (Long Term Evolution), NR (New Radio), Wi-Fi (registered trademark), etc.).

[0073] The broadcast distribution device 20 is a device for distributing broadcasts to televisions (TV). Here, a television (television receiver) is a device having at least one function of receiving terrestrial broadcasts, broadcasts from broadcast satellites (BS: Broadcasting Satellite) / communication satellites (CS: Communications Satellite), Internet broadcasts (Internet TV), etc. For example, the television may be a multi-function TV, a smart TV, an IP (Internet Protocol) TV, a set-top box, etc. In addition, broadcast refers to the transmission of telecommunications for the purpose of direct reception by the public, and may also include wireless broadcasts, Internet broadcasts, etc.

[0074] The prediction device 30 is a device that aggregates the actual ratings of the programs broadcast on the television. In addition, the prediction device 30 is a device that predicts the ratings of the programs (target programs) that will be broadcast on the television later. The ratings can be stored in predetermined time units (e.g., seconds, minutes, hours). The prediction device 30 can send information related to the ratings (e.g., the ratings of the target programs inserted in the television commercials (predicted ratings)) to other devices (e.g., the cost calculation device 10) via the network.

[0075] The prediction device 30 can be, for example, a device managed by a television ratings investigator. The prediction device 30 can also obtain and manage viewing history data from a measuring device connected to a television of a specified household (for example, a survey cooperative household). In addition, the ratings can also be calculated or predicted based on the data. In the prediction device 30, the ratings of each specified target (users who meet the specified attributes) (the target's personal ratings) can also be stored. In addition, in the present disclosure, it is assumed that the target is a gender and age classification used in the marketing field (children (C: Child) group, teenagers (T: Teen) group, male (M: Male) 1-M3 group, female (F: Female) 1-F3 group), but it is not limited to this classification.

[0076] The prediction device 30 is configured to be able to predict the "audience's advertising ratings" of television with high accuracy using a new analysis method such as processing big data. The prediction device 30 may include a data input unit for inputting data source information, an explanation variable conversion unit for performing specific processing on the data source information input from the data input unit to convert the data source information into explanation variables, a program installation / execution unit for installing a program of a predetermined algorithm, and a data output unit for outputting the result of the program installation / execution unit. The algorithm of the program installed in the program installation / execution unit is executed using the data source information input from the data input unit and the explanation variables obtained by converting the explanation variables by the explanation variable conversion unit, and the predicted ratings data obtained by executing the algorithm is output from the output unit. Based on the output predicted ratings data, the advertising position can be optimized. Here, as the data source information for calculating the ratings (predicted ratings) of the target program in which the television commercial is inserted, the ratings of related programs related to the target program (for example, the broadcast before the last episode of the series, the broadcast before the last episode of the special program, etc.) can be used.

[0077] In addition, the system structure is an example and is not limited to this. For example, each device Figure 6 The system 1 may include one each, but the number of each device is not limited thereto and may include multiple devices. The system 1 may include a structure that does not include some devices, or may include a structure that implements the function of one device through multiple devices. Conversely, a structure in which one device implements the functions of multiple devices may be adopted. For example, at least two of the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30 may be installed on one server.

[0078] Figure 7 This is a functional block diagram showing an example of the internal structure of the prediction device 30. The prediction device 30 includes a storage unit 31, a setting unit 32, an extraction unit 33, and an acquisition unit 34. The prediction method and prediction program disclosed herein are implemented by causing a computer included in the prediction device 30 (storage unit 31, setting unit 32, extraction unit 33, acquisition unit 34) to execute a predetermined processing step.

[0079] The storage unit 31 stores the reference feature of the ratings of previous programs. The storage unit 31 stores the feature of the ratings during a specified period of multiple previous programs, and stores the time series ratings on the same day of the week as the prediction base date as the reference feature. The storage unit 31 stores the feature of the ratings during a specified period of multiple previous programs, and stores the difference between the ratings of the week including the prediction base date and the exponential moving average ratings during the specified period before the prediction base date as the reference feature.

[0080] The setting unit 32 sets a request period for at least one of the target programs. The setting unit 32 sets a sub-program (mini-program) of relatively short broadcast minutes adjacent to a main program of relatively long broadcast minutes in the target program in the request period.

[0081] The extraction unit 33 extracts the feature quantity of the request period. As the feature quantity of the request period, the extraction unit 33 extracts the number of minutes played during the request period, the holiday flag of the weekday, the time span flag, and the day span flag. As the feature quantity of the request period, the extraction unit 33 extracts the number of minutes played during the request period, the day of the week, the time period, the broadcasting station, the Saturday and Sunday flag, the holiday flag, the holiday flag of the weekday, the day span flag, and the day span flag.

[0082] The acquisition unit 34 acquires the predicted feature value related to the prediction of the audience rating of the requested period based on the reference feature value and the feature value of the requested period. The acquisition unit 34 acquires the predicted feature value related to the prediction of the audience rating of the requested period by correcting the reference feature value based on the feature value of the requested period.

[0083] In addition, the present disclosure also covers the content of replacing "ratings" with specific indicators (or the degree of improvement, deviation, etc. of specific indicators). The specific indicators can be indicators for judging advertising effectiveness, intermediate indicators for improving advertisers' key performance indicators (KPIs), etc., such as recognition rate, coverage, ROAS (Return On Advertising Spend), ROI (Return on Investment), etc. At least a part of the record of "ratings" can also be replaced by any of the above-mentioned specific indicators.

[0084] Figure 8 1 is a diagram showing an example of the functional configuration of the cost calculation device 10, the broadcast distribution device 20, and the prediction device (audience rating prediction device) 30. Figure 8As shown, the cost calculation device 10, the broadcast distribution device 20 and the prediction device 30 include a control unit 110, a storage unit 120, a communication unit 130, an input unit 140 and an output unit 150. Figure 8 In the example of the functional blocks of the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30, it is also possible to have Figure 8 Other functional blocks not depicted in the figure. In addition, the structure may be configured so as not to include some functional blocks.

[0085] The control unit 110 controls the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30. The control unit 110 may be composed of a controller, a control circuit, or a control device described based on the common knowledge in the technical field of the present invention.

[0086] The storage unit 120 stores information used in the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30. The storage unit 120 may be composed of, for example, a memory, a storage device, or the like described based on the common knowledge in the technical field of the present disclosure.

[0087] The communication unit 130 performs communication (e.g., mutual communication) between the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30. The communication unit 130 may be composed of a transmitter / receiver, a transmission / reception circuit, or a transmission / reception device described based on the common understanding of the technical field involved in the present disclosure. In addition, the communication unit 130 may also be composed of a transmission unit and a reception unit.

[0088] The input unit 140 receives input from the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30. In addition, the input unit 140 may be connected to a predetermined device, a storage medium, etc., and receive input of data. The input unit 140 may be composed of input devices such as a keyboard, a mouse, and a button, input / output terminals, and input / output circuits, etc., which are described based on the common understanding in the technical field of the present disclosure. In addition, the input unit 140 may also be a structure integrated with the display unit (for example, a touch panel).

[0089] The output unit 150 performs output of the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30. For example, the output unit 150 may also include a display unit that displays images, a sound output unit that outputs sounds, and the like. The display unit may be composed of, for example, a display device such as a display or a monitor that is described based on the common understanding in the technical field of the present disclosure. In addition, the sound output unit may be composed of an output device such as a speaker that is described based on the common understanding in the technical field of the present disclosure.

[0090] In addition, the block diagram used in the description of the above-mentioned embodiment represents a block diagram of a functional unit. These functional modules (structural parts) are implemented by any combination of hardware and / or software. In addition, the implementation method of each functional module is not particularly limited. That is, each functional module can be implemented by a physically combined device, or it can be implemented by two or more physically separated devices connected by wire or wirelessly, and implemented by these multiple devices.

[0091] For example, the apparatus according to one embodiment of the present disclosure (the cost calculation apparatus 10 , the broadcast distribution apparatus 20 , and the prediction apparatus 30 ) can function as a computer that executes the prediction process (audience rating prediction process) of the present disclosure. Fig. 9 1 is a diagram showing an example of the hardware configuration of the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30. The cost calculation device 10, the broadcast distribution device 20, and the prediction device 30 may be physically configured as a computer device including a processor 1001, a memory 1002, a storage 1003, a communication device 1004, an input device 1005, an output device 1006, a bus 1007, and the like.

[0092] In addition, in the present disclosure, devices, circuits, equipment, units, servers, etc. can be interchangeable. The hardware structure of the cost calculation device 10, broadcast distribution device 20, and prediction device 30 can be configured to include one or more of the devices shown in the figure, or can be configured to exclude some of the devices.

[0093] For example, although only one processor 1001 is shown, there may be multiple processors. In addition, the processing may be performed by one processor, or the processing may be performed by two or more processors simultaneously, sequentially, or using other methods. In addition, the processor 1001 may also be implemented by one or more chips.

[0094] The functions of the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30 are realized by reading the specified software (program) into the hardware such as the processor 1001 and the memory 1002, and the processor 1001 performs calculations to control the communication of the communication device 1004, the reading and / or writing of data in the memory 1002 and the storage 1003, etc.

[0095] The processor 1001 controls the entire computer by, for example, running an operating system. The processor 1001 may be composed of a central processing unit (CPU) including an interface with peripheral devices, a control device, a computing device, registers, etc. In addition, the above-mentioned control unit 110 and other units may also be implemented by the processor 1001.

[0096] In addition, the processor 1001 reads the program (program code), software module, data, etc. from at least one of the storage 1003 and the communication device 1004 to the memory 1002, and performs various processes based on them. As the program, a program that causes the computer to execute at least a part of the actions described in the above embodiment is used. For example, the control unit 110 can be implemented by a control program stored in the memory 1002 and running in the processor 1001, and other functional modules can also be implemented in the same way.

[0097] The memory 1002 is a computer-readable storage medium and may be composed of at least one of, for example, a ROM (Read Only Memory), an EPROM (Erasable Programmable ROM), an EEPROM (Electrically EPROM), a RAM (Random Access Memory), and other suitable storage media. The memory 1002 may be referred to as a register, a cache, a main memory (main storage device), etc. The memory 1002 may store a program (program code), a software module, etc., which may be executed to implement a method of an embodiment.

[0098] The memory 1003 may be a computer-readable storage medium, and may be composed of at least one of, for example, a floppy disk, a floppy disk (registered trademark), a magneto-optical disk (e.g., a compact disc (CD-ROM) (Compact Disc ROM), a digital versatile disc, a Blu-ray (registered trademark) disc), a removable disk, a hard disk drive, a smart card, a flash memory device (e.g., a card, a stick, a key drive), a magnetic stripe, a database, a server, and other suitable storage media. The memory 1003 may be referred to as an auxiliary storage device. In addition, the storage unit 120 described above may also be implemented by the memory 1002 and / or the memory 1003.

[0099] The communication device 1004 is hardware (transceiver) for communicating between computers via at least one of a wired network and a wireless network, and is also called a network device, a network controller, a network card, a communication module, etc. The communication device 1004 may include a SIM card. In addition, the communication unit 130 may also be implemented by the communication device 1004.

[0100] The input device 1005 is an input device (e.g., a keyboard, a mouse, etc.) for receiving input from the outside. The output device 1006 is an output device (e.g., a display, a speaker, etc.) for realizing output to the outside. In addition, the input device 1005 and the output device 1006 may also be an integrated structure (e.g., a touch panel). In addition, the above-mentioned input unit 140 and the output unit 150 may also be realized by the input device 1005 and the output device 1006, respectively.

[0101] In addition, the processor 1001, the memory 1002 and other devices are connected via a bus 1007 for communicating information. The bus 1007 may be composed of a single bus or different buses between devices.

[0102] In addition, the cost calculation device 10, the broadcast distribution device 20, and the prediction device 30 may include hardware such as a microprocessor, a digital signal processor (DSP), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), and an FPGA (Field Programmable Gate Array), and may also implement a part or all of each functional module through the hardware. For example, the processor 1001 may be implemented by at least one of these hardware.

[0103] (Variation Example)

[0104] Furthermore, terms explained in the present disclosure and / or terms required for understanding the present disclosure may be replaced with terms having the same or similar meanings.

[0105] The information, parameters, etc. described in this disclosure may be expressed using absolute values, relative values ​​to specified values, or other corresponding information. In addition, the names used for parameters, etc. in this disclosure are not limited in any way.

[0106] Any of a variety of different technologies may be used to represent the information, signals, etc. described in the present disclosure. For example, data, instructions, commands, information, signals, bits, symbols, chips, etc. mentioned in the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, light fields or photons, or any combination thereof.

[0107] Information, signals, etc. can also be input and output via multiple network nodes. The input and output information, signals, etc. can be stored in a specific place (such as memory), or can be managed using a table. The input and output information, signals, etc. can be overwritten, updated, or appended. The output information, signals, etc. can also be deleted. The input information, signals, etc. can also be sent to other devices.

[0108] Furthermore, notification of the prescribed information (for example, notification of "it is X") is not limited to explicit notification, but may be performed implicitly (for example, by not notifying the prescribed information or by notifying other information).

[0109] Software shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subroutines, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

[0110] In addition, software, instructions, information, etc. may also be sent and received via at least one of a transmission medium and a signal waveform. For example, when the software is sent from a website, server or other remote source using at least one of wired technology (coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL: Digital Subscriber Line), etc.) and wireless technology (infrared, microwave, etc.), at least one of these wired technology and wireless technology is included in the definition of transmission medium.

[0111] The terms "system" and "network" used in the present invention may be used interchangeably.

[0112] The various methods and implementations described in this disclosure may be used alone, in combination, or switched as they are executed. In addition, the processing order, sequence, flow chart, etc. of the various methods and implementations described in this disclosure may be changed in order as long as there is no contradiction. For example, for the methods described in this disclosure, the elements of the various steps are presented in an exemplary order and are not limited to the specific order presented.

[0113] The description of “based on” used in the present disclosure does not mean “only based on” unless otherwise specified. In other words, “based on” means both “only based on” and “at least based on”.

[0114] Any reference to an element using the terms "first", "second", etc. used in this disclosure does not limit the quantity or order of these elements as a whole. These terms can be used as a convenient method to distinguish between two or more elements in this disclosure. Therefore, reference to a first and a second element does not mean that only two elements can be used, or that the first element must precede the second element in some form.

[0115] In this disclosure, if “include,” “including,” and variations thereof are used, these terms are meant to be as comprehensive as the term “comprising.” In addition, the term “or” used in the present invention means not an exclusive or.

[0116] In the present disclosure, for example, like a, an, and the in English, when articles are added by translation, the present disclosure may include that nouns following these articles are plural.

[0117] The invention involved in the present disclosure is described in detail above, but for those skilled in the art, the invention involved in the present disclosure is not limited to the embodiments described in the present disclosure. The invention disclosed in the present disclosure can be implemented as a modification and variation without departing from the gist and scope of the invention determined according to the records of the claims. Therefore, the description of the present disclosure is intended to illustrate and does not provide any restrictive meaning for the invention according to the present disclosure.

[0118] This application is based on Japanese Patent Application No. 2022-112232 filed on July 13, 2022, the entire contents of which are expressly incorporated herein by reference.

Claims

1. A prediction device, characterized in that: have: a storage unit storing a reference feature value of audience ratings of past programs; a setting unit, which sets a request time period for at least one target program; an extraction unit, which extracts a feature quantity of the request period; as well as An acquisition unit acquires a predicted feature amount related to prediction of the audience rating of the requested time period based on the reference feature amount and the feature amount of the requested time period.

2. The prediction device according to claim 1, characterized in that The extraction unit extracts the number of minutes played during the requested time period, a holiday flag of a weekday, a cross-time flag, and a cross-day flag as the feature amount of the requested time period.

3. The prediction device according to claim 1 or 2, characterized in that: The extraction unit extracts the number of minutes of broadcasting, day of the week, time period, broadcasting station, Saturday and Sunday flags, holiday flags, holiday flags on weekdays, cross-time flags, and cross-day flags of the requested time period as the feature quantities of the requested time period.

4. The prediction device according to claim 1 or 2, characterized in that: The setting section sets a sub-program of a relatively short broadcasting number of minutes in the request period, the sub-program being adjacent to a main program of a relatively long broadcasting number of minutes in the target program.

5. The prediction device according to claim 1 or 2, characterized in that: The storage unit stores feature quantities of audience ratings during a predetermined period of a plurality of past programs, and stores time-series audience ratings on the same day of the week as a prediction reference date as the reference feature quantity.

6. The prediction device according to claim 1 or 2, characterized in that: The storage unit stores feature values ​​of audience ratings in a predetermined period of a plurality of past programs, and stores a difference between the audience ratings of the week including a prediction base date and the exponential moving average audience ratings in a predetermined period before the prediction base date as the reference feature value.

7. A prediction method, characterized in that: have: A step of storing a reference feature value of the audience rating of a past program; a step of setting a request time period for at least one of the target programs; A step of extracting a characteristic quantity of the request period; as well as A step of acquiring a predicted feature amount related to prediction of the audience rating of the request time period based on the reference feature amount and the feature amount of the request time period.

8. A prediction program, characterized in that Causes the computer to execute: A step of storing a reference feature value of the audience rating of a past program; a step of setting a request time period for at least one of the target programs; A step of extracting a characteristic quantity of the request period; as well as A step of acquiring a predicted feature amount related to prediction of the audience rating of the request time period based on the reference feature amount and the feature amount of the request time period.

Images