Exogenous event-based viewership causal attribution methods, systems, devices, and media

Abstract

The present disclosure relates to an exogenous event-based viewing causal attribution method, system, device and medium. The method comprises: acquiring multi-source viewing behavior data and external dimension data of a target channel; constructing a structural causal model based on the multi-source viewing behavior data and the external dimension data and a preset intervention variable; identifying variables of the external dimension data based on a preset instrumental variable condition to obtain a target instrumental variable; in the case that the number of the target instrumental variable is consistent with the number of the preset intervention variable, estimating a causal effect of the preset intervention variable on the viewing rate based on the target instrumental variable and the structural causal model through a two-stage regression model to obtain a causal effect estimation value; and performing viewing causal attribution analysis of a target program within a preset time window according to the causal effect estimation value to obtain an attribution result.

Description

Technical Field

[0001] This disclosure relates to the field of broadcast television data processing technology, and more specifically, to a method, system, device, and medium for attributing viewing causality based on exogenous events. Background Technology

[0002] In broadcast television operations, accurately assessing the impact of programming on viewership is crucial for optimizing content strategies. Current technologies often employ correlation analysis or traditional regression models to quantify programming effectiveness. However, viewership behavior is influenced by a complex interplay of factors, including program quality, competitor activity, and seasonal cycles. Furthermore, these factors commonly exhibit bidirectional causality or omitted variables in their relationship with programming decisions. For instance, high-viewership periods often coincide with both high-quality content and favorable scheduling. Traditional methods struggle to isolate the independent contributions of each factor, leading to significant attribution biases and even erroneous conclusions. While some studies have attempted to mitigate endogeneity by introducing difference methods or fixed-effects models, these still rely on strong assumptions, lack systematic utilization of exogenous shocks, and suffer from inadequate verification mechanisms. Summary of the Invention

[0003] This disclosure provides a method, system, device, and medium for attributing viewership causality based on exogenous events.

[0004] According to a first aspect of this disclosure, a viewership causal attribution method based on exogenous events is provided, the method comprising: Acquire multi-source viewing behavior data and external dimension data of the target channel; wherein, the multi-source viewing behavior data includes at least a portion of two-way digital television viewing behavior data, interactive network television viewing behavior data, and Internet television viewing behavior data; Based on the multi-source viewing behavior data, external dimension data, and preset intervention variables, a structural causal model is constructed; wherein, the structural causal model is a directed acyclic graph model that characterizes the causal relationship between the viewing behavior variables of the target channel. Based on preset instrumental variable conditions, the external dimension data is used to identify variables to obtain target instrumental variables; wherein, the target instrumental variables are variables derived from exogenous events and whose explanatory power for viewing behavior meets preset correlation conditions, and the target instrumental variables include at least some of the following: workday adjustment variables, breaking news event variables, and channel exclusive broadcasting rights change variables; When the number of target instrumental variables is consistent with the number of preset intervention variables, the causal effect of the preset intervention variables on viewership is estimated by a two-stage regression model based on the target instrumental variables and the structural causal model, and the causal effect estimate is obtained. Based on the estimated causal effect, a causal attribution analysis of the target program's viewership within a preset time window is performed to obtain the attribution results.

[0005] Optionally, the construction of a structural causal model based on the multi-source viewing behavior data, external dimension data, and preset intervention variables includes: Based on the multi-source viewing behavior data and external dimension data, the basic viewing indicators of the target channel are calculated; wherein, the basic viewing indicators include at least a portion of viewership rating, viewership share, and viewership loyalty. Based on the control variables and ratings in the basic viewership indicators, as well as the preset intervention variables, the causal side orientation logic is determined; wherein, the causal side orientation logic is the causal logical relationship between any two variables among the control variables, preset intervention variables, and ratings. Based on the causal edge orientation logic, a structural causal model is constructed.

[0006] Optionally, the step of identifying target instrumental variables from the external dimension data based on preset instrumental variable conditions includes: Based on the historical viewing behavior data of the target channel, at least one candidate instrumental variable is set; For each candidate instrumental variable, a first regression model is constructed based on the control variables in the structural causal model; wherein the first regression model uses a preset intervention variable as the dependent variable. Based on the historical viewing behavior data of the target channel and the first regression model corresponding to each candidate instrumental variable, regression statistics were performed to obtain the F-statistic corresponding to each candidate instrumental variable. Candidate instrumental variables whose F-statistics reach a preset F-statistic threshold are identified as target instrumental variables; wherein, the preset correlation condition includes the preset F-statistic threshold.

[0007] Optionally, the step of estimating the causal effect of the preset intervention variable on viewership ratings using a two-stage regression model based on the target instrumental variable and the structural causal model to obtain the causal effect estimate includes: Based on the target instrumental variable and the control variables in the structural causal model, a second regression model corresponding to the target instrumental variable is constructed; wherein, the second regression model uses a preset intervention variable as the dependent variable; Based on the historical viewing behavior data of the target channel and the second regression model, the model coefficient set of the second regression model is obtained; wherein, the model coefficient set includes the intercept term and the coefficients of each variable in the second regression model; Regression statistics are performed based on the model coefficient set and the second regression model to obtain the predicted values ​​of the preset intervention variables; Based on the predicted values ​​of the preset intervention variables and the control variables in the structural causal model, a third regression model is constructed; wherein, the third regression model uses viewership rating as the dependent variable. Based on the historical viewing behavior data of the target channel and the third regression model, regression statistics are performed to obtain the causal effect estimate; wherein the causal effect estimate is the coefficient of the preset intervention variable.

[0008] Optionally, after obtaining the causal effect estimate, the method further includes: Based on the historical viewership data of the target channel, the target instrumental variable is verified by matching and comparing at least two independent events on the corresponding days to obtain the verification estimate for each independent event. If the validation estimates corresponding to each independent event are consistent with the effect direction of the causal effect estimates, then the target instrumental variable is an effective instrumental variable.

[0009] Optionally, when the number of target instrumental variables is greater than the number of pre-defined intervention variables, the method further includes: Based on the target instrumental variables and the control variables in the structural causal model, a fourth regression model is constructed corresponding to all target instrumental variables; wherein, the fourth regression model uses a preset intervention variable as the dependent variable; Based on the historical viewership data of the target channel, all instrumental variables, and control variables, a two-stage least squares method was performed to estimate the causal effect and the residuals of the fourth regression model. Based on the residuals, least squares regression was performed on all target instrumental variables and all control variables to obtain the goodness of fit. A test statistic is constructed by multiplying the number of data sets of historical viewership data for the target channel by the goodness of fit. Calculate the significance probability value based on the test statistic and its corresponding chi-square distribution; When the significance probability value is greater than the preset probability threshold, all target instrumental variables are valid instrumental variables.

[0010] Optionally, the step of performing a causal attribution analysis on the target program's viewership within a preset time window based on the estimated causal effect, to obtain the attribution results, includes: Based on the estimated causal effect and the causal effect values ​​corresponding to each control variable, the viewership causal attribution model of the target program is obtained; wherein, the control variable is the exogenous variable in the structural causal model, and the causal effect value corresponding to each control variable is the coefficient of each control variable in the second stage regression model of the two-stage regression model; Based on the aforementioned viewership causal attribution model and a preset time window, attribution results are obtained; wherein, the attribution results are the causal contribution values ​​or proportions of each dimension within the preset time window; the causal contribution values ​​of each dimension are determined by the product of the causal effect value corresponding to each variable and the variation range of each variable within the preset time window.

[0011] According to a second aspect of this disclosure, a viewership causal attribution system based on exogenous events is provided, the system comprising: The data acquisition module is used to acquire multi-source viewing behavior data and external dimension data of the target channel; wherein, the multi-source viewing behavior data includes at least a portion of two-way digital television viewing behavior data, interactive network television viewing behavior data, and Internet television viewing behavior data; The causal modeling module is used to construct a structural causal model based on the multi-source viewing behavior data, external dimension data, and preset intervention variables; wherein, the structural causal model is a directed acyclic graph model that characterizes the causal relationship between the viewing behavior variables of the target channel; The instrumental variable identification module is used to identify variables in the external dimension data based on preset instrumental variable conditions to obtain target instrumental variables; wherein, the target instrumental variables are variables derived from exogenous events and whose explanatory power for viewing behavior meets preset correlation conditions, and the target instrumental variables include at least a portion of the following: workday adjustment variables, breaking news event variables, and channel exclusive broadcasting rights change variables; The causal effect estimation module is used to estimate the causal effect of the preset intervention variables on viewership ratings based on the target instrumental variables and the structural causal model, using a two-stage regression model, when the number of target instrumental variables is consistent with the number of preset intervention variables, and to obtain the causal effect estimate. The attribution analysis module is used to perform causal attribution analysis on the target program's viewership within a preset time window based on the estimated causal effect value, and to obtain the attribution results.

[0012] According to a third aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and Memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, such that the instructions are executed by at least one processor to enable the at least one processor to perform the method of the first aspect.

[0013] According to a fourth aspect of this disclosure, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method of the first aspect.

[0014] This disclosure embodiment constructs a structural causal model of the causal relationship between viewership variables by integrating multi-source viewing behavior data and external dimension data. Based on this, it rigorously selects target instrumental variables that meet the relevance conditions, and then uses a two-stage regression model to overcome endogeneity problems such as omitted variable bias and bidirectional causality. It estimates the net causal effect of the preset intervention variables on viewership. Finally, it constructs an attribution model by combining the coefficients of control variables, realizing the explainable attribution of viewership fluctuations within a specific time window, thereby significantly improving the scientific nature of viewership analysis and its decision support capabilities.

[0015] Other features and advantages of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments of the present disclosure and, together with their description, serve to explain the principles of the present disclosure.

[0017] Figure 1 This is a flowchart illustrating a viewership causal attribution method based on exogenous events provided in an embodiment of this disclosure.

[0018] Figure 2 This is a schematic diagram of a cause-and-effect graph of viewing behavior structure provided in an embodiment of this disclosure.

[0019] Figure 3 This is a schematic diagram of the structure of a viewership causal attribution system based on exogenous events provided in an embodiment of this disclosure.

[0020] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0021] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present disclosure.

[0022] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit this disclosure or its application or use.

[0023] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0024] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0025] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0026] Figure 1 The diagram illustrates a flowchart of a viewership causal attribution method based on exogenous events provided in an embodiment of this disclosure, as shown below. Figure 1 As shown, the method includes steps S110 to S150: Step S110: Obtain multi-source viewing behavior data and external dimension data of the target channel.

[0027] The multi-source viewing behavior data includes at least a portion of viewing behavior data from two-way digital video broadcasting (DVB), interactive network television (IPTV), and over-the-top (OTT) television. Two-way digital video broadcasting behavior data reflects the linear viewing habits and on-demand viewing behavior of traditional cable TV users; interactive network television viewing behavior data includes user interaction logs from operator-owned networks, providing high-precision, second-level channel-changing records; and over-the-top (OTT) television viewing behavior data covers smart TV terminals accessed via the public internet, reflecting the viewing characteristics of younger demographics and a preference for on-demand viewing.

[0028] Multi-source viewing behavior data can be obtained in the form of logs, which may include unique identifiers, timestamps, viewing types, channel codes, program codes, behavior types, and viewing durations.

[0029] External data can include national statutory holidays and work schedules, historical meteorological data, timelines of major breaking news events, programming schedules of key programs on competitor channels, and social media sentiment indices. These data, as potential exogenous or control variables, can help the model isolate the interference of non-content factors on viewership ratings, providing a clean external context for subsequent causal inferences.

[0030] External dimension data can be obtained in the form of program metadata or event data. Program metadata may include program type tags, start time, end time, broadcast channel, and premiere status. Event data may include at least some of the following: holiday tags, major event tags, and seasonal tags.

[0031] By cleaning, deduplicating, aligning timestamps, and mapping user identity documents (IDs) to the aforementioned multi-source viewing behavior data and external dimension data, a unified full-domain viewing behavior dataset can be constructed, effectively avoiding sample bias caused by a single data source. The cleaning process can refer to the industry standard GY / T418—2024 "Specification for Cleaning Big Data of Broadcast Television and Online Audiovisual Viewing".

[0032] Step S120: Based on multi-source viewing behavior data, external dimension data, and preset intervention variables, construct a structural causal model.

[0033] Among them, the structural causal model is a directed acyclic graph (DAG) model that represents the causal relationship between viewing behavior variables of the target channel.

[0034] Pre-defined intervention variables refer to the core strategies or events whose causal effects the analysis subject wishes to assess, such as prime-time scheduling, the premiere of a new variety show, or a specific marketing campaign. Because the viewership system is a complex dynamic system, direct regression analysis is highly susceptible to endogeneity issues such as omitted variable bias and bidirectional causality.

[0035] In some examples, step S120 specifically includes: Based on multi-source viewing behavior data and external dimension data, the basic viewership indicators for the target channel are calculated. Based on the control variables and viewership ratings within the basic viewership indicators, as well as preset intervention variables, the causal side-orientation logic is determined. According to the causal side-orientation logic, a structural causal model is constructed.

[0036] In this example, the basic viewership metrics include at least some of the following: viewership ratings, viewership share, and viewership loyalty.

[0037] Ratings are calculated as the ratio of viewed duration to the ideal total viewed duration (the product of program length and number of viewers), comprehensively reflecting the program's viewership level. Market share is calculated as the ratio of the total viewed duration of competing products in the previous statistical period to the total viewed duration of all channels in the same time slot, reflecting the competitiveness of those competing products within the same time slot. Loyalty is calculated as the ratio of the average viewed duration of the target channel in historical periods (excluding the impact of current programming strategies) to the program length, reflecting user engagement.

[0038] In this example, the causal edge orientation logic refers to the causal logical relationship between any two of the control variables, the preset intervention variable, and the viewership rating. Control variables are confounding factors that may affect the pre-defined intervention variables and / or outcome variables (ratings), such as program quality, competitor ratings, seasonality, etc. Causal side-direction logic is determined based on domain knowledge and statistical tests (such as Granger causality tests and PC algorithms).

[0039] For example, Figure 2 A schematic diagram of a viewing behavior structure causal graph provided in an embodiment of this disclosure is shown, such as... Figure 2 As shown, when the preset intervention variable is broadcast scheduling and the control variables are program quality, competitor ratings, and seasonality, the causal edges include broadcast scheduling-ratings, program quality-ratings, competitor ratings-ratings, seasonality-ratings, program quality-broadcast scheduling, and competitor ratings-broadcast scheduling. It is necessary to ensure that the model is a directed acyclic graph (DAG) to avoid circular causality. The final DAG clearly depicts the causal path from the control variables and preset intervention variables to ratings, providing a theoretical basis for subsequent instrumental variable selection.

[0040] Step S130: Based on preset instrumental variable conditions, identify variables in the external dimension data to obtain target instrumental variables. Target instrumental variables are variables derived from exogenous events and whose explanatory power for viewing behavior meets preset relevance conditions. Target instrumental variables include at least a portion of variables related to adjusted workdays, breaking news events, and changes in channel exclusive broadcasting rights.

[0041] Instrumental variables (IVs) are key to solving endogeneity problems. They are typically identified and selected from event data in external dimension data.

[0042] Effective instrumental variables must satisfy the requirements of relevance and exogeneity. Relevance means that the instrumental variable is strongly correlated with the endogenous explanatory variable (the pre-defined intervention variable), i.e., it has sufficient explanatory power. Exogeneity means that it only affects the ratings through the pre-defined intervention variable and is not correlated with the error term. This step focuses on the statistical verification of relevance.

[0043] In some examples, step S130 specifically includes: Based on the historical viewing behavior data of the target channel, at least one candidate instrumental variable is set.

[0044] The instrumental variables to be selected should be exogenous events, such as variables related to adjusted workdays, breaking news events, changes in exclusive channel broadcasting rights, and extreme weather. These variables naturally satisfy the exogeneity constraint and are not affected by viewership feedback.

[0045] For each candidate instrumental variable, a first regression model is constructed based on the control variables in the structural causal model.

[0046] In this model, the pre-defined intervention variable is the dependent variable, and the candidate instrumental variable and control variable are the independent variables. This is the first stage regression of the two-stage least squares (2SLS) method.

[0047] Regression statistics were performed based on historical viewing behavior data of the target channel and the first regression model corresponding to each candidate instrumental variable to obtain the F-statistic for each candidate instrumental variable. The F-statistic is used to test the correlation between the instrumental variables and the viewership rating.

[0048] Candidate instrumental variables whose F-statistics reach a preset F-statistic threshold are identified as target instrumental variables. According to the Stock-Yogo weak instrumental variable test criteria, the preset F-statistic threshold is typically set to 10 (or more strictly, 16.38). If the F-statistic is greater than 10, the variable is considered not a weak instrumental variable, meets the preset correlation condition, and can be used as a target instrumental variable.

[0049] Step S140: When the number of target instrumental variables is consistent with the number of preset intervention variables, the causal effect of the preset intervention variables on viewership is estimated by a two-stage regression model based on the target instrumental variables and the structural causal model, and the causal effect estimate is obtained.

[0050] In a two-stage regression model, the first stage regression is used to obtain the predicted values ​​of the pre-defined intervention variables, and the second stage regression is used to estimate the causal effect.

[0051] In some examples, step S140 specifically includes: Based on the target instrumental variable and the control variables in the structural causal model, a second regression model corresponding to the target instrumental variable is constructed. The second regression model uses the pre-defined intervention variable as the dependent variable. This step constitutes the first-stage regression.

[0052] Based on historical viewing behavior data of the target channel and a second regression model, the model coefficient set of the second regression model is obtained. The model coefficient set includes the intercept term and the coefficients of each variable in the second regression model.

[0053] Regression statistics were performed based on the model coefficient set and the second regression model to obtain the predicted values ​​of the preset intervention variables. These predicted values ​​eliminated the endogenous components related to the error term and retained only the exogenous variations driven by the instrumental variables.

[0054] Based on the predicted values ​​of the pre-defined intervention variables and the control variables in the structural causal model, a third regression model is constructed. In this model, viewership ratings are the dependent variable. This step constitutes the second-stage regression.

[0055] Regression statistics were performed based on historical viewership data of the target channel and a third regression model to obtain an estimate of the causal effect. The causal effect estimate is the coefficient of the pre-defined intervention variable. This coefficient is the Local Average Treatment Effect (LATE), representing the pure causal effect of the intervention variable on viewership under the influence of the instrumental variable.

[0056] In some examples, to ensure the reliability of causal effect estimates, this disclosure also provides a mechanism for validating the validity of instrumental variables, as follows: After obtaining the causal effect estimate, the method further includes: Based on historical viewership data of the target channel, the instrumental variable is validated by matching at least two independent events to corresponding days, yielding validation estimates for each independent event. For example, the causal effect is estimated by selecting the same work schedule adjustment in two different years.

[0057] If the validation estimates for each independent event are consistent with the causal effect estimates in terms of effect direction, then the target instrumental variable is an effective instrumental variable.

[0058] In this example, the direction of the effect refers to the causal effect on the ratings. Consistent direction of the effect means that when the preset intervention variable changes from 0 to 1, the trend of the ratings changes in the same way.

[0059] In some examples, when the number of target instrumental variables is greater than the number of pre-defined intervention variables, steps S130-S140 can be replaced with the following steps: Based on the target instrumental variables and the control variables in the structural causal model, a fourth regression model is constructed corresponding to all target instrumental variables; wherein, the fourth regression model uses a preset intervention variable as the dependent variable.

[0060] Based on the historical viewership data of the target channel, all instrumental variables, and control variables, a two-stage least squares estimation was performed on the fourth regression model to obtain the causal effect estimate and the residuals of the fourth regression model.

[0061] Based on the residuals, ordinary least squares (OLS) regression was performed on all target instrumental variables and all control variables to obtain the goodness of fit (R²).

[0062] Based on the product of the number of data sets (N) of the historical viewership data of the target channel and the goodness of fit, a test statistic (i.e., the Sargan-Hansen J statistic: J=N×R²) is constructed.

[0063] The significance probability value is calculated based on the test statistic and its corresponding chi-square distribution. The J statistic follows a chi-square distribution under the null hypothesis that all instrumental variables are exogenous, and the degrees of freedom of the chi-square distribution are the difference between the number of instrumental variables and the number of endogenous variables (viewership ratings).

[0064] When the significance probability value is greater than the preset probability threshold (e.g., 0.05), the null hypothesis that all instrumental variables are exogenous cannot be rejected, and all target instrumental variables are valid instrumental variables.

[0065] Step S150: Perform causal attribution analysis on the target program's viewership within a preset time window based on the causal effect estimate to obtain the attribution results.

[0066] Counterfactual deductions can be performed using causal effect estimates. Based on these estimates, the impact on viewership ratings when only the pre-defined intervention variable is changed can be determined. For example, if the pre-defined intervention variable is the prime-time broadcast schedule T... peak In the case of program P broadcasting under normal viewership conditions (weekday evenings, T...), peak =0) Adjust to prime-time broadcast (holiday evenings, T peak =1), the counterfactual change in viewership was 1.32 percentage points. This change reflects only the pure causal effect of the programming strategy.

[0067] In some examples, multi-dimensional attribution decomposition can also be performed using causal effect estimates: Based on the estimated causal effect and the corresponding causal effect values ​​of each control variable, a causal attribution model for the target program's viewership is obtained. The control variables are exogenous variables in the structural causal model, and the causal effect values ​​corresponding to each control variable are the coefficients of each control variable in the second-stage regression model of the two-stage regression model. The attribution model is in the form: Causal Contribution = Σ Unit Causal Effect × Variation of the Factor during the Sample Period.

[0068] Based on the viewership causal attribution model and a preset time window, the attribution results are obtained. The attribution results are the causal contribution values ​​or proportions of each dimension within the preset time window; the causal contribution values ​​of each dimension are determined by the product of the causal effect value corresponding to each variable and the magnitude of change of each variable within the preset time window.

[0069] Substitute the actual data from the preset time window (such as during the broadcast of a program) into the attribution model, decompose the portion of the total viewership change contributed by the preset intervention variable (estimated causal effect × magnitude of change of the preset intervention variable) and the portion contributed by each control variable (Σ causal effect value of the control variable × magnitude of change of the control variable), and output the attribution results in the form of percentage or absolute value.

[0070] This disclosure embodiment constructs a structural causal model of the causal relationship between viewership variables by integrating multi-source viewing behavior data and external dimension data. Based on this, it rigorously selects target instrumental variables that meet the relevance conditions, and then uses a two-stage regression model to overcome endogeneity problems such as omitted variable bias and bidirectional causality. It estimates the net causal effect of the preset intervention variables on viewership. Finally, it constructs an attribution model by combining the coefficients of control variables, realizing the explainable attribution of viewership fluctuations within a specific time window, thereby significantly improving the scientific nature of viewership analysis and its decision support capabilities.

[0071] Figure 3 A schematic diagram of the structure of a viewership causal attribution system based on exogenous events provided in an embodiment of this disclosure is shown. Figure 3 As shown, the system 300 includes a data acquisition module 310, a causal modeling module 320, an instrumental variable identification module 330, a causal effect estimation module 340, and an attribution analysis module 350.

[0072] The data acquisition module 310 is used to acquire multi-source viewing behavior data and external dimension data of the target channel; wherein, the multi-source viewing behavior data includes at least a portion of two-way digital television viewing behavior data, interactive network television viewing behavior data and Internet television viewing behavior data.

[0073] The causal modeling module 320 is used to construct a structural causal model based on multi-source viewing behavior data, external dimension data, and preset intervention variables; wherein, the structural causal model is a directed acyclic graph model that represents the causal relationship between viewing behavior variables of the target channel.

[0074] The instrumental variable identification module 330 is used to identify variables in external dimension data based on preset instrumental variable conditions to obtain target instrumental variables. The target instrumental variables are variables derived from exogenous events and whose explanatory power for viewing behavior meets preset relevance conditions. The target instrumental variables include at least some of the variables of adjusted workdays, breaking news events, and changes in channel exclusive broadcasting rights.

[0075] The causal effect estimation module 340 is used to estimate the causal effect of the preset intervention variables on viewership ratings based on the target instrumental variables and the structural causal model, using a two-stage regression model, when the number of target instrumental variables is consistent with the number of preset intervention variables, and to obtain the causal effect estimate.

[0076] The attribution analysis module 350 is used to perform causal attribution analysis on the viewership of the target program within a preset time window based on the causal effect estimate, and obtain the attribution results.

[0077] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. For example... Figure 4 As shown, the electronic device 400 may include a memory 410 and a processor 420. The memory 410 may be used to store computer instructions, and the processor 420 may be used to retrieve computer instructions from the memory 410 to execute all or part of the steps of any of the methods in the foregoing embodiments of this disclosure. It should be noted that the processor 420 may include one or more processors to execute instructions, and the memory 410 may also include one or more memories to store computer instructions. In some embodiments, the processor 420 may be used to control the overall operation of the electronic device 400. For example, the processor 420 may execute instructions to implement all or part of the steps of the methods in any of the foregoing embodiments of this disclosure, thereby enabling one or more of operations such as voice communication, data communication, database operation, display control, component control, and multimedia processing. The aforementioned components may be internal components of the electronic device itself, or external components connected to the electronic device wirelessly or wiredly. For example, the components may include sensors, cameras, headphones, input / output devices, etc. The aforementioned multimedia may include one or more of voice, images, video, and text. In some embodiments, the memory 410 may include one or more memories, and the contents stored in different memories may be the same or different. The memory 410 can be configured to store various types of data to support the operation of the electronic device 400. Examples of such data include instructions for any application or method operating on the electronic device 400, contact data, phonebook data, messages, pictures, videos, etc. The memory 410 can be implemented by any type of temporary or non-temporary storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0078] This disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the methods in the foregoing embodiments of this disclosure.

[0079] The various embodiments in this disclosure are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the method and device embodiments are basically similar to the system embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0080] The foregoing has described specific embodiments of this disclosure. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0081] This disclosure can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of this disclosure.

[0082] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0083] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0084] Computer program instructions used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as C or similar languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing the status information of the computer-readable program instructions to implement various aspects of this disclosure.

[0085] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0086] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0087] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0088] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. It will be known to those skilled in the art that implementation in hardware, implementation in software, and implementation in a combination of software and hardware are equivalent.

[0089] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, and are not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of this disclosure is defined by the appended claims.

Claims

1. A viewership causal attribution method based on exogenous events, characterized in that, The method includes: Acquire multi-source viewing behavior data and external dimension data of the target channel; wherein, the multi-source viewing behavior data includes at least a portion of two-way digital television viewing behavior data, interactive network television viewing behavior data, and Internet television viewing behavior data; Based on the multi-source viewing behavior data, external dimension data, and preset intervention variables, a structural causal model is constructed; wherein, the structural causal model is a directed acyclic graph model that characterizes the causal relationship between the viewing behavior variables of the target channel. Based on preset instrumental variable conditions, the external dimension data is used to identify variables to obtain target instrumental variables; wherein, the target instrumental variables are derived from exogenous events and are variables whose explanatory power for viewing behavior meets preset correlation conditions, and the target instrumental variables include at least some of the following: workday adjustment variables, breaking news event variables, and channel exclusive broadcasting rights change variables; When the number of target instrumental variables is consistent with the number of preset intervention variables, the causal effect of the preset intervention variables on viewership is estimated by a two-stage regression model based on the target instrumental variables and the structural causal model, and the causal effect estimate is obtained. Based on the estimated causal effect, a causal attribution analysis of the target program's viewership within a preset time window is performed to obtain the attribution results.

2. The method according to claim 1, characterized in that, The construction of a structural causal model based on the multi-source viewing behavior data, external dimension data, and preset intervention variables includes: Based on the multi-source viewing behavior data and external dimension data, the basic viewing indicators of the target channel are calculated; wherein, the basic viewing indicators include at least a portion of viewership rating, viewership share, and viewership loyalty. Based on the control variables and ratings in the basic viewership indicators, as well as the preset intervention variables, the causal side orientation logic is determined; wherein, the causal side orientation logic is the causal logical relationship between any two variables among the control variables, preset intervention variables, and ratings. Based on the causal edge orientation logic, a structural causal model is constructed.

3. The method according to claim 1, characterized in that, The process of identifying target instrumental variables from the external dimension data based on preset instrumental variable conditions includes: Based on the historical viewing behavior data of the target channel, at least one candidate instrumental variable is set; For each candidate instrumental variable, a first regression model is constructed based on the control variables in the structural causal model; wherein the first regression model uses a preset intervention variable as the dependent variable. Based on the historical viewing behavior data of the target channel and the first regression model corresponding to each candidate instrumental variable, regression statistics were performed to obtain the F-statistic corresponding to each candidate instrumental variable. Candidate instrumental variables whose F-statistics reach a preset F-statistic threshold are identified as target instrumental variables; wherein, the preset correlation condition includes the preset F-statistic threshold.

4. The method according to claim 1, characterized in that, Based on the target instrumental variable and the structural causal model, a two-stage regression model is used to estimate the causal effect of the preset intervention variable on viewership ratings, yielding an estimated causal effect value, including: Based on the target instrumental variable and the control variables in the structural causal model, a second regression model corresponding to the target instrumental variable is constructed; wherein, the second regression model uses a preset intervention variable as the dependent variable; Based on the historical viewing behavior data of the target channel and the second regression model, the model coefficient set of the second regression model is obtained; wherein, the model coefficient set includes the intercept term and the coefficients of each variable in the second regression model; Regression statistics are performed based on the model coefficient set and the second regression model to obtain the predicted values ​​of the preset intervention variables; Based on the predicted values ​​of the preset intervention variables and the control variables in the structural causal model, a third regression model is constructed; wherein, the third regression model uses viewership rating as the dependent variable. Based on the historical viewing behavior data of the target channel and the third regression model, regression statistics are performed to obtain the causal effect estimate; wherein the causal effect estimate is the coefficient of the preset intervention variable.

5. The method according to claim 4, characterized in that, After obtaining the causal effect estimate, the method further includes: Based on the historical viewership data of the target channel, the target instrumental variable is verified by matching and comparing at least two independent events on the corresponding days to obtain the verification estimate for each independent event. If the validation estimates corresponding to each independent event are consistent with the effect direction of the causal effect estimates, then the target instrumental variable is an effective instrumental variable.

6. The method according to claim 1, characterized in that, When the number of target instrumental variables is greater than the number of pre-defined intervention variables, the method also includes: Based on the target instrumental variables and the control variables in the structural causal model, a fourth regression model is constructed corresponding to all target instrumental variables; wherein, the fourth regression model uses a preset intervention variable as the dependent variable; Based on the historical viewership data of the target channel, all instrumental variables, and control variables, a two-stage least squares method was performed to estimate the causal effect and the residuals of the fourth regression model. Based on the residuals, least squares regression was performed on all target instrumental variables and all control variables to obtain the goodness of fit. A test statistic is constructed by multiplying the number of data sets of historical viewership data for the target channel by the goodness of fit. Calculate the significance probability value based on the test statistic and its corresponding chi-square distribution; When the significance probability value is greater than the preset probability threshold, all target instrumental variables are valid instrumental variables.

7. The method according to claim 1, characterized in that, The step of performing a causal attribution analysis on the target program's viewership within a preset time window based on the estimated causal effect, and obtaining the attribution results, includes: Based on the estimated causal effect and the causal effect values ​​corresponding to each control variable, the viewership causal attribution model of the target program is obtained; wherein, the control variable is the exogenous variable in the structural causal model, and the causal effect value corresponding to each control variable is the coefficient of each control variable in the second stage regression model of the two-stage regression model; Based on the aforementioned viewership causal attribution model and a preset time window, attribution results are obtained; wherein, the attribution results are the causal contribution values ​​or proportions of each dimension within the preset time window; the causal contribution values ​​of each dimension are determined by the product of the causal effect value corresponding to each variable and the variation range of each variable within the preset time window.

8. A viewership causal attribution system based on exogenous events, characterized in that, The system includes: The data acquisition module is used to acquire multi-source viewing behavior data and external dimension data of the target channel; wherein, the multi-source viewing behavior data includes at least a portion of two-way digital television viewing behavior data, interactive network television viewing behavior data, and Internet television viewing behavior data; The causal modeling module is used to construct a structural causal model based on the multi-source viewing behavior data, external dimension data, and preset intervention variables; wherein, the structural causal model is a directed acyclic graph model that characterizes the causal relationship between the viewing behavior variables of the target channel; The instrumental variable identification module is used to identify variables in the external dimension data based on preset instrumental variable conditions to obtain target instrumental variables; wherein, the target instrumental variables are variables derived from exogenous events and whose explanatory power for viewing behavior meets preset correlation conditions, and the target instrumental variables include at least a portion of the following: workday adjustment variables, breaking news event variables, and channel exclusive broadcasting rights change variables; The causal effect estimation module is used to estimate the causal effect of the preset intervention variables on viewership ratings based on the target instrumental variables and the structural causal model, using a two-stage regression model, when the number of target instrumental variables is consistent with the number of preset intervention variables, and to obtain the causal effect estimate. The attribution analysis module is used to perform causal attribution analysis on the target program's viewership within a preset time window based on the estimated causal effect value, and to obtain the attribution results.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory that is communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-7.

10. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1-7.

Images