Viewing information processing device and viewing information processing method

Abstract

Based on viewing information obtained by people meters or the like in specific areas other than the survey area, the viewing rate for the survey area is predicted from viewing log data for the survey area. [Solution] This invention generates predicted values ​​for individual audience ratings using a model obtained by learning based on viewing situation information obtained from dedicated equipment (people meters, etc.) installed in sample households in areas other than the survey area, using viewing log information obtained from television devices installed in each household in the survey area as a prediction model. The components other than the prediction model perform processing similar to the processing of predicting individual audience ratings by applying a prediction model learned from viewing situation information in the survey area to viewing log information.

Description

[Technical Field]

[0001] The present invention relates to an audiovisual information processing device and an audiovisual information processing method. [Background technology]

[0002] A program's audience rating is generally calculated by installing a dedicated device (such as a people meter system or online meter system, hereinafter referred to as "people meter") to measure audience status in a sampling of multiple households, and collecting and analyzing audience status information for programs viewed on television sets, etc.

[0003] Furthermore, as shown in Patent Document 1, a technology has been proposed that uses a model (mathematical model) that has been subjected to machine learning processing in advance to estimate the attribute composition of viewers of a television device (more specifically, one or more viewers who can watch television broadcasts received via the television device) and attribute-specific viewing information indicating the viewing status of television broadcasts for each attribute of the viewers, from viewing log information collected from each television device by each television manufacturer. Here, the attribute composition of viewers of a television device indicates which of multiple types of attributes, classified according to gender, age, etc., the viewers are made up of. Furthermore, the attribute-specific viewing information of viewers of a television device indicates, for each viewer attribute, whether or not viewers of each attribute watched a television broadcast broadcast on a certain channel (or broadcast station affiliate) at a certain date and time.

[0004] In this regard, because viewing log information differs from the actual viewing conditions in the surveyed area, there is a possibility that the attribute-specific viewing information obtained by estimation solely from viewing log information, as in Patent Document 1, is not accurate. Therefore, Patent Document 2 proposes estimating attribute-specific viewing information by applying the viewing log information to be analyzed, obtained from each television device in the surveyed area, to a prediction model obtained by learning from sample viewing log data in the surveyed area and viewing information obtained from people meters or the like installed in sample households in the surveyed area. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2019-220825 [Patent Document 2] Japanese Patent Application Publication No. 2023-153658 Summary of the Invention [Problem to be solved by the invention]

[0006] However, although people meters and other devices are currently installed in each household in multiple regions across the country (for example, 32 regions), considering the profitability of future audience rating surveys, it is not necessarily possible to install people meters and other devices in the survey target areas.

[0007] Therefore, after careful consideration, the inventors constructed a prediction model using viewing situation information (information obtained by people meters, etc.) from an area different from the area where viewing log information was obtained (survey area), and found that the estimated viewing rate obtained by applying the viewing log information to the prediction model did not deviate significantly from the actual viewing rate, and therefore reliability was guaranteed.

[0008] In view of this situation, the present invention proposes a technology for predicting the audience rating of a survey area from the audience log data of the survey area, based on audience status information obtained using people meters or the like in specific areas other than the survey area. [Means for solving the problem]

[0009] In order to solve the above problems, the present invention provides, as an example, A viewing information processing device that applies viewing log information obtained from television devices installed in each household in a survey area to a prediction model to predict individual audience ratings in the survey area, a storage device that stores at least a program for predicting the individual audience rating in the survey area and the prediction model; a control device that reads the program, the viewing log information, and the prediction model from the storage device and executes the program to predict the individual audience rating in the survey target area; The prediction model is a model obtained by learning based on viewing situation information obtained from dedicated devices installed in sample households in areas other than the survey target area, and includes a first model for calculating the presence probability by attribute of each household and estimating attributes indicating the composition of each household, and a second model for separating the viewing time of each attribute of each household, The control device A process of applying the viewing log information of each of the households to the first model to estimate the attributes; applying the viewing log information of each of the households to the second model to estimate the viewing time of each television broadcast for each of the attributes of each of the households during a predetermined period; A process of identifying a frequency distribution of the number of viewers with respect to the viewing time for each of the attributes of each of the households in the survey area; a process of comparing the frequency distribution of the number of viewers for each attribute with a standard frequency distribution for each attribute that is predetermined based on the viewing situation information of areas other than the survey target area, and selecting viewers to be surveyed in the survey target area for each attribute so that the frequency distribution of the number of viewers to be surveyed coincides with the standard frequency distribution; a process of calculating the audience rating of the television broadcasting subject to the survey based on an estimation result of whether or not the selected viewers to be surveyed have watched the television broadcasting subject to the survey; The present invention proposes a viewing information processing device that executes the above.

[0010] Further features related to the present invention will become apparent from the description and accompanying drawings of this specification, and the aspects of the present invention may be realized and realized by the elements and combinations of various elements and aspects set forth in the following detailed description and the appended claims. The descriptions in this specification are exemplary and illustrative only and are not intended to limit the scope or application of the present invention in any way. [Effects of the Invention]

[0011] According to the technology of the present invention, it is possible to predict the audience rating of a survey area from the audience log data of the survey area based on audience status information obtained using a people meter or the like in a specific area other than the survey area. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram for explaining an overview of viewing information processing according to the present embodiment. [Figure 2] 1 is a diagram showing an example of a schematic configuration of an audience information processing system according to an embodiment of the present invention; [Figure 3] FIG. 10 is a diagram illustrating an example of the internal configuration of an attribute configuration estimation unit 113. [Figure 4] 10 is a table showing an example of affiliated key stations assigned to local stations in each region. [Figure 5] 10 is a diagram illustrating an example of the internal configuration of an individual separation processing unit 114. FIG. [Figure 6] 10 is a diagram illustrating an example of the internal configuration of a viewing minute distribution adjustment processing unit 115. FIG. [Figure 7] 11 is a diagram for explaining the processing performed by the frequency distribution specifying unit 1151 and the survey target viewer selecting unit 1152. FIG. [Figure 8] 8 is a substitution pattern table 800 showing examples of substitution patterns based on viewing situation information such as people meters in other areas. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In the accompanying drawings, functionally identical elements may be designated by the same numerals. Note that the accompanying drawings show specific embodiments and implementation examples in accordance with the principles of the present invention, but these are for understanding the present invention and are by no means to be used to interpret the present invention in a limiting manner.

[0014] Although the present embodiment has been described in sufficient detail to enable those skilled in the art to practice the present invention, it should be understood that other implementations and forms are possible, and that changes in configuration and structure and substitutions of various elements are possible without departing from the scope and spirit of the technical concept of the present invention. Therefore, the following description should not be interpreted as being limited thereto.

[0015] <Outline of viewing information processing> 1 is a diagram for explaining an overview of viewing information processing according to this embodiment. The viewing information processing includes a viewing log information acquisition process 101, an attribute composition estimation process 102, an individual separation process 103, a viewing minute distribution adjustment process 104, and an individual viewing rate estimation process 105.

[0016] The viewing log information acquisition process 101 is a process of acquiring (purchasing) from each television manufacturer viewing log information (log information for each television device in each household) collected by each television manufacturer via a network from households that own their television devices.

[0017] The attribute composition estimation process 102 is a process for estimating the attribute composition indicating which members of which attributes are included in the device-specific household (in other words, the attribute composition of viewers of the television devices in the device-specific household) from device-specific viewing log information for a predetermined period (e.g., one month, several months, one year, etc.) output from the television devices of each device-specific household.

[0018] The individual separation process 103 is a process for estimating the television broadcast viewing time (e.g., viewing time per day or viewing time per month) of viewers with each attribute in a predetermined unit time period for each attribute included in the attribute configuration estimated by the attribute configuration estimation process for each device-specific household.

[0019] The viewing minute distribution adjustment process 104 includes a process (frequency distribution determination process) of identifying the frequency distribution of viewing time (viewing time per day) in the survey area (more specifically, the frequency distribution of the number of viewers relative to viewing time) for each of the n types of attributes At(i) (i=1, 2, ..., n), and a process (survey target viewer selection process) of selecting, for each of the n types of attributes At(i) (i=1, 2, ..., n), survey target viewers who are the viewers to be surveyed for investigating viewing conditions such as television broadcast ratings.

[0020] The individual audience rating estimation process 105 is a process of estimating viewing situation data (data equivalent to audience ratings) for each of n types of attributes At(i) regarding the television broadcast being surveyed in the survey area (hereinafter referred to as the survey target broadcast).

[0021] <Configuration example of viewing information processing system> Fig. 2 is a diagram showing a schematic configuration example of a viewing information processing system according to this embodiment. The viewing information processing system includes a viewing information processing device 100, television manufacturer 1 server devices 300_1 to 300_N connected to the viewing information processing device 100 via a network 1_400, and television devices 1 200_1 to M 200_M of each household connected to the television manufacturer 1 server devices 300_1 to 300_N via a network 2_500. Note that although the network 1_400 and the network 2_500 are separate networks in Fig. 2, they may be the same network.

[0022] One device-specific household is associated with each of television devices 1 200_1 to M 200_M. Therefore, even if a dwelling unit is equipped with multiple television devices, one device-specific household is associated with each of the multiple television devices. In this case, each device-specific household corresponding to each of the multiple television devices may include the same members as viewers. In other words, one or more members of one device-specific household may be members of another device-specific household.

[0023] Television device 1 200_1 to television device M 200_M of each device-specific household may be composed of a television, or may be composed of a television and a recording device connected to it. Each television device is capable of detecting the channel of the television broadcast viewed through it, as well as the date and time when the television broadcast of that channel was viewed (more specifically, the date and time when viewing of the television broadcast of that channel began and ended). From this detection information, viewing log information of the device-specific household (hereinafter referred to as device-specific viewing log information) is generated. In other words, device-specific viewing log information is the viewing log information of all members included in the device-specific household.

[0024] Each of the television manufacturer 1 server device 300_1 to television manufacturer M server device 300_M collects viewing log information from the television device manufactured by the television manufacturer corresponding to the television device among television devices 1 200_1 to M 200_M installed in each household.

[0025] The audience rating information processing device 100 includes a control device 110 consisting of a processor (CPU, MPU, GPU, MCU, etc.) that executes various processes, a storage device 120 consisting of a ROM, RAM, etc., an input device 130 consisting of a mouse, keyboard, microphone, etc., an output device 140 consisting of a display device, printer, speaker, etc., and a communication device 150.

[0026] The storage device 120 stores and holds various processing programs and parameters, viewing log information for the survey target area acquired from each television manufacturer, viewing situation information acquired from people meters and the like in areas other than the survey target area, and a prediction model that predicts individual audience ratings using the viewing log information. The prediction model is a model obtained by training a learning model using viewing situation information collected from people meters and the like installed in areas other than the survey target area.

[0027] The control device 110 reads various programs from the storage device 120 and expands them into internal memory (not shown), thereby constructing a survey area viewing log information acquisition unit 111, a reference area viewing situation information acquisition unit 112, an attribute composition estimation unit 113, an individual separation processing unit 114, a viewing minute distribution adjustment processing unit 115, and an individual viewing rate estimation unit 116.

[0028] The survey target area viewing log information acquisition unit 111 controls the communication device 150 to acquire survey target area viewing log information from a desired television manufacturer's server device (at least one of the television manufacturer 1 server device 300_1 to the television manufacturer N server device 300_N) via the network 1_400, and stores it in the storage device 120. Note that the survey target area viewing log information may be input directly from the input device 130.

[0029] The reference area viewing situation information acquisition unit 112 controls the communication device 150 to collect viewing situation information acquired by a people meter or the like in a reference area (an area other than the survey target area: for example, the Kanto region when the survey target area is the Hiroshima area) to be used when learning the prediction models of the attribute composition estimation process 102 and the individual separation process 103, and stores the information in the storage device 120. Note that the viewing situation information acquired by a people meter or the like in the reference area may be input directly from the input device 130.

[0030] The attribute composition estimation unit 113 applies the viewing log information (target viewing log information) of each television device of each household to a prediction model (first model), determines the attribute composition, and performs processing to estimate the family composition linked to the corresponding television device. Details of the attribute composition estimation unit 113 will be described later.

[0031] The individual separation processing unit 114 applies the target viewing log information to a prediction model (second model), determines the time each family member watches the television device, and performs processing to estimate the viewing time by attribute. The individual separation processing unit 114 will be described in detail later.

[0032] Viewing minute distribution adjustment processing unit 115 extracts individuals with similar viewing tendencies based on viewing situation information obtained by a people meter, etc., and adjusts sampling. Details of viewing minute distribution adjustment processing unit 115 will be described later. The individual audience rating estimation unit 116 estimates viewing situation data (viewing rating) for each individual. Details of the individual audience rating estimation unit 116 will be described later.

[0033] <Details of the attribute configuration estimation unit 113> The attribute composition estimation unit 113 estimates an attribute composition indicating which members of which attributes are included in the device-specific household (in other words, the attribute composition of viewers of the device-specific household's television device k 200_k) from device-specific viewing log information for a predetermined period (e.g., one month, several months, one year, etc.) output from the television device k 200_k (k=1, 2, ..., M) of each device-specific household.

[0034] Here, the attributes of the members of each device-specific household (viewers of television device k 200_k) are classified into multiple types of attributes based on, for example, the age and gender of the members. For example, the attributes are classified into multiple types such as children under y1, men aged y1 or older but younger than y2, women aged y1 or older but younger than y2, men aged y2 or older, and women aged y2 or older. Hereinafter, the number of types of attributes will be designated as N, and each of the N types of attributes will be appropriately represented as At(i) (i=1, 2, ..., n). Note that the attributes of the members of each device-specific household 20 can be classified not only by age and gender, but also by various parameters such as occupation and educational background.

[0035] FIG. 3 is a diagram showing an example of the internal configuration of the attribute composition estimation unit 113. As shown in FIG. 3, the attribute composition estimation unit 113 estimates the attribute composition of members of each device-specific household using a first model that has been previously subjected to machine learning processing. Here, the first model is a model that has been previously subjected to machine learning processing so that the attribute composition of members of a device-specific household can be estimated from device-specific viewing log information for a predetermined period of any device-specific household 20 that belongs to the survey target area. In the machine learning processing for the first model, viewing situation information for each sample household obtained from people meters or the like installed in each of multiple sample households (sample households in areas other than the survey target area) whose attribute composition is known, and the attribute composition for each sample household are used as learning data. Furthermore, a publicly known algorithm can be used as the algorithm for the machine learning processing.

[0036] In this case, in this embodiment, the first model is configured to be able to determine, for each of n types of attributes At(i) (i = 1, 2, ..., n), an attribute-specific presence probability, which is the probability that a member of each attribute At(i) is present in the device-specific household 20, from the device-specific viewing log information for a predetermined period of time for each device-specific household 20. As such a first model, for example, a model similar to the mathematical model for determining household composition in Patent Document 1 may be adopted. However, the first model may be of another form as long as it is able to determine the attribute-specific presence probability (or an index value similar thereto) from the device-specific viewing log information for a predetermined period of time for the device-specific household 20.

[0037] Then, the attribute composition estimation unit 113 estimates an attribute whose attribute-specific existence probability specified by the first model is higher than a predetermined threshold (for example, 0.5) (or an attribute whose attribute-specific existence probability is higher than or matches the predetermined threshold) as an attribute of the constituent members of the device-specific household 20. As an example, Fig. 3 illustrates the processing of the attribute composition estimation unit 113 for one device-specific household (in Fig. 3, a device-specific household whose identification information ID is x1).

[0038] 3, the attribute-specific presence probabilities of members of a device-specific household are identified by the first model from device-specific viewing log information for a predetermined period of time for a device-specific household with identification information ID x1, as shown in the figure. In this example, of n types of attributes At(i) (i=1, 2, ..., n), the attribute-specific presence probabilities of each of the attributes At(1), At(3), and At(n) are higher than a predetermined threshold (here, for example, 0.5), and the attribute-specific presence probabilities of each of the other attributes At(2), At(4) to At(n-1) are lower than the threshold.

[0039] In this case, the attribute composition estimation unit 113 estimates that the attributes At(1), At(3), and At(n) whose attribute-specific existence probabilities are higher than the threshold are attributes of the members (viewers of the television device 23) of the device-specific household 20, and estimates that the attributes At(2), At(4) to At(n-1) whose attribute-specific existence probabilities are lower than the threshold are not attributes of the members of the device-specific household. In this way, the attribute composition of the members of the device-specific household is estimated.

[0040] In this way, the attribute composition estimation unit 113 estimates the attribute composition of the members of each device-specific household by estimating whether or not members (viewers) of each attribute are present in the device-specific household based on whether the attribute-specific presence probability identified by the first model is high or low relative to a predetermined threshold.

[0041] <Example of allocation of viewing status information from people meters, etc. to survey target areas> Figure 4 is a table showing examples of affiliated key stations assigned to local stations in each region. Figure 4 shows the affiliated key stations to be used in place of each local station in each survey target region (XX region, XX region, △△ region, ●● region, ▲▲ region) when predicting the audience ratings for each local station in that region. In other words, the table shows which affiliated key station's viewing situation information, such as people meters, is used to build a prediction model to be applied when predicting the audience ratings for each local station.

[0042] For example, when predicting the viewership rating of A1 TV in a certain area (survey area), the viewing log information of A1 TV in that area (log information obtained from the television devices of each household) is applied to a prediction model trained on viewing status information (information collected from people meters, etc.) of G TV (A1 TV's affiliated key station) in an area different from the certain area (for example, the Kanto region), and the viewership rating of A1 TV in that certain area is predicted.

[0043] <Details of the personal separation processing unit 114> 5 is a diagram showing an example of the internal configuration of the individual separation processing unit 114. In this embodiment, for each device-specific household, the individual separation processing unit 114 estimates the television broadcast viewing time, for example, the viewing time per day, of viewers with each attribute for each attribute included in the attribute configuration estimated by the attribute configuration estimation unit 113. Note that the unit time is not limited to one day, and may be one week, one month, or the like.

[0044] In this case, the individual separation processing unit 114 includes an attribute-specific viewing information generation unit 114a that uses a second model to generate (estimate) attribute-specific viewing information for each device-specific household from the viewing log information of each device-specific household, which generates (estimates) attribute-specific viewing information indicating whether or not a viewer of each attribute included in the attribute configuration of the device-specific household watched a television broadcast corresponding to any combination of a channel (or broadcast station network) and date and time data indicating the month, day of the week, time slot, etc., and a viewing time aggregation unit 114b that aggregates, from the attribute-specific viewing information, the viewing time (total viewing time over a period of a specified unit time width) for each attribute included in the attribute configuration of the device-specific household.

[0045] A television broadcast corresponding to a combination of a channel (or a broadcasting station affiliate) and date and time data indicating a month, a day of the week, and a time slot means a television broadcast that is broadcast on that channel (or a broadcasting station affiliate) at the date and time indicated by the date and time data. The time slot in the date and time data is, for example, a time slot divided into predetermined time intervals (10 minutes, 30 minutes, 1 hour, etc.).

[0046] Here, the second model used in the attribute-specific viewing information generation unit 114a is a model created for each of n types of attributes At(i) (i = 1, 2, ..., n). The second model corresponding to each attribute At(i) is a model that has been subjected to machine learning processing in advance so that, when a set of channel (or broadcasting station affiliate) and date and time data is specified, the probability (hereinafter referred to as attribute-specific viewing probability) that a viewer with attribute At(i) watched a television broadcast (hereinafter referred to as channel / date and time specified broadcast) corresponding to the specified set of channel (or broadcasting station affiliate) and date and time data can be determined from device-specific viewing log information for a predetermined period of device-specific household 20 to which members (viewers) of the attribute At(i) belong.

[0047] In addition, Figure 5 shows that the attribute-specific viewing probability, which is the probability that a member (viewer) with attribute At(n) watched a channel / date-specified broadcast, can be determined from viewing log information by device for a household with a member (viewer) with a certain attribute At(n) using the second model corresponding to the attribute At(n). The same applies to the second models corresponding to other attributes.

[0048] In the machine learning process for the second model, the learning data is the viewing situation information of each member of each sample household (sample households in areas other than the survey area) whose attribute composition is known, obtained from people meters or the like installed in each sample household, and the attribute composition of each sample household. A publicly known algorithm may be used as the algorithm for the machine learning process. Furthermore, the second model may be similar to the mathematical model for determining individual viewing behavior in Patent Document 1, for example. However, the second model may be of another form, as long as it can identify, for each attribute of the members of the device-specific household, the attribute-specific viewing probability (or a similar index value) for a channel / date / time-specific broadcast in which the channel (or broadcast station affiliate) and date / time data are arbitrarily specified from the device-specific viewing log information for the device-specific household for a predetermined period.

[0049] The attribute-specific viewing information generation unit 114a is configured to determine (estimate) for each attribute constituting the attribute configuration of a device-specific household whether or not members of each attribute have viewed the channel / date / time specified broadcast, based on the attribute-specific viewing information estimated by the second model and the device-specific viewing log information of the device-specific household.

[0050] Specifically, the attribute-specific viewing information generation unit 114a determines that a member of a device-specific household 20 has viewed a channel and date / time specified broadcast if the following conditions are met: for each attribute of the members of the device-specific household 20, the attribute-specific viewing probability value (probability value) for the channel and date / time specified broadcast specified by the second model is equal to or greater than a predetermined threshold (for example, 0.5 or greater); and the viewing log information for the device-specific household confirms that actual viewing occurred on the channel (or broadcast station affiliate) of the channel and date / time specified broadcast specified at the broadcast date and time indicated by the date and time data of the channel and date / time specified broadcast. If these conditions are not met, the unit determines that a member of the device-specific household has not viewed a channel and date / time specified broadcast specified by the member of the device-specific household.

[0051] The individual separation processing unit 114 sequentially specifies all pairs of channel (or broadcasting station network) and date and time data within a predetermined period (for example, within a one-month period) to the attribute-specific viewing information generation unit 114a configured as described above for each attribute included in the attribute configuration of each device-specific household, and estimates whether or not the channel / date and time specified broadcast corresponding to each of the pairs was viewed from the device-specific viewing log for the device-specific household for the predetermined period.

[0052] Next, the individual separation processing unit 114 executes the processing of the viewing time aggregation unit 114b. This viewing time aggregation unit 114b calculates the sum of the time durations of all channel / date / time specified broadcasts that are estimated to have been viewed for each attribute of the members of each device-specific household, and estimates the viewing time per day by dividing the sum of the time durations by the number of days in the above-mentioned predetermined period. In this way, the viewing time per day is obtained for each attribute of the members (viewers) of each device-specific household.

[0053] <Details of the viewing score distribution adjustment processing unit 115> FIG. 6 is a diagram showing an example of the internal configuration of the viewing score distribution adjustment processing unit 115. The viewing score distribution adjustment processing unit 115 includes a frequency distribution specifying unit 1151 that specifies a frequency distribution of viewing time (viewing time per day) in the survey target area (specifically, the frequency distribution of the number of viewers with respect to the viewing time) for each of n types of attributes At(i) (i = 1, 2, ···, n), and a survey target viewer selection unit 1152 that selects survey target viewers who are viewers to be surveyed for investigating viewing status such as TV viewing rate for each of n types of attributes At(i) (i = 1, 2, ···, n).

[0054] (1) Frequency distribution specifying unit 1151 The viewing time is divided into m types of time ranges T(k) (k = 1, 2, ···, m). For example, it is divided into m types of time ranges such that T(1) = 0, 0 < T(2) ≤ Tx_2, Tx_2 < T(3) ≤ Tx_3, ……, Tx_m - 2 < T(m - 1) ≤ Tx_m - 1, Tx_m - 1 < T(m). Note that the boundary values Tx_2, Tx_3, …, Tx_m - 1 of adjacent time ranges are predetermined constant values.

[0055] <{ The frequency distribution specifying unit 1151 determines, for each of n types of attributes At(i) (i = 1, 2, …, n), the total number of viewers (total number in the survey target area) having a viewing time belonging to each time range T(k) as the frequency corresponding to the time range T(k). For example, in the whole household by device type where viewers (constituent members) with a certain attribute At(j) are presumed to exist, if the total numbers of viewers (constituent members with attribute At(j)) having viewing times belonging to T(1), T(2), ···, T(m) are X1, X2, ···, Xm respectively, then X1, X2, ···, Xm are determined as the frequencies corresponding to the viewing times of each time range T(1), T(2), ···, T(m) for attribute At(j) respectively.

[0056] In this way, the frequency distribution determination unit 1151 determines a frequency distribution of viewing time for each of the n types of attributes At(i) (i=1, 2, . . . , n). FIG. 7 is a diagram for explaining the processing by the frequency distribution determination unit 1151 and the survey target audience selection unit 1152. The solid bar graph in FIG. 7 illustrates an example of the frequency distribution of viewing time determined as described above for one of the attributes At(i). In this example, the number m of types of time ranges T(k) is, for example, 10. The frequency distribution of viewing time is determined similarly for other attributes.

[0057] Additionally, the number m of types of time ranges T(k) or the boundary values ​​Tx_2, Tx_3, . . . , Tx_m-1 between adjacent time ranges do not need to be the same for all N types of attributes At(i) (i = 1, 2, . . . , n), and may be different depending on the type of attribute At(i). For example, the number m of types of time ranges T(k) corresponding to attributes whose viewing times tend to be distributed over a wide range may be greater than the number m of types of time ranges T(k) corresponding to attributes whose viewing times are relatively uniform.

[0058] (2) Survey Target Viewer Selection Division 1152 The survey target viewer selection unit 1152 performs a process of selecting survey target viewers, who are viewers to be surveyed for investigating viewing conditions such as television broadcast ratings, for each of n types of attributes At(i) (i=1, 2,..., n).

[0059] In this case, the survey target viewer selection unit 1152 selects survey target viewers based on a comparison between the frequency distribution identified by the frequency distribution identification unit 1151 and a predetermined standard frequency distribution corresponding to each attribute At(i), for each of the n types of attributes At(i) (i=1, 2,..., n).

[0060] Specifically, in this embodiment, a frequency distribution of viewing time for each attribute At(i) for all sample households is specified in advance from viewing situation information for each member of a plurality of sample households (households whose attributes are known) selected in advance in an area other than the survey area (for example, an area in the Kanto region when the survey area is a rural area such as Hiroshima) obtained from each of the sample households via a people meter or the like, and this frequency distribution is used as the reference frequency distribution. Here, the sample households are selected as households that represent the viewing situation of television broadcasts in areas other than the survey area, and the frequency distribution of viewing time for each attribute At(i) for all sample households is considered to correspond to the frequency distribution of actual viewing time for each attribute At(i) in the survey area.

[0061] Then, the survey target viewer selection unit 1152 selects the survey target viewers for each attribute At(i) so that the frequency distribution of the viewing time of the selected survey target viewers matches (or almost matches) the reference frequency distribution. Such selection can be performed for each attribute At(i), for example, as follows.

[0062] Hereinafter, any one attribute among the N types of attributes At(i) (i=1, 2, . . . , n) will be referred to as the target attribute At(x), and the process of selecting viewers to be surveyed for this target attribute At(x) will be described below. In the following description, the frequency distribution of viewing times specified by the frequency distribution specifying unit 1151 for each attribute At(i) will be referred to as the pre-selection frequency distribution, the frequency within each time range T(k) (k=1, 2, . . . , m) within the pre-selection frequency distribution will be referred to as the pre-selection frequency Xa_k, the frequency distribution of viewing times of viewers to be surveyed selected for each attribute At(i) will be referred to as the post-selection frequency distribution, and the frequency within each time range T(k) (k=1, 2, . . . , m) within the post-selection frequency distribution will be referred to as the post-selection frequency Xb_k. In addition, in the reference frequency distribution, the frequency (normalized frequency) corresponding to each time range T(k) (k=1, 2, . . . , m) is defined as the reference frequency Xs_k. As an example of the pre-selection frequency distribution for the target attribute At(x), the frequency distribution shown by the solid bar graph in Figure 7 is used, and as an example of the reference frequency distribution corresponding to the target attribute At(x), the frequency distribution shown by the dashed-dotted line with the open circles in Figure 7 as the break points is used.

[0063] (Step 1) The survey target viewer selection unit 1152 calculates the value Xa_k / Xs_k (hereinafter referred to as the reference frequency ratio Xa_k / Xs_k) by dividing the pre-selection frequency Xa_k corresponding to each of m (here, 10) time ranges T(1) to T(10) in the pre-selection frequency distribution of viewers with the target attribute At(x) by the reference frequency Xs_k corresponding to each of m time ranges T(1) to T(10) in the reference frequency distribution. Then, from the m (10) time ranges T(1) to T(10), it identifies the time range corresponding to the smallest reference frequency ratio Xa_k / Xs_k (excluding zero) (hereinafter this time range will be referred to as the minimum frequency ratio time range T(min)). For example, in the example of the pre-selection frequency distribution (frequency distribution shown by the solid bar graph) and the reference distribution shown in Figure 7, of the pre-selection frequencies Xa_1 to Xa_10 corresponding to each of the 10 time ranges T(1) to T(10), the frequency ratio to the reference Xa_9 / Xs_9 (>0) corresponding to the time range T(9) is the smallest, so T(9) is identified as the time range T(min) with the smallest frequency ratio.

[0064] (Step 2) The survey target viewer selection unit 1152 calculates, as a reference ratio, the ratio of the frequency in each of the other time ranges to the frequency in the frequency ratio minimum time range T(min) in the reference frequency distribution corresponding to the target attribute At(x). For example, in the reference frequency distribution shown in Fig. 7, the reference ratios corresponding to the time ranges T(1), T(2), , T(8), and T(10) other than the frequency ratio minimum time range T(min) (= T(9)) are calculated by the division operations Xs_1 / Xs_9, Xs_2 / Xs_9, , Xs_8 / Xs_9, and Xs_10 / Xs_9, respectively.

[0065] (Step 3) The survey target viewer selection unit 1152 sets the pre-selection frequency Xa_9 in the minimum frequency time range T(min) (=T(9)) in the pre-selection frequency distribution of viewers with the target attribute At(x) as a tentative target value for the post-selection frequency Xb_9 in the minimum frequency ratio time range T(min), and sets the integer part (integer value with decimal points rounded down) of the value obtained by multiplying the tentative target value (=pre-selection frequency Xa_9) of the post-selection frequency Xb_9 in the minimum frequency ratio time range T(min) (=T(9)) by the standard ratio corresponding to each of the other time ranges T(1), T(2), ..., T(8), T(10) as the tentative target values ​​for the post-selection frequencies Xb_1, Xb_2, ..., Xb_8, Xb_10 in each of the other time ranges T(1), T(2), ..., T(8), T(10).

[0066] Therefore, when any one of the time ranges T(1), T(2), . . . , T(8), T(10) other than the minimum time range T(min) (= T(9)) of frequency ratio is expressed as T(y) (y is 1, 2, . . . , 8, 10), the tentative target value of the post-selection frequency Xb_y corresponding to the time range T(y) is set to the integer part of the value obtained by multiplying the tentative target value (= pre-selection frequency Xa_9) of the post-selection frequency Xb_9 in the minimum time range T(min) (= T(9)) of frequency ratio by the reference ratio (= Xs_y / Xs_9) corresponding to the time range T(y).

[0067] (Step 4) The survey target viewer selection unit 1152 determines, for each time range T(k) (k=1, 2, . . . , 10), whether the tentative target value of the post-selection frequency Xb_k set in step 3 is equal to or less than the pre-selection frequency Xa_k corresponding to the time range T(k) (whether the tentative target value of the post-selection frequency Xb_k≦the pre-selection frequency Xa_k). Note that, with regard to the frequency ratio minimum time range T(min) (=T(9)), the result of the determination process will necessarily be positive, so the determination process for the frequency ratio minimum time range T(min) may be omitted.

[0068] (Step 5) If the determination result in step 4 is positive for all time ranges T(1) to T(10), the survey target viewer selection unit 1152 determines the tentative target value of the post-selection frequency Xb_k in each time range T(k) as the target value of the post-selection frequency Xb_k. For example, in FIG. 7, the determination result in step 4 is positive for all time ranges T(1) to T(10). In this case, the tentative target value of the post-selection frequency Xb_k in each time range T(k) (k=1, 2, . . . , 10) is determined as the target value of the post-selection frequency Xb_k.

[0069] (Step 6) If the judgment result of step 4 becomes negative for one or more of the time ranges T(1) to T(m), the survey target viewer selection unit 1152 sets a new tentative target value for the post-selection frequency Xb_9 in the frequency ratio minimum time range T(min) (=T(9)) to a value smaller than the pre-selection frequency Xa_9 in the frequency ratio minimum time range T(min).

[0070] In this case, the new tentative target value of the post-selection frequency Xb_9 in the frequency ratio minimum time range T(min) (=T(9)) is set so as to satisfy the condition that the judgment result of step 4 becomes positive for all time ranges T(1) to T(10) when the tentative target value is used to set the post-selection frequency in other time ranges by the same processing as in step 3. Furthermore, the new tentative target value of the post-selection frequency Xb_9 in the frequency ratio minimum time range T(min) is set so as to be as close as possible to the pre-selection frequency Xa_9 in the frequency ratio minimum time range T(min) within the range that satisfies the above condition.

[0071] Setting a new tentative target value for the post-selection frequency Xa_9 in the frequency ratio minimum time range T(min) in this manner can be achieved, for example, by repeating the process of decreasing the tentative target value from the pre-selection frequency Xa_9 by a predetermined amount until the above condition is met.

[0072] Then, after setting the new tentative target value of the post-selection frequency Xb_k in each time range T(k) so as to satisfy the above conditions, the survey target viewer selection unit 1152 determines the tentative target value of the post-selection frequency Xb_k in each time range T(k) as the target value of the post-selection frequency Xb_k in each time range T(k), as in step 5.

[0073] (Step 7) After determining the target value of the post-selection frequency Xb_k for each time range T(k) in step 5 or step 6, the survey target viewer selection unit 1152 removes, for each time range T(k) (k=1, 2, . . . , m), from all viewers (viewers of the target attribute At(x)) who have viewing times belonging to each time range T(k), a number of viewers equivalent to the difference between the pre-selection frequency Xa_k corresponding to that time range T(k) and the target value of the post-selection frequency Xb_k, and selects the remaining viewers as survey target viewers for the target attribute At(x). In this case, viewers to be removed for each time range T(k) are selected, for example, randomly. As a result, the survey target viewers (survey target viewers of the target attribute At(y)) for each time range T(k) (k=1, 2, . . . , m) are selected so that their total number matches the target value of the post-selection frequency Xb_k.

[0074] For example, in FIG. 7, within the time range T(k) (k=1, 2, . . . , 10), for each time range in which the pre-selection frequency is greater than the target value for the post-selection frequency, the number of viewers corresponding to the frequency in the gray area is removed. As a result, for the attribute of interest At(x), survey target viewers are selected so that the post-selection frequency distribution of viewing time matches the reference frequency distribution. Survey target viewers are then selected in the same manner as above for other attributes. The selected survey target viewers are stored in association with the television devices 23 of the device-specific households 20 to which they belong.

[0075] <Details of the Individual Viewer Rating Estimation Unit 116> After the target viewers for each attribute At(i) are selected, the individual viewership estimation unit 116 uses the selected target viewers to estimate viewership (corresponding to individual viewership) for each of n types of attributes At(i) for the television broadcast (hereinafter referred to as the target broadcast) that is the target of the survey of viewing conditions in the target area.

[0076] In this case, the individual audience rating estimation unit 116 estimates whether or not the target surveyed viewer watched the surveyed broadcast by performing the same processing as the attribute-specific viewing information generation unit 114a of the individual separation processing unit 114 (processing that uses a second model corresponding to the attributes of the surveyed viewer) from the viewing log information for a specified period of the device-specific household to which each surveyed viewer belongs, and the channel (or broadcasting station network) and date and time data of the surveyed broadcast.

[0077] Then, the individual audience rating estimation unit 116 estimates whether or not all of the surveyed viewers in the surveyed area have watched the surveyed broadcast, and then calculates the viewing status data of the surveyed broadcast for each attribute At(i) based on the estimation results, for example, using equation (1). Viewing status data for attribute At(i) = m(i) / M(i) (1)

[0078] Here, M(i) is the total number of surveyed viewers with attribute At(i), and m(i) is the total number of surveyed viewers with attribute At(i) who are estimated to have watched the surveyed broadcast.

[0079] This allows us to obtain viewing status data equivalent to the viewing rate for each attribute At(i) in the survey area. In this case, the survey target viewers are selected so that the post-selection frequency distribution of viewing time matches the standard frequency distribution based on viewing data for sample households in the survey area, so we can obtain highly reliable viewing data.

[0080] <Alternative patterns based on viewing status information such as people meters in other areas> FIG. 8 is a substitution pattern table 800 showing examples of substitution patterns based on viewing situation information such as people meters in other areas.

[0081] The substitution pattern table 800 includes as its constituent items an information collection area 801 indicating the area where viewing status information and viewing log information were collected using a people meter or the like (denoted as PM in Figure 7), a sample count 802 indicating the number of samples of collected PM viewing status information, a current status 803 indicating the information pattern used in calculating the current viewership rating, pattern 1_804 which is the first substitution pattern, pattern 2_805 which is the second substitution pattern, and pattern 3_806 which is the third substitution pattern.

[0082] In Figure 8, the information collection area 801 is divided into the Kanto area, the Kansai area, the Nagoya area, the Northern Kyushu / Sapporo area, and other areas, but the division method is not limited to these and any division method may be used.

[0083] As shown by sample number 802, the number of PM samples is small in some areas, which may make it difficult to maintain PM surveys. As shown by current situation 803, currently, audience ratings are calculated in all areas based on the number of PM samples in that area, but for areas with a small number of PM samples, there is a possibility that the reliability of the calculated audience ratings may be questionable. For this reason, as mentioned above, as shown in pattern 1_804 to pattern 3_806, viewing log information obtained from each television manufacturer is used for the survey area, and individual audience ratings are estimated by applying the viewing log information of the survey area to a prediction model obtained by learning from viewing situation information of PMs in areas other than the survey area.

[0084] In pattern 1_804, if the survey area is set to "areas other than those listed above," the viewing log information for the survey area "areas other than those listed above" can be used to estimate individual viewership ratings by applying the viewing log information for the survey area "areas other than those listed above" to a prediction model trained on viewing status information from PMs in at least one of the following areas: the Kanto region, the Kansai region, the Nagoya region, or the Northern Kyushu / Sapporo region. For example, if the survey area is the Kumamoto region, the individual viewership rating can be predicted by applying the viewing log information for the Kumamoto region to a prediction model trained on PMs in the Northern Kyushu region. As shown in Figure 4, when predicting individual viewership ratings for a specific local television station, a prediction model trained on viewing status information (information obtained by PMs) from affiliated key stations is used. Therefore, when using a prediction model trained on viewing situation information for at least one of the Kanto, Kansai, Nagoya, or Northern Kyushu / Sapporo regions for the "areas other than the above" as shown in Pattern 1_804, it is possible to use a prediction model trained on at least one of the affiliated stations corresponding to the television stations in the "areas other than the above" (survey target area) (key affiliated stations in the Kanto region, or corresponding affiliated stations in the Kansai, Nagoya, Northern Kyushu / Sapporo regions).

[0085] In Pattern 2_805, if the survey area is either "areas other than those listed above" or "Northern Kyushu and Sapporo," the individual audience rating can be estimated by applying the viewing log information for the survey areas "areas other than those listed above" and "Northern Kyushu and Sapporo" to a prediction model obtained by training on PM viewing status information for at least one of the following areas other than the survey area: "Kanto," "Kansai," or "Nagoya." For example, if the survey area is the Toyama area, the individual audience rating can be predicted by applying the viewing log information for the Toyama area to a prediction model obtained by training on PM in the Nagoya area. Note that, in the case of Pattern 2_805, as with Pattern 1_804, viewing status information for competing affiliated stations can be used when constructing a prediction model.

[0086] In pattern 3_806, if the survey target area is either "areas other than the above," "northern Kyushu and Sapporo area," "Nagoya area," or "Kansai area," the individual audience rating can be estimated by using the viewing log information of "areas other than the above," "northern Kyushu and Sapporo area," "Nagoya area," or "Kansai area" and applying the viewing log information of either of the survey target areas "areas other than the above," "northern Kyushu and Sapporo area," "Nagoya area," or "Kansai area" to a prediction model obtained by training with viewing status information of PMs in the "Kanto area," which is not the survey target area. For example, if the Kansai area is the survey target area, the individual audience rating can be predicted by applying the viewing log information of the Kansai area to a prediction model obtained by training with PMs in the Kanto area.

[0087] <Other> In this embodiment, the provisional target value of the post-selection frequency Xb_k set as described above for each time range T(k) is determined as the target value of the post-selection frequency Xb_k for each time range T(k), with the necessary condition that the provisional target value of the post-selection frequency Xb_k is less than or equal to the pre-selection frequency Xa_k for all time ranges T(1) to T(10).

[0088] However, even if only the provisional target value of the post-selection frequency corresponding to each of a portion of the time ranges T(1) to T(10) is larger than the pre-selection frequency corresponding to each of the portion of the time ranges, if the difference between the provisional target value of the post-selection frequency corresponding to each of the portion of the time ranges and the pre-selection frequency is sufficiently small (below a predetermined threshold), the target value of the post-selection frequency corresponding to each of the portion of the time ranges can be set to match the pre-selection frequency, and for each time range other than the portion of the time ranges (time ranges in which the provisional target value of the post-selection frequency is less than the pre-selection frequency), the provisional target value of the post-selection frequency can be set as the target value as is.

[0089] Furthermore, in the present embodiment, the viewing information processing device 100 is configured to include the individual viewing rating estimation unit 116, but it may be configured not to include the individual viewing rating estimation unit 116. In this case, the individual viewing rating estimation unit 116 is provided in a device separate from the viewing information processing device 100.

[0090] Furthermore, in this embodiment, the attribute configuration and attribute-specific viewing information of each device-specific household are estimated from device-specific viewing log information, and the viewing time for each attribute of viewers in each device-specific household (viewing time in a predetermined unit time period) is estimated from the estimated attribute-specific viewing information. However, in the present invention, the method for acquiring the viewing time for each attribute of viewers in each device-specific household is not limited to the above method. For example, each device-specific household may be equipped with an appropriate device capable of measuring or estimating the viewing time for each attribute of viewers in each device-specific household, and the viewing time for each attribute of viewers in each device-specific household may be acquired from that device. Alternatively, for example, each device-specific household may be equipped with an appropriate device capable of generating and outputting viewing log information for each attribute of viewers in each device-specific household, and the viewing time for each attribute of viewers in each device-specific household may be estimated from the viewing log information output from that device.

[0091] <Summary> (i) This embodiment discloses a viewing information processing device 100 that predicts individual audience ratings for a survey area by applying viewing log information (device log information) obtained from television devices installed in each household in the survey area to a prediction model. While the prediction model is typically trained using viewing situation information obtained from people meters or the like installed in sample households in the same area as the survey area, this embodiment uses a model trained based on viewing situation information from people meters or the like installed in sample households in areas other than the survey area. The prediction model includes a first model for calculating the presence probability by attribute for each household to estimate attributes indicating the composition of each household, and a second model for separating the viewing time for each attribute for each household. The viewing information processing device 100 then applies the viewing log information (device log information) of each household to the first model to estimate attributes (family composition of the target household), and applies the viewing log information of each household to the second model to estimate the viewing time for each television broadcast for each attribute of each household during a predetermined period. The viewing information processing device 100 also identifies the frequency distribution of the number of viewers relative to viewing time for each attribute of each household in the survey area (see FIG. 7 ), compares the frequency distribution of the number of viewers for each attribute with a predetermined reference frequency distribution for each attribute based on viewing situation information for areas other than the survey area, and selects viewers to be surveyed in the survey area for each attribute so that the frequency distribution of the number of viewers to be surveyed matches the reference frequency distribution. Furthermore, the viewing information processing device 100 calculates the audience rating of the surveyed television broadcast based on an estimation result of whether or not the selected viewers watched the surveyed television broadcast. In this way, the prediction model uses a model trained based on viewing situation information collected via a people meter or the like in areas other than the survey area. Therefore, even if a sufficient amount of viewing situation information cannot be obtained in the survey area using a people meter or the like, audience ratings for the surveyed area can be efficiently obtained and the validity of the calculated individual audience rating can be ensured.

[0092] (ii) The prediction model of this embodiment can be a model obtained by learning based on viewing situation information of key stations affiliated with the target television station in the survey area. This makes it possible to build a model that outputs more accurate prediction values.

[0093] (iii) As shown in Figure 8, the prediction model can also be a model obtained by learning based on viewing situation information in multiple regions other than the survey target region. By using viewing situation information in multiple regions, the prediction model can be set to output more accurate prediction values. Note that the prediction model may also be a model obtained by learning based on viewing situation information of multiple affiliated stations (affiliated television stations in regions other than the survey target region) corresponding to the target television station in the survey target region.

[0094] (iv) The functions of the present embodiment can also be realized by software program code. In this case, a storage medium on which the program code is recorded is provided to a system or device, and the computer (or CPU or MPU) of the system or device reads the program code stored on the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the above-mentioned embodiment, and the program code itself and the storage medium on which it is stored constitute the present invention. Examples of storage media for providing such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, optical disks, magneto-optical disks, CD-Rs, magnetic tapes, non-volatile memory cards, and ROMs.

[0095] In addition, an operating system (OS) running on a computer may perform some or all of the actual processing based on instructions in the program code, and the functions of the above-described embodiments may be realized by this processing. Furthermore, after the program code is read from a storage medium and written to a memory on a computer, a CPU of the computer may perform some or all of the actual processing based on instructions in the program code, and the functions of the above-described embodiments may be realized by this processing.

[0096] Furthermore, the program code of the software that realizes the functions of the embodiments and each example may be distributed via a network and stored in a storage means such as a hard disk or memory of the system or device, or in a storage medium such as a CD-RW or CD-R, so that when in use, the computer (or CPU or MPU) of the system or device reads and executes the program code stored in the storage means or storage medium.

[0097] The processes and techniques described herein are not inherently related to any specific device and can be implemented by a combination of components. Various types of general-purpose devices can also be added. A dedicated device may be constructed to perform the functions of this embodiment and each example. Various functions can also be formed by appropriately combining multiple components disclosed in this embodiment and each example. For example, some components may be omitted from all the components shown in the embodiment and each example, or components from different examples may be appropriately combined.

[0098] Although specific embodiments are described in the present invention, they are in all respects for the purpose of explanation (understanding the technology of the present invention) and not for the purpose of limitation. Those skilled in the art will recognize that there are many combinations of hardware, software, and firmware suitable for implementing the technology of the present invention. For example, the described software can be implemented in a wide range of programming or scripting languages, such as assembler, C / C++, Perl, Shell, PHP, Java (registered trademark), etc.

[0099] Furthermore, in the above-described embodiment, the control lines and information lines are those that are considered necessary for the explanation, and not all control lines and information lines in the product are necessarily shown. All components may be interconnected.

[0100] In addition, other implementations of the present invention will become apparent to those skilled in the art from consideration of the present embodiments and examples. The specification and examples are exemplary only, with the scope and spirit of the present invention being indicated by the following claims. [Explanation of symbols]

[0101] 100 Viewing information processing device 110 Processor (control device) 111 Survey Area Viewing Log Information Acquisition Department 112 Reference area viewing situation information acquisition unit 113 Attribute composition estimation part 114 Personal Separation Processing Unit 115 Viewing Minute Distribution Adjustment Processing Unit 116 Individual Viewership Estimation Department 120 Storage Devices 130 Input Devices 140 output devices 150 Communication Devices 200_1 to 200_M TV devices 1 to M 300_1 to 300_N TV manufacturer 1 server device to TV manufacturer N server device 400, 500 network

Claims

1. A viewing information processing device that applies viewing log information obtained from television devices installed in each household in a survey area to a prediction model to predict individual audience ratings in the survey area, a storage device that stores at least a program for predicting the individual audience rating in the survey area and the prediction model; a control device that reads the program, the viewing log information, and the prediction model from the storage device and executes the program to predict the individual audience rating in the survey target area; The prediction model is a model obtained by learning based on viewing situation information obtained from dedicated devices installed in sample households in areas other than the survey target area, and includes a first model for calculating the presence probability by attribute of each household and estimating attributes indicating the composition of each household, and a second model for separating the viewing time of each attribute of each household, The control device a process of applying the viewing log information of each of the households to the first model to estimate the attributes; applying the viewing log information of each of the households to the second model to estimate the viewing time of each television broadcast for each of the attributes of each of the households during a predetermined period; A process of identifying a frequency distribution of the number of viewers with respect to the viewing time for each of the attributes of each of the households in the survey area; a process of comparing the frequency distribution of the number of viewers for each attribute with a predetermined standard frequency distribution for each attribute based on the viewing situation information of areas other than the survey target area, and selecting viewers to be surveyed in the survey target area for each attribute so that the frequency distribution of the number of viewers to be surveyed coincides with the standard frequency distribution; a process of calculating the audience rating of the television broadcasting subject to the survey based on an estimation result of whether or not the selected viewers to be surveyed have watched the television broadcasting subject to the survey; A viewing information processing device that executes the above.

2. In claim 1, The viewing information processing device, wherein the prediction model is a model obtained by learning based on the viewing situation information of key stations affiliated with the target television station in the survey target area.

3. In claim 1, The viewing information processing device, wherein the prediction model is a model obtained by learning based on the viewing situation information in a plurality of regions other than the survey target region.

4. In claim 3, The viewing information processing device, wherein the prediction model is a model obtained by learning based on the viewing situation information of a plurality of affiliated stations corresponding to the target television station in the survey target area.

5. A viewing information processing method using a computer to predict individual audience ratings in a survey area by applying viewing log information obtained from television devices installed in each household in the survey area to a prediction model, comprising: a program for the computer to predict individual audience ratings in the survey area; reading the program, the viewing log information, and the prediction model from a storage device that stores at least the program, the viewing log information, and the prediction model; The computer executes the program to predict the individual audience rating in the survey area, The prediction model is a model obtained by learning based on viewing situation information obtained from dedicated devices installed in sample households in areas other than the survey target area, and includes a first model for calculating the presence probability by attribute of each household and estimating attributes indicating the composition of each household, and a second model for separating the viewing time of each attribute of each household, Predicting the individual viewership rate in the survey target area includes: applying the viewing log information of each of the households to the first model to estimate the attributes; applying the viewing log information of each of the households to the second model to estimate the viewing time of each television broadcast for each of the attributes of each of the households during a predetermined period; Identifying a frequency distribution of the number of viewers with respect to the viewing time for each of the attributes of each of the households in the survey area; comparing the frequency distribution of the number of viewers for each attribute with a predetermined standard frequency distribution for each attribute based on the viewing situation information of areas other than the survey target area, and selecting the viewers to be surveyed in the survey target area for each attribute so that the frequency distribution of the number of viewers to be surveyed coincides with the standard frequency distribution; calculating an audience rating for the television broadcasting subject to the survey based on an estimation result of whether or not the selected viewers to be surveyed have watched the television broadcasting subject to the survey; A viewing information processing method, including:

Images