Movie viewing person-time dynamic prediction method based on review plot revelation degree analysis

Abstract

The application discloses a movie viewing person-time dynamic prediction method based on movie review spoiler degree analysis, relates to the technical field of data processing, and comprises the following steps: in response to a movie viewing person number prediction instruction, acquiring daily movie reviews and historical movie viewing person numbers corresponding to a target movie, quantifying the spoiler degree of the daily movie reviews by using a preset LLM model, obtaining a spoiler index corresponding to the daily movie reviews, predicting the movie viewing person number of the next day by using a pre-constructed movie viewing person number prediction model based on the spoiler index corresponding to the daily movie reviews and the historical movie viewing person numbers, and obtaining a movie viewing person number prediction result corresponding to the target movie. By introducing the LLM technology, the application takes the spoiler as a quantifiable key variable into the movie viewing person number prediction model, considers the influence of the spoiler index on the predicted movie viewing person number when the movie viewing person number prediction model is used to predict the movie viewing person number of the next day, and makes the movie viewing person number prediction result more accurate.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method for dynamically predicting movie attendance based on the analysis of spoiler content in movie reviews. Background Technology

[0002] Current moviegoer prediction technologies primarily rely on historical box office data, scheduling experience, and macro-level release window popularity to construct time-series models. This makes it difficult to effectively integrate real-time textual information such as online film reviews, resulting in low prediction accuracy and delayed dynamic response. With the development of the internet, platforms like Douban aggregate massive amounts of user-generated content, rich in emotional expression, thematic discussions, and narrative evaluations, providing a potential data foundation for accurate prediction. However, traditional natural language processing methods have significant limitations when handling long Chinese film reviews: dictionary-based matching and shallow classifiers struggle to identify context-dependent irony, metaphors, and complex semantic structures, fail to capture implicit word-of-mouth signals, and lack the systematic quantification capability for multi-dimensional semantic features, preventing the full conversion of the film review text's value into predictive effectiveness. Although some studies have attempted to introduce sentiment analysis or topic modeling, they still face problems such as weak cross-contextual understanding, poor cross-language adaptation, and limited logical reasoning ability, making it difficult to support the demand for high-timeliness and high-accuracy predictions from refined cinema scheduling.

[0003] Therefore, there is an urgent need for a new technology that can deeply understand the context of film reviews, accurately extract multi-level semantic features and dynamically integrate them into the prediction framework, so as to break through the bottleneck of traditional methods, improve the accuracy of moviegoer prediction, and achieve the optimal allocation of screening resources.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is related technology. Summary of the Invention

[0005] The main purpose of this application is to provide a method for dynamically predicting movie attendance based on the analysis of spoiler content in movie reviews, aiming to solve the technical problem of how to improve the accuracy of predicting the number of moviegoers.

[0006] To achieve the above objectives, this application proposes a method for dynamically predicting movie attendance based on the degree of spoiler analysis in film reviews. The method includes:

[0007] In response to the audience prediction command, obtain the daily film reviews and historical audience numbers for the target film; The spoiler level of the film review on the day is quantified using a preset LLM model to obtain the spoiler index corresponding to the film review on the day. The preset LLM model includes the DeepSeek-Chat model. Based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, a pre-built audience prediction model is used to predict the number of viewers for the next day, thus obtaining the predicted audience number for the target film. The audience prediction model is based on the film-date unit and is constructed based on the differential effect of the spoiler index of the sample film under different scenarios. The scenarios are determined by the film type, word-of-mouth structure, and user engagement.

[0008] In one embodiment, before the step of quantifying the spoiler level of the film review using a preset LLM model to obtain the spoiler index corresponding to the film review, the method further includes: Obtain research samples, which are the original film reviews corresponding to sample films on the film platform; Based on preset standards, N comments in the original film reviews are manually annotated to obtain a spoiler rating for each of the N comments, where the value of N is preset by the staff. Based on N comments and the spoiler rating for each comment, a Few-Shot example is constructed. Based on the Few-Shot example, the LLM model is given hints and corrections to obtain the preset LLM model.

[0009] In one embodiment, the step of using a preset LLM model to quantify the spoiler level of the film review for the day and obtaining the spoiler index corresponding to the film review for the day further includes: Using a pre-defined LLM model, semantic recognition and automated scoring are performed on the comments in the film reviews of the day to obtain the spoiler index corresponding to the film reviews of the day.

[0010] In one embodiment, before the step of predicting the number of viewers for the next day using a pre-built viewer prediction model based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, and obtaining the predicted number of viewers for the target film, the method further includes: Obtain panel data corresponding to the sample videos and construct multiple different hypotheses; Based on different assumptions, construct different forms of two-way fixed-effects panel models; Based on different forms of two-way fixed-effects panel models and panel data, we conduct an empirical test on the dynamic relationship between the spoiler index, word-of-mouth structure, film type, release time, whether it is a holiday or not, and the number of moviegoers for the sample films, in order to verify whether the hypothesis is correct. Based on correct assumptions, determine the differentiated effects of the spoiler index under different scenarios; Based on the differentiated effects of the spoiler index under different scenarios, a movie attendance prediction model is constructed using movie-date as the unit.

[0011] In one embodiment, the step of obtaining panel data corresponding to the sample video further includes: The original film reviews corresponding to the sample films are cleaned and organized to obtain multiple valid film reviews. The cleaning includes removing null values ​​and outliers. Based on the unique identifier and date information corresponding to each sample film, the daily box office information and sample film corresponding to each valid film review are determined to construct panel data. The panel data is in "film-day" units. The dependent variable in the panel data is the number of daily moviegoers. The core independent variable in the panel data is the spoiler index corresponding to the valid film review. The control variables in the panel data are film rating, number of reviews, screening rate, market ranking, and number of days the film has been released.

[0012] In one embodiment, the step of predicting the number of viewers for the next day using a pre-built viewer prediction model based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, and obtaining the predicted number of viewers for the target film, further includes: Determine the target film's genre, word-of-mouth structure, and user engagement. Based on the film type, word-of-mouth structure, and user engagement of the target film, determine the context corresponding to the target film; Based on the context of the target film, the spoiler index of the film reviews on that day, and the historical number of viewers, a pre-built audience prediction model is used to predict the number of viewers for the next day, thus obtaining the predicted audience number for the target film.

[0013] Furthermore, to achieve the above objectives, this application also proposes a device for dynamically predicting movie attendance based on the degree of spoiler analysis in film reviews. This device includes: The acquisition module is used to acquire the daily film reviews and historical viewership of the target film in response to the movie attendance prediction command. The quantification module is used to quantify the spoiler level of the film review of the day using a preset LLM model to obtain the spoiler index corresponding to the film review of the day. The preset LLM model includes the DeepSeek-Chat model. The prediction module is used to predict the number of viewers for the next day based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, using a pre-built viewer prediction model to obtain the viewer prediction result for the target film. The viewer prediction model is based on the film-date unit and is constructed based on the differential effect of the spoiler index corresponding to the sample film under different scenarios. The scenarios are determined by the film type, word-of-mouth structure, and user participation.

[0014] Furthermore, to achieve the above objectives, this application also proposes a device for dynamically predicting movie attendance based on the analysis of movie review spoilers. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the method for dynamically predicting movie attendance based on the analysis of movie review spoilers as described above.

[0015] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the method for dynamically predicting the number of moviegoers based on the analysis of movie review spoilers as described above.

[0016] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the method for dynamically predicting movie attendance based on movie review spoiler analysis as described above.

[0017] One or more technical solutions proposed in this application have at least the following technical effects: This application proposes a dynamic prediction method for movie attendance based on the analysis of spoiler content in film reviews, belonging to the field of data processing technology. Among related technologies, existing movie attendance prediction techniques mainly rely on historical box office data, screening experience, and macro-level release period popularity to construct time-series models, making it difficult to effectively integrate real-time text information such as online film reviews. This results in low prediction accuracy and delayed dynamic response. With the development of the internet, platforms such as Douban gather massive amounts of user-generated content, containing rich emotional expressions, thematic discussions, and narrative evaluations, providing a potential data foundation for accurate prediction. However, traditional natural language processing methods have significant limitations when processing long Chinese film reviews: dictionary-based matching and shallow classifiers struggle to identify context-dependent irony, metaphors, and complex semantic structures, are insufficient in capturing implicit word-of-mouth signals, and lack the ability to systematically quantify multi-dimensional semantic features. Consequently, the value of film review texts cannot be fully converted into predictive effectiveness. Although some studies have attempted to introduce sentiment analysis or topic modeling, these still face challenges. The limitations of weak cross-contextual understanding, poor cross-language adaptation, and limited logical reasoning ability make it difficult to support the demand for high-timeliness and high-accuracy predictions in refined cinema scheduling. In this application, firstly, in response to the audience prediction command, the daily film reviews and historical audience numbers for the target film are obtained. Then, a pre-set LLM model is used to quantify the spoiler level of the daily film reviews, obtaining the spoiler index corresponding to the daily film reviews. The pre-set LLM model includes the DeepSeek-Chat model. Finally, based on the spoiler index corresponding to the daily film reviews and the historical audience numbers, a pre-constructed audience prediction model is used to predict the audience numbers for the next day, obtaining the audience prediction result for the target film. The audience prediction model is based on a movie-date unit and is constructed based on the differentiated effect of the spoiler index corresponding to the sample film under different scenarios, where the scenarios are determined by film type, word-of-mouth structure, and user engagement.

[0018] Understandably, this application incorporates LLM technology to include spoilers as a quantifiable key variable in the movie attendance prediction model. When using the movie attendance prediction model to predict the number of moviegoers the next day, the impact of the spoiler index on the predicted number of moviegoers is taken into account, making the movie attendance prediction results more accurate. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating an embodiment of the method for dynamically predicting movie attendance based on the analysis of spoiler levels in film reviews, as provided in this application. Figure 2 This is a flowchart illustrating Embodiment 2 of the method for dynamic prediction of movie attendance based on movie review spoiler analysis in this application; Figure 3 This is a schematic diagram of the module structure of the movie attendance dynamic prediction device based on movie review spoiler analysis, as described in this application embodiment. Figure 4 This is a schematic diagram of the hardware operating environment involved in the dynamic prediction method for movie attendance based on movie review spoiler analysis in the embodiments of this application.

[0022] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0023] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0024] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0025] The main solution in this application's embodiments is: In this embodiment, for ease of description, the following description will focus on a dynamic prediction device for movie attendance based on movie review spoiler analysis.

[0026] Due to limitations in existing technologies, current movie attendance prediction techniques primarily rely on historical box office data, scheduling experience, and macro-level release window popularity to construct time-series models. This makes it difficult to effectively integrate real-time textual information such as online film reviews, resulting in low prediction accuracy and delayed dynamic response. With the development of the internet, platforms like Douban have gathered massive amounts of user-generated content, containing rich emotional expressions, thematic discussions, and narrative evaluations, providing a potential data foundation for accurate prediction. However, traditional natural language processing methods have significant limitations when processing long Chinese film reviews: dictionary-based matching and shallow classifiers struggle to identify context-dependent irony, metaphors, and complex semantic structures, are insufficient in capturing implicit word-of-mouth signals, and lack the ability to systematically quantify multi-dimensional semantic features. Consequently, the value of film review texts cannot be fully converted into predictive effectiveness. Although some studies have attempted to introduce sentiment analysis or topic modeling, they still face problems such as weak cross-contextual understanding, poor cross-language adaptation, and limited logical reasoning ability, making it difficult to support the demand for high-timeliness and high-accuracy predictions from cinemas' refined scheduling.

[0027] This application provides a solution in which: first, in response to a moviegoer prediction command, the daily movie reviews and historical moviegoer numbers for the target movie are obtained; then, a preset LLM model is used to quantify the spoiler level of the daily movie reviews to obtain a spoiler index for the daily movie reviews, wherein the preset LLM model includes the DeepSeek-Chat model; finally, based on the spoiler index for the daily movie reviews and historical moviegoer numbers, a pre-built moviegoer prediction model is used to predict the moviegoer numbers for the next day, thus obtaining the moviegoer prediction result for the target movie, wherein the moviegoer prediction model is based on movie-date units, and the moviegoer prediction model is constructed based on the differential effect of the spoiler index for the sample movie under different scenarios, wherein the scenarios are determined by movie type, word-of-mouth structure, and user engagement.

[0028] Understandably, this application incorporates LLM technology to include spoilers as a quantifiable key variable in the movie attendance prediction model. When using the movie attendance prediction model to predict the number of moviegoers the next day, the impact of the spoiler index on the predicted number of moviegoers is taken into account, making the movie attendance prediction results more accurate.

[0029] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device capable of performing the above functions, such as a dynamic prediction device for movie attendance based on movie review spoiler analysis. The following description uses a dynamic prediction device for movie attendance based on movie review spoiler analysis as an example to illustrate this embodiment and the subsequent embodiments.

[0030] Based on this, embodiments of this application provide a method for dynamically predicting movie attendance based on the analysis of spoiler levels in film reviews, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the method for dynamically predicting movie attendance based on the analysis of spoiler levels in movie reviews, as described in this application.

[0031] In this embodiment, the method for dynamically predicting movie attendance based on the analysis of spoiler levels in movie reviews includes steps S10 to S30: Step S10: In response to the moviegoer prediction instruction, obtain the movie reviews and historical viewership of the target movie for the current day; It should be noted that, in this application, the audience prediction instruction refers to the request issued by the cinema scheduling system to "please provide the audience count for film i on the next day".

[0032] It should be noted that "film reviews of the day" refers to all text comments (including long and short reviews) added on the Douban platform between 0:00 and 24:00 on the same day. This is not limited to the Douban platform and can also include other film platforms.

[0033] It should be noted that the historical audience numbers refer to the daily and cumulative audience numbers up to date t, which are calculated by converting "daily box office revenue ÷ average ticket price".

[0034] In this embodiment, after receiving the instruction, the system retrieves the "Daily Film Review" text field and rating field in real time through the Douban open interface. Simultaneously, it retrieves the historical daily box office and historical daily average ticket price for the target film, and calculates the historical daily audience size using the formula "Historical Audience Size = Daily Box Office ÷ Daily Average Ticket Price".

[0035] Step S20: Use a preset LLM model to quantify the spoiler level of the film review of the day to obtain the spoiler index corresponding to the film review of the day. The preset LLM model includes the DeepSeek-Chat model. It should be noted that the default LLM model refers to the DeepSeek-Chat model, which is used for automated scoring after N manually labeled sample few-shot hints.

[0036] It's important to note that the Spoiler Index (SI) is a quantitative score calculated by the model based on whether a film review "reveals key plot points." It's typically a continuous score from 0 to 10. 0 points indicate no spoilers, only mentioning atmosphere, performance, or viewing experience, such as "The last fifteen minutes were breathless," or "The cinematography and soundtrack created a strong sense of tension." 1-3 points indicate mild spoilers, using vague expressions like "A character's fate was unexpected," or "The plot has two twists but still makes sense," without revealing core information. 4-6 points indicate moderate spoilers, involving specific character identity changes, plot twists, or key events in the middle of the film, such as "It's revealed midway through that he's actually an undercover agent," or "The second female lead's death in a car accident changed the plot's direction." 7-9 points indicate severe spoilers, explicitly revealing the ending or major plot twists, such as "The final murderer is the female lead's father," or "The entire movie was actually the protagonist's dream." 10 points indicate a complete spoiler, equivalent to a written summary of the plot, such as "The film begins with X, then Y, and finally Z; the male and female leads get together."

[0037] Understandably, in this embodiment, the large language model, with its superior contextual modeling capabilities, can accurately and efficiently extract the key predictive variable of "spoiler index," laying a solid technical foundation for developing a new generation of high-precision, fine-grained dynamic prediction models for moviegoers.

[0038] Specifically, before the step of using a preset LLM model to quantify the spoiler level of the film review for the day and obtain the spoiler index corresponding to the film review for the day, steps S21 to S23 are also included: Step S21: Obtain research samples, which are the original film reviews corresponding to sample films on the movie platform; It should be noted that in this application, the research samples are all from the Douban Movie platform, with a total of 732,172 original movie review data, covering the time span from July 2016 to July 2019.

[0039] Step S22: Based on preset standards, manually annotate N comments in the original film review to obtain the spoiler score corresponding to each of the N comments, where the value of N is preset by the staff; Step S23: Based on N comments and the spoiler rating for each of the N comments, construct a Few-Shot example. Based on the Few-Shot example, provide hints and corrections to the LLM model to obtain the preset LLM model.

[0040] It should be noted that in this embodiment, the DeepSeek-Chat model is used as the large language model, and human rating results are used as examples to prompt and correct the model. First, N comments (e.g., 30 comments) are extracted from the original film review data, and researchers manually annotate them according to preset standards, with a rating range of 0–10. Subsequently, some of the manually annotated results are used as Few-Shot examples to input into the model to improve its consistency in automated rating of large-scale film review texts. To balance accuracy and processing efficiency, the model makes a trade-off in the selection of the number of Few-Shot examples: although experiments show that increasing the number of examples can improve accuracy, it will significantly prolong the processing time. Therefore, this paper seeks a trade-off between accuracy and speed and determines the final parameter settings. Parallel computing is used in the data processing process, which enables the model to process multiple comments simultaneously, thereby effectively improving the overall computational efficiency. Finally, the model generates a variable for each film review, the Spoiler Index (SI), as the core independent variable in the empirical analysis of this paper.

[0041] Specifically, the step of using a preset LLM model to quantify the spoiler level of the film review for the day and obtaining the spoiler index corresponding to the film review for the day further includes step S24: Step S24: Use a preset LLM model to perform semantic recognition and automated scoring on the comments in the film reviews of the day to obtain the spoiler index corresponding to the film reviews of the day.

[0042] Understandably, this application utilizes an LLM model to perform semantic recognition and automated scoring on long Chinese film reviews, constructing a daily spoiler index and comprehensive reputation index for the film, thus overcoming the shortcomings of traditional NLP in contextual understanding and emotion capture.

[0043] Step S30: Based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, use a pre-built audience prediction model to predict the number of viewers for the next day, and obtain the audience prediction result for the target film. The audience prediction model is based on the film-date unit, and is constructed based on the differential effect of the spoiler index corresponding to the sample film under different scenarios. The scenarios are determined by the film type, word-of-mouth structure, and user participation.

[0044] It should be noted that the movie attendance prediction model is used to predict the number of moviegoers the following day. Spoilers, as content that reveals plot developments, key twists, or endings in advance, have extended their influence from psychological experiments to online consumer behavior and are a key variable affecting the prediction results of the movie attendance prediction model.

[0045] Specifically, the step of predicting the number of viewers for the next day based on the spoiler index corresponding to the film review of the day and the historical number of viewers, and obtaining the predicted number of viewers for the target film, further includes steps S31 to S33: Step S31: Determine the film type, word-of-mouth structure, and user engagement corresponding to the target film; In this application, after determining the Movie ID of the target film, relevant film review data is directly read from the Douban daily film review field to determine the film type, word-of-mouth structure, and user engagement of the target film.

[0046] Step S32: Determine the context corresponding to the target film based on the film type, word-of-mouth structure, and user engagement. In this embodiment, different variables are transformed into "differentiated scenarios that can be recognized by the model" to ensure that subsequent predictions can call the "spoiler index's exclusive coefficient under that scenario", thus providing a data foundation for subsequent predictions based on the spoiler index.

[0047] Step S33: Based on the context of the target film, the spoiler index of the film review on the day and the historical number of viewers, use the pre-built audience prediction model to predict the number of viewers for the next day, and obtain the audience prediction result for the target film.

[0048] In this embodiment, based on the scene corresponding to the target film, the effect of the spoiler index corresponding to the film review on that day in that scene is determined. Based on the effect of the spoiler index in that scene and the spoiler index, the number of viewers for the next day is predicted.

[0049] It is understood that in this embodiment, the "spoiler index" is used as the most critical feature variable for predicting the daily number of moviegoers. Compared with conventional moviegoer data prediction methods, this application combines the spoiler effect to correct the prediction results and obtain more accurate moviegoer prediction results.

[0050] This application proposes a dynamic prediction method for movie attendance based on the analysis of spoiler content in film reviews, belonging to the field of data processing technology. Among related technologies, existing movie attendance prediction techniques mainly rely on historical box office data, screening experience, and macro-level release period popularity to construct time-series models, making it difficult to effectively integrate real-time text information such as online film reviews. This results in low prediction accuracy and delayed dynamic response. With the development of the internet, platforms such as Douban gather massive amounts of user-generated content, containing rich emotional expressions, thematic discussions, and narrative evaluations, providing a potential data foundation for accurate prediction. However, traditional natural language processing methods have significant limitations when processing long Chinese film reviews: dictionary-based matching and shallow classifiers struggle to identify context-dependent irony, metaphors, and complex semantic structures, are insufficient in capturing implicit word-of-mouth signals, and lack the ability to systematically quantify multi-dimensional semantic features. Consequently, the value of film review texts cannot be fully converted into predictive effectiveness. Although some studies have attempted to introduce sentiment analysis or topic modeling, these still face challenges. The limitations of weak cross-contextual understanding, poor cross-language adaptation, and limited logical reasoning ability make it difficult to support the demand for high-timeliness and high-accuracy predictions in refined cinema scheduling. In this application, firstly, in response to the audience prediction command, the daily film reviews and historical audience numbers for the target film are obtained. Then, a pre-set LLM model is used to quantify the spoiler level of the daily film reviews, obtaining the spoiler index corresponding to the daily film reviews. The pre-set LLM model includes the DeepSeek-Chat model. Finally, based on the spoiler index corresponding to the daily film reviews and the historical audience numbers, a pre-constructed audience prediction model is used to predict the audience numbers for the next day, obtaining the audience prediction result for the target film. The audience prediction model is based on a movie-date unit and is constructed based on the differentiated effect of the spoiler index corresponding to the sample film under different scenarios, where the scenarios are determined by film type, word-of-mouth structure, and user engagement.

[0051] Understandably, this application incorporates LLM technology to include spoilers as a quantifiable key variable in the movie attendance prediction model. When using the movie attendance prediction model to predict the number of moviegoers the next day, the impact of the spoiler index on the predicted number of moviegoers is taken into account, making the movie attendance prediction results more accurate.

[0052] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2Before the step of predicting the number of viewers for the next day based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, and obtaining the predicted number of viewers for the target film, steps A10 to A50 are included: Step A10: Obtain panel data corresponding to the sample videos and construct multiple different hypotheses; It should be noted that in this implementation, panel data was constructed using "shadow-day" units, containing 25,516 observations, with fields including dailyboxoffice, avrgspoilerindex, FiveStarRatio, PlotDriven, Holiday, etc.

[0053] It should be noted that the verified hypotheses in this application are as follows: Hypothesis 1: Under the condition of controlling for the previous day's box office performance and holidays, a higher spoiler index (SI) is significantly negatively correlated with the next day's box office performance (and thus the corresponding number of viewers). Hypothesis 2: Film type moderates the impact of the spoiler index on box office (and corresponding number of viewers), with the negative effect of the spoiler index on box office and number of viewers being significantly stronger in plot-driven films. Hypothesis H3.a: In films with a high proportion of high ratings, the positive impact of the spoiler index on box office (and number of viewers) growth is more significant. Hypothesis H3.b: In films with a high proportion of low ratings, the negative effect of the spoiler index on box office (and number of viewers) growth is stronger. Hypothesis H3.c: In films with a high degree of rating polarization, the positive impact of the spoiler index on box office (and number of viewers) growth is more significant. Hypothesis H3.d: In films with a high proportion of neutral ratings, the promoting effect of the spoiler index on box office (and number of viewers) growth is weaker.

[0054] Specifically, the step of obtaining the panel data corresponding to the sample video further includes steps A11 to A12: Step A11: Clean and organize the original film reviews corresponding to the sample films to obtain multiple valid film reviews. The cleaning includes removing null values ​​and outliers. It should be noted that the data used in this application (the original film reviews corresponding to the sample films) still needs to be cleaned and organized. The cleaning process includes removing null values ​​and outliers.

[0055] Step A12: Based on the unique identifier and date information corresponding to each sample film, determine the daily box office information and sample film corresponding to each valid film review to construct panel data. The panel data is in "film-day" units. The dependent variable in the panel data is the number of daily moviegoers. The core independent variable in the panel data is the spoiler index corresponding to the valid film review. The control variables in the panel data are film rating, number of reviews, screening rate, market ranking, and number of days the film has been released.

[0056] It should be noted that in this application, the film review table and box office information table are merged based on the film's unique identifier MovielD and date information to ensure that each observation accurately corresponds to a specific film and its daily box office performance. The processed data is used to construct panel data on a "film-day" basis, with the dependent variable being daily box office revenue (RMB), the core independent variable being the proposed Film Review Spoiler Index (SI), and other variables such as ratings, number of reviews, screening rate, market ranking, and number of days the film has been in theaters being introduced as control variables. Through the above processing, the study has formed a comprehensive dataset that can systematically reflect film market performance and user feedback, laying a solid data foundation for the subsequent construction of the Film Review Spoiler Index and the empirical model testing.

[0057] Step A20: Based on different assumptions, construct different forms of two-way fixed-effects panel models; It should be noted that this application unfolds sequentially around the assumptions.

[0058] The corresponding hypothesis H1 is to construct a benchmark model (1) to test the direct relationship between the spoiler index and the next day's box office change, control for the lagged box office level and holiday effect, and verify whether the spoiler and box office performance are significantly negatively correlated; in addition, to ensure that the model for verifying the main hypothesis is rigorous and convincing, models (2) and (3) are constructed as robustness tests.

[0059]

[0060]

[0061]

[0062] For hypothesis H2: Model (4) is constructed to examine the heterogeneity of the spoiler effect in different types of films. Film type variables are introduced and grouped regression is performed to test whether the negative effect of spoilers on box office is stronger in plot-driven films.

[0063]

[0064] For hypotheses H3.a-H3.d: Model (5)-(8) is constructed. Considering that users’ perception of the rating structure may modulate the spoiler effect, four types of variables are constructed respectively: high rating ratio, low rating ratio, rating polarization degree and medium rating ratio. After interacting with the spoiler index, they are introduced into the model to explore the changes in the direction and intensity of the spoiler effect under different word-of-mouth structures.

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072]

[0073] It should be noted that, through In other words, box office revenue = average ticket price per moviegoer, and the box office revenue is proportionally converted into the predicted number of moviegoers.

[0074] Here, ln(BoxOffice) represents the natural logarithm of the daily box office, used to measure the box office size; ΔBoxOffice represents the change in box office revenue; ln(ΔBoxOffice) represents taking the symmetric log of the box office change to mitigate the impact of extreme values. PlotDriven is a dummy variable that measures whether the film is primarily driven by plot. SI represents the average spoiler index, measuring the degree to which film reviews reveal plot details. Holiday is a dummy variable representing holidays, with holidays = 1 and non-holidays = 0. FiveStarRatio represents the percentage of five-star reviews; FourStarRatio represents the percentage of four-star reviews; and so on. Polarization measures the degree of extreme distribution of reviews.

[0075] Step A30: Based on different forms of two-way fixed effects panel models and panel data, conduct an empirical test on the dynamic relationship between the spoiler index, word-of-mouth structure, film type, release time, whether it is a holiday or not, and the number of movie viewers for the sample films, in order to verify whether the hypothesis is correct. It should be noted that, in order to gain a deeper understanding of the sample characteristics and provide a foundation for subsequent empirical analysis, this application conducted descriptive statistical analysis on key variables. The study used a total of 945 observations in "movie-date" units, including core indicators such as daily box office revenue, box office ranking, movie review spoiler index, number of users rating the films, and number of questions asked in the comments section.

[0076] It should be noted that, after empirical testing, this application found that the sample data exhibited high heterogeneity in dimensions such as film market performance, the degree of spoiler in comments, and user participation behavior. This provides the necessary statistical basis for the subsequent use of a fixed-effects panel model to identify the dynamic impact mechanism of spoilers on box office revenue.

[0077] It should be noted that, in this application, after empirical testing, hypotheses 1, 2, 3.a, 3.b, 3.c, and 3.d were all found to be correct. The specific judgment process is as follows: To test Hypothesis 1, namely, "under the condition of controlling for the previous day's box office performance and holidays, a higher spoiler index (SI) is significantly negatively correlated with the next day's box office performance," this paper constructs three different forms of panel regression models, with the natural logarithm of the daily box office level, the change in box office revenue, and the symmetric logarithm of the change in box office revenue as dependent variables, respectively. Among them, model (1) is the main model for empirical testing, while models (2) and (3) are used to ensure rigor and protect the robustness of the dependent variables. The model adopts a fixed effects setting, controls for the film and the date, and introduces holiday dummy variables and the previous period's box office level as control terms to ensure the robustness of the model identification. The core explanatory variable is the spoiler index lagged by one day, which is used to capture the delayed effect after the release of film review content.

[0078] Based on the above methods, the results of the regression experiments show that the previous day's box office performance exhibits a highly significant positive effect in all three models, indicating a significant dynamic inertia in box office trends. In other words, the empirical results verify the negative impact of spoiler information on box office performance, supporting Hypothesis 1.

[0079] Regarding Hypothesis 2: To further explore the heterogeneity of the spoiler effect, this application introduces a film type variable on the basis of the baseline model and constructs interaction terms to test whether film type will moderate the impact of the spoiler index on box office.

[0080] Based on the above methods, the results of the regression experiment show that the interaction term coefficient for drama films is -0.1105, which is highly significant at the 1% significance level (p=0.0003), verifying that the higher the level of spoilers, the stronger the adverse impact on the box office of drama films. In other words, the empirical results support Hypothesis 2, namely that film genre does indeed moderate the direction and intensity of the impact of spoilers on box office revenue.

[0081] Regarding hypotheses 3.a-3.d: Based on the regression results of the interaction term between different rating distribution variables and the spoiler index, the interaction effect between the overall word-of-mouth rating structure and spoilers was verified. After further refining the rating structure into specific star rating proportions and polarization levels, the model test results showed significant differences in the moderating effect. In summary, under a more detailed rating dimension, the word-of-mouth structure does indeed significantly moderate the path of spoilers' influence on box office performance; that is, the above empirical results support hypotheses 3.a-3.d.

[0082] Step A40: Based on the correct assumptions, determine the differentiated effects of the spoiler index under different scenarios; Understandably, this application, through the verification of hypotheses, concludes that empirical analysis shows the spoiler effect has significant context-dependent and moderating mechanisms. First, in terms of film genre, the negative effect of spoilers is significantly stronger in plot-driven films than in other genres. Second, regarding the moderating effect on film reputation structure, empirical results indicate that when film ratings are concentrated in high, low, or extremely polarized areas, the negative effect of spoilers is significantly weakened, and may even transform into a positive effect, reflecting that spoilers may increase audience interest and motivation in "high trust" or "high controversy" reputation contexts.

[0083] Understandably, from a predictive perspective, the above heterogeneity results indicate that the spoiler index provides the most significant incremental predictive information for plot-driven films, while its marginal predictive value is relatively limited for genres such as comedy and romance. At the same time, there are also significant differences in the uncertainty level of model predictions under different word-of-mouth structures. This provides a practical basis for subsequent genre-based predictive modeling and hierarchical optimization of screening decisions based on film characteristics.

[0084] Step A50: Based on the differentiated effects of the spoiler index under different scenarios, construct a movie attendance prediction model using movie-date as the unit.

[0085] It's important to note that the audience attendance prediction model uses "movie-date" as the smallest analytical unit, aiming to capture the dynamic changes in daily audience attendance for each film during its theatrical run. Its core logic lies in the fact that the impact of spoilers in film reviews on audience decisions is not static, but rather exhibits significant heterogeneity depending on the film's attributes and external environment. Therefore, the model must learn the weight and impact patterns of spoiler indices under different scenarios.

[0086] Based on the above assumptions, in order to accurately characterize the differentiated effect of the spoiler index, a three-dimensional scenario space is constructed in this embodiment, and the specific division and quantification methods are as follows: Method 1 (Based on Film Genre): Different types of films have vastly different levels of sensitivity to spoilers. For example, suspense, thriller, and plot twist films are extremely sensitive to spoilers (high-sensitivity scenario); while special effects blockbusters, comedies, and animated films prioritize audiovisual experience or the density of jokes, and have a higher tolerance for spoilers (low-sensitivity scenario). Film genre embedding vectors, using one-hot encoding or pre-trained film genre embedding vectors, are used as the basic gating condition for scenario segmentation.

[0087] Method Two (Based on the Reputation Structure Dimension): Spoilers have a dual effect—"destroying suspense" and "reducing decision-making risk." When film reviews are polarized, spoilers, as a "mine-clearing" method, may actually have positive value, helping users determine if the film suits them, thus attracting specific audiences. When reviews are generally positive, spoilers mainly have a negative effect. We introduce reputation dispersion indicators (such as the variance and skewness of ratings) and positive review rates to construct a reputation structure feature vector, which is used to adjust the influence coefficient of the spoiler index.

[0088] Method 3 (User Engagement Dimension): In environments with extremely high user engagement (such as social media discussion volume and number of film reviews), spoilers spread extremely quickly. In this case, a surge in the daily spoiler index may trigger a sharp decline (for highly sensitive films) or an abnormal increase (for specific films that generate curiosity through spoilers) in the following day's attendance. Indicators such as film review growth rate and social media mentions are introduced as moderating variables to measure the intensity of the situation.

[0089] Based on the above three methods, the following mechanism is used to achieve the differentiated effect of the spoiler index when constructing specific mathematical prediction models: Interactive feature construction: This involves not only inputting the original "spoiler index," but also constructing higher-order interaction terms between the spoiler index and the aforementioned contextual dimensions (genre, word-of-mouth structure, user engagement). For example: Spoiler index × suspense film identifier, spoiler index × rating variance.

[0090] Gating / Attention Mechanism: If using a deep learning model, introduce a gating linear unit or attention mechanism layer. Enable the model to learn dynamically: Given the current "context vector" (composed of type, reputation, and engagement), assign appropriate attention and positive / negative influence weights to the "spoiler index" feature.

[0091] Based on the above logic and mechanism, a feature set for predicting the number of moviegoers the next day is constructed. In this application, the feature set is mainly divided into three categories: basic time series features, core scenario features, and scenario interaction features.

[0092] It should be noted that the basic time-series characteristics include: the number of moviegoers on the day, the moving average number of moviegoers over the previous 7 days (trend item), the number of days of release (life cycle stage), and whether it is a weekend / holiday (dummy variable).

[0093] It should be noted that the core context features include: the overall spoiler index of the film reviews on the day (calculated through an NLP model), the film type embedding vector, the structural features of the word-of-mouth on the day (positive review rate, rating entropy, variance), and the user engagement on the day (number of new film reviews, number of likes, number of shares).

[0094] It should be noted that the contextual interaction features include: spoiler index × type sensitivity coefficient and spoiler index × word-of-mouth dispersion.

[0095] After constructing the feature set, the model is trained and used for prediction.

[0096] It should also be noted that, in the model selection process, considering the non-linear relationships and interactive complexity of features, gradient boosting trees or deep neural networks based on attention mechanisms were employed. These models can automatically capture the threshold effect and turning points of the spoiler index under different scenarios. Panel data was formed using historical release cycle data of multiple films as samples. Then, the model was trained with the goal of minimizing the mean squared error between the predicted and actual number of viewers the following day.

[0097] It should be noted that during the model training process, the input includes the spoiler index, scene characteristics, and historical number of viewers for the day, and the output is the predicted number of viewers for the next day.

[0098] Understandably, this application uses the differentiated effects of the spoiler index under different scenarios as a basis, incorporating the impact of the spoiler index on the prediction results into the audience prediction model to make the prediction results more accurate. Simultaneously, the constructed audience prediction model can provide cinemas with actionable decision-making basis for scheduling, such as dynamically adjusting the number of screenings, screening times, and auditorium sizes for different films based on the predicted audience numbers for the following day. This effectively reduces empty theaters and overcrowding risks in a market environment with large demand fluctuations, thereby improving the overall allocation efficiency of screening resources.

[0099] This application also provides a device for dynamically predicting movie attendance based on the analysis of spoiler levels in film reviews. Please refer to [link / reference]. Figure 3 The device for dynamically predicting movie attendance based on the analysis of spoiler levels in movie reviews includes: Acquisition module 10 is used to acquire the daily film reviews and historical viewership of the target film in response to the movie attendance prediction instruction. Quantization module 20, the quantization module is used to quantify the spoiler level of the film review of the day using a preset LLM model to obtain the spoiler index corresponding to the film review of the day, wherein the preset LLM model includes the DeepSeek-Chat model; The prediction module 30 is used to predict the number of viewers for the next day based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, using a pre-built viewer prediction model to obtain the viewer prediction result for the target film. The viewer prediction model is based on the film-date unit and is constructed based on the differential effect of the spoiler index corresponding to the sample film under different scenarios. The scenarios are determined by the film type, word-of-mouth structure, and user participation.

[0100] In one embodiment, the dynamic prediction device for movie attendance based on movie review spoiler analysis further includes a large language model correction module, which further includes: The acquisition unit is used to acquire research samples, which are the original film reviews corresponding to sample films in the film platform. The annotation unit is used to manually annotate N comments in the original film review based on preset standards, so as to obtain the spoiler score corresponding to each of the N comments, where the value of N is preset by the staff; The correction unit is used to construct a Few-Shot example based on N comments and the spoiler rating corresponding to each of the N comments. Based on the Few-Shot example, the LLM model is given hints and corrections to obtain the preset LLM model.

[0101] In one embodiment, the quantization module further includes: The quantization unit is used to perform semantic recognition and automated scoring on the comments in the film reviews of the day using a preset LLM model, so as to obtain the spoiler index corresponding to the film reviews of the day.

[0102] In one embodiment, the dynamic prediction device for movie attendance based on movie review spoiler analysis further includes a movie attendance prediction model construction module, which further includes: Hypothesis building units are used to obtain panel data corresponding to sample videos and build multiple different hypotheses; Two-way fixed effects panel model building unit, used to construct different forms of two-way fixed effects panel models based on different assumptions; The empirical testing unit is used to conduct empirical tests on the dynamic relationship between the spoiler index, word-of-mouth structure, film type, release time, holidays, and number of moviegoers of the sample films based on different forms of two-way fixed effects panel models and panel data, in order to verify whether the hypothesis is correct. The first determining unit is used to determine the differentiated effects of the spoiler index under different scenarios based on correct assumptions; The movie attendance prediction model building unit is used to build a movie attendance prediction model based on the differentiated effects of the spoiler index under different scenarios, using movie-date as the unit.

[0103] In one embodiment, the audience attendance prediction model building module further includes: The data preprocessing unit is used to clean and organize the original film reviews corresponding to the sample films to obtain multiple valid film reviews. The cleaning includes removing null values ​​and outliers. Panel data units are constructed to determine the daily box office information and sample films corresponding to each valid film review based on the unique identifier and date information of each sample film. The panel data is in "film-day" units. The dependent variable in the panel data is the number of daily moviegoers. The core independent variable in the panel data is the spoiler index corresponding to the valid film review. The control variables in the panel data are film rating, number of reviews, screening rate, market ranking, and number of days the film has been released.

[0104] In one embodiment, the prediction module further includes: The second determining unit is used to determine the film type, word-of-mouth structure, and user engagement of the target film. The scenario determination unit is used to determine the scenario corresponding to the target film based on the film type, word-of-mouth structure, and user engagement of the target film. The prediction unit is used to predict the number of viewers for the next day based on the context of the target film, the spoiler index of the film review on that day, and the historical number of viewers, using a pre-built audience prediction model, and obtain the audience prediction result for the target film.

[0105] The movie attendance dynamic prediction device based on movie review spoiler analysis provided in this application adopts the movie attendance dynamic prediction method based on movie review spoiler analysis in the above embodiments, and can solve the technical problem of dynamic prediction of movie attendance based on movie review spoiler analysis. Compared with related technologies, the beneficial effects of the movie attendance dynamic prediction device based on movie review spoiler analysis provided in this application are the same as the beneficial effects of the movie attendance dynamic prediction method based on movie review spoiler analysis provided in the above embodiments, and other technical features in the movie attendance dynamic prediction device based on movie review spoiler analysis are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0106] This application provides a device for dynamically predicting movie attendance based on movie review spoiler analysis. The device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the movie attendance dynamic prediction method based on movie review spoiler analysis in the above embodiment 1.

[0107] The following is for reference. Figure 4This document illustrates a structural diagram of a device suitable for implementing the movie attendance dynamic prediction device based on movie review spoiler analysis in the embodiments of this application. The movie attendance dynamic prediction device based on movie review spoiler analysis in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 4 The illustrated device for dynamically predicting movie attendance based on movie review spoiler analysis is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments in this application.

[0108] like Figure 4 As shown, the movie attendance dynamic prediction device based on movie review spoiler analysis may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM) 1004. The RAM 1004 also stores various programs and data required for the operation of the movie attendance dynamic prediction device based on movie review spoiler analysis. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the moviegoer prediction device based on movie review spoiler analysis to exchange data wirelessly or via wired communication with other devices. Although the figure shows a moviegoer prediction device based on movie review spoiler analysis with various systems, it should be understood that implementing or having all the systems shown is not required. More or fewer systems can be implemented alternatively.

[0109] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0110] The movie attendance dynamic prediction device based on movie review spoiler analysis provided in this application adopts the movie attendance dynamic prediction method based on movie review spoiler analysis in the above embodiments, and can solve the technical problem of dynamic prediction of movie attendance based on movie review spoiler analysis. Compared with related technologies, the beneficial effects of the movie attendance dynamic prediction device based on movie review spoiler analysis provided in this application are the same as the beneficial effects of the movie attendance dynamic prediction method based on movie review spoiler analysis provided in the above embodiments, and other technical features in the movie attendance dynamic prediction device based on movie review spoiler analysis are the same as the features disclosed in the previous embodiment method, and will not be repeated here.

[0111] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0112] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0113] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the movie attendance dynamic prediction method based on movie review spoiler analysis in the above embodiments.

[0114] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0115] The aforementioned computer-readable storage medium may be included in a device for dynamically predicting movie attendance based on movie review spoiler analysis; or it may exist independently and not be assembled into a device for dynamically predicting movie attendance based on movie review spoiler analysis.

[0116] The aforementioned computer-readable storage medium carries one or more programs that, when executed by the moviegoer dynamic prediction device based on movie review spoiler analysis, cause the moviegoer dynamic prediction device based on movie review spoiler analysis to: In response to the audience prediction command, obtain the daily film reviews and historical audience numbers for the target film; The spoiler level of the film review on the day is quantified using a preset LLM model to obtain the spoiler index corresponding to the film review on the day. The preset LLM model includes the DeepSeek-Chat model. Based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, a pre-built audience prediction model is used to predict the number of viewers for the next day, thus obtaining the predicted audience number for the target film. The audience prediction model is based on the film-date unit and is constructed based on the differential effect of the spoiler index of the sample film under different scenarios. The scenarios are determined by the film type, word-of-mouth structure, and user engagement.

[0117] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed 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 remote computers, the remote computer can 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 can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0118] 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 this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated 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 operation, or using a combination of dedicated hardware and computer instructions.

[0119] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0120] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described method for dynamically predicting movie attendance based on movie review spoiler analysis, thereby solving the technical problem of dynamically predicting movie attendance based on movie review spoiler analysis. Compared with related technologies, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the method for dynamically predicting movie attendance based on movie review spoiler analysis provided in the above embodiments, and will not be repeated here.

[0121] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method for dynamically predicting movie attendance based on movie review spoiler analysis.

[0122] The computer program product provided in this application can solve the technical problem of dynamically predicting movie attendance based on the analysis of movie review spoilers. Compared with related technologies, the beneficial effects of the computer program product provided in this application are the same as those of the movie attendance dynamic prediction method based on the analysis of movie review spoilers provided in the above embodiments, and will not be repeated here.

[0123] The above description is only a part of the embodiments of this application and does not limit the scope of protection of this application. All equivalent structural transformations made under the technical concept of this application and using the content of this application specification and drawings, or direct / indirect applications in other related technical fields, are included in the scope of protection of this application.

Claims

1. A dynamic prediction method of movie viewership based on the degree of analysis of movie review spoilers, characterized by, The method for dynamically predicting movie attendance based on the analysis of spoiler levels in film reviews includes: In response to the audience prediction command, obtain the daily film reviews and historical audience numbers for the target film; The spoiler level of the film review on the day is quantified using a preset LLM model to obtain the spoiler index corresponding to the film review on the day. The preset LLM model includes the DeepSeek-Chat model. Based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, a pre-built audience prediction model is used to predict the number of viewers for the next day, thus obtaining the audience prediction result for the target film. The audience prediction model is based on the film-date unit and is constructed based on the differential effect of the spoiler index corresponding to the sample film under different scenarios. The scenarios are determined by the film type, word-of-mouth structure, and user participation.

2. The method for dynamically predicting movie attendance based on the analysis of spoiler levels in film reviews as described in claim 1, characterized in that, Before the step of quantifying the spoiler level of the film review using a preset LLM model to obtain the spoiler index corresponding to the film review, the following steps are also included: Obtain research samples, which are the original film reviews corresponding to sample films on the film platform; Based on preset standards, N comments in the original film reviews are manually annotated to obtain a spoiler rating for each of the N comments, where the value of N is preset by the staff. Based on N comments and the spoiler rating for each comment, a Few-Shot example is constructed. Based on the Few-Shot example, the LLM model is given hints and corrections to obtain the preset LLM model.

3. The method for dynamically predicting movie attendance based on the analysis of spoiler levels in film reviews as described in claim 1, characterized in that, The step of quantifying the spoiler level of the film review on that day using a preset LLM model to obtain the spoiler index corresponding to the film review on that day also includes: Using a pre-defined LLM model, semantic recognition and automated scoring are performed on the comments in the film reviews of the day to obtain the spoiler index corresponding to the film reviews of the day. 4.The movie attendance dynamic prediction method based on the movie review spoiler degree analysis of claim 1, wherein, Before the step of predicting the audience numbers for the next day using a pre-built audience prediction model based on the spoiler index corresponding to the film reviews of the day and the historical audience numbers, and obtaining the audience prediction result for the target film, the method further includes: Obtain panel data corresponding to the sample videos and construct multiple different hypotheses; Based on different assumptions, construct different forms of two-way fixed-effects panel models; Based on different forms of two-way fixed-effects panel models and panel data, we conduct an empirical test on the dynamic relationship between the spoiler index, word-of-mouth structure, film type, release time, whether it is a holiday or not, and the number of moviegoers for the sample films, in order to verify whether the hypothesis is correct. Based on correct assumptions, determine the differentiated effects of the spoiler index under different scenarios; Based on the differentiated effects of the spoiler index under different scenarios, a movie attendance prediction model is constructed using movie-date as the unit. 5.The movie attendance dynamic prediction method based on the degree of movie review plot revelation analysis of claim 4, wherein, The step of obtaining panel data corresponding to the sample video further includes: The original film reviews corresponding to the sample films are cleaned and organized to obtain multiple valid film reviews. The cleaning includes removing null values ​​and outliers. Based on the unique identifier and date information corresponding to each sample film, the daily box office information and sample film corresponding to each valid film review are determined to construct panel data. The panel data is in "film-day" units. The dependent variable in the panel data is the number of daily moviegoers. The core independent variable in the panel data is the spoiler index corresponding to the valid film review. The control variables in the panel data are film rating, number of reviews, screening rate, market ranking, and number of days the film has been released. 6.The movie attendance dynamic prediction method based on the movie review spoiler degree analysis of claim 1, wherein, The step of predicting the number of viewers for the next day based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, and obtaining the predicted number of viewers for the target film, further includes: Determine the target film's genre, word-of-mouth structure, and user engagement. Based on the film type, word-of-mouth structure, and user engagement of the target film, determine the context corresponding to the target film; Based on the context of the target film, the spoiler index of the film reviews on that day, and the historical number of viewers, a pre-built audience prediction model is used to predict the number of viewers for the next day, thus obtaining the predicted audience number for the target film.

7. A movie viewing person-time dynamic prediction device based on a movie review plot reveal degree analysis, characterized by, The dynamic prediction device for movie attendance based on the analysis of movie review spoilers includes: The acquisition module is used to acquire the daily film reviews and historical viewership of the target film in response to the movie attendance prediction command. The quantification module is used to quantify the spoiler level of the film review of the day using a preset LLM model to obtain the spoiler index corresponding to the film review of the day. The preset LLM model includes the DeepSeek-Chat model. The prediction module is used to predict the number of viewers for the next day based on the spoiler index corresponding to the film reviews of the day and the historical number of viewers, using a pre-built viewer prediction model to obtain the viewer prediction result for the target film. The viewer prediction model is based on the film-date unit and is constructed based on the differential effect of the spoiler index corresponding to the sample film under different scenarios. The scenarios are determined by the film type, word-of-mouth structure, and user participation.

8. A movie viewing person-time dynamic prediction device based on a movie review plot reveal degree analysis, characterized by, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the method for dynamically predicting movie attendance based on movie review spoiler analysis as described in any one of claims 1 to 6.

9. A storage medium, characterized by The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the method for dynamically predicting the number of moviegoers based on the analysis of movie review spoilers as described in any one of claims 1 to 6.

10. A computer program product, characterised in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the method for dynamically predicting movie attendance based on the analysis of movie review spoilers as described in any one of claims 1 to 6.

Images