Short episode putting optimization method and system based on artificial intelligence, and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a short episode delivery optimization method and system based on artificial intelligence, and a storage medium. The method comprises the following steps: collecting short-drama related data, and performing feature extraction and fusion by using a cross-modal feature fusion network to generate a feature vector; based on the feature vectors, using depth sequential network modeling to analyze user behaviors, obtaining interest features and constructing portraits; putting modeling and prediction are carried out by means of multi-agent reinforcement learning according to the portraits and the feature vectors; recombining the materials by using a delivery strategy, and creating multi-version materials through a generative adversarial network; and analyzing the putting effect by adopting a distributed framework, and generating a resource configuration scheme. According to the method, the cross-modal feature fusion network is adopted to deeply integrate the visual, auditory and user interaction data of a short drama, and the deep sequential network is combined to model the user interest evolution rule, so that the user interest features are accurately captured, and the delivery effect is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and particularly relates to a short drama delivery optimization method and system based on artificial intelligence and a storage medium. BACKGROUND

[0002] With the rapid development of short videos and short drama contents, the demand of users for high-quality contents is continuously increasing. The existing short drama delivery method mainly relies on artificial experience and simple data analysis, and the delivery strategy is usually decided based on the basic portrait features and historical viewing data of users. In the content distribution process, fixed delivery rules and preset target audience groups are usually adopted, and there is a lack of dynamic perception of changes in user interests and real-time optimization of delivery effects. At the same time, the traditional short drama creative optimization method mainly relies on artificial editing and experience summary, and cannot fully utilize multi-modal data features for intelligent content adaptation.

[0003] However, this traditional delivery method has obvious deficiencies. First, a single data dimension cannot accurately capture the deep association between users and contents, resulting in insufficient delivery accuracy. Second, the static user portrait construction method cannot timely reflect the dynamic changes of user interests, affecting the delivery effect. Third, the mechanical delivery strategy optimization lacks deep perception of the market competition environment, making it difficult to achieve optimal allocation of resources. In addition, the optimization process of creative materials lacks consideration of the user emotional resonance dimension, affecting the content reach effect. SUMMARY

[0004] The present application provides a short drama delivery optimization method and system based on artificial intelligence and a storage medium, which solves the technical problem of inaccurate user interest description in the traditional delivery method by introducing multi-modal feature fusion and deep time series modeling technology. The method uses a cross-modal feature fusion network to deeply integrate the visual, auditory and user interaction data of short dramas, combines a deep time series network to model the evolution law of user interests, and realizes accurate capture of user interest features, significantly improving the delivery effect.

[0005] In the first aspect, the present application provides an artificial intelligence-based short drama delivery optimization method, which includes: collecting short drama picture data, audio data and user interaction data through a multimodal data acquisition unit, and using a cross-modal feature fusion network to perform feature extraction and fusion operations on the short drama picture data, audio data and user interaction data to generate a short drama fusion feature vector; based on the short drama fusion feature vector, using a deep time series network to model and analyze the user viewing behavior sequence to obtain user interest migration characteristics, and construct user portrait data based on the user interest migration characteristics; based on the user portrait data and the short drama fusion feature vector, using a multi-agent reinforcement learning network to model the delivery environment and predict the delivery action to obtain short drama delivery strategy data; using the short drama delivery strategy data to segment and reorganize the short drama materials, and generating multiple versions of creative materials through a conditional generative adversarial network to form a short drama creative material library; for the materials in the short drama creative material library, using a distributed computing framework to analyze the delivery effect, obtain delivery effect data, and generate a delivery resource allocation plan based on the delivery effect data.

[0006] In a second aspect, the present application provides an artificial intelligence-based short play delivery optimization system, the artificial intelligence-based short play delivery optimization system comprising: an acquisition module for collecting skit image data, audio data, and user interaction data through a multimodal data acquisition unit, and performing feature extraction and fusion operations on the skit image data, audio data, and user interaction data using a cross-modal feature fusion network to generate a skit fusion feature vector; A modeling module is used to model and analyze the user viewing behavior sequence based on the short drama fusion feature vector using a deep temporal network to obtain user interest migration features and construct user portrait data based on the user interest migration features; A prediction module is used to perform delivery environment modeling and delivery action prediction through a multi-agent reinforcement learning network based on the user portrait data and the short play fusion feature vector to obtain short play delivery strategy data; A reorganization module is used to segment and reorganize the short play materials using the short play delivery strategy data, and generate multiple versions of creative materials through a conditional generative adversarial network to form a short play creative material library; The analysis module is used to analyze the delivery effect of the materials in the short drama creative material library using a distributed computing framework, obtain delivery effect data, and generate a delivery resource configuration plan based on the delivery effect data.

[0007] The third aspect of the present invention provides a computer device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the computer device executes the above-mentioned artificial intelligence-based short drama delivery optimization method.

[0008] A fourth aspect of the present invention provides a computer-readable storage medium, which stores instructions that, when executed on a computer, enable the computer to execute the above-mentioned method for optimizing the delivery of short dramas based on artificial intelligence.

[0009] In the technical solution provided by this application, short play picture data, audio data and user interaction data are collected through a multimodal data acquisition unit, and feature extraction and fusion operations are performed using a cross-modal feature fusion network, thereby achieving all-round capture of short play content features, effectively overcoming the problem of insufficient feature expression caused by a single data dimension in traditional methods; the user viewing behavior sequence is modeled and analyzed through a deep temporal network, the user interest migration characteristics are obtained and user portrait data is constructed, which accurately depicts the dynamic evolution of user interests and improves the timeliness and accuracy of user portraits; the delivery environment modeling and delivery action prediction are performed based on a multi-agent reinforcement learning network, which realizes accurate perception and strategy optimization of complex delivery environments and significantly improves the adaptability of delivery decisions; the short play materials are intelligently segmented and reorganized through a conditional generative adversarial network to generate multiple versions of creative materials, thereby enhancing the differentiated expression ability of the content; the delivery effect analysis is performed using a distributed computing framework to achieve efficient allocation of delivery resources. The innovation of the present invention at the algorithm level is mainly reflected in the following aspects: the cross-modal feature fusion network realizes the deep integration of multimodal data through the attention mechanism, extracting more expressive content features; the deep temporal network introduces long and short-term memory units, effectively modeling the temporal dependencies of user interests; the multi-agent reinforcement learning network adopts a layered architecture design to improve the generalization ability of the delivery strategy; the conditional generative adversarial network combines domain prior knowledge to optimize the generation quality of creative materials. These algorithmic innovations are deeply integrated with specific application scenarios and play an important role in the distribution process of short drama content: they realize the accurate identification of user interests, intelligent optimization of delivery strategies, automatic generation of creative materials, and real-time monitoring of delivery effects, ultimately achieving the goal of improving the delivery effect of short dramas. Through practical application verification, the present invention significantly improves the reach accuracy of short drama content, optimizes the efficiency of delivery resources, enhances the expressiveness of creative materials, and provides effective technical support for the intelligent distribution of short drama content. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 This is a schematic diagram of an embodiment of a method for optimizing short play delivery based on artificial intelligence in an embodiment of the present application; Figure 2 This is a user behavior diagram of a user viewing behavior sequence in an embodiment of the present application; Figure 3 This is a schematic diagram of an embodiment of a short drama delivery optimization system based on artificial intelligence in an embodiment of the present application; Figure 4 It is a schematic block diagram of the structure of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a method, system and storage medium for optimizing the delivery of short dramas based on artificial intelligence. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0013] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiments of the present application, an embodiment of the method for optimizing the delivery of short dramas based on artificial intelligence includes: Step S101: collecting skit image data, audio data, and user interaction data through a multimodal data collection unit, and performing feature extraction and fusion operations on the skit image data, audio data, and user interaction data using a cross-modal feature fusion network to generate a skit fusion feature vector; Step S102: Based on the short drama fusion feature vector, a deep temporal network is used to model and analyze the user viewing behavior sequence to obtain user interest migration features, and user portrait data is constructed based on the user interest migration features; Step S103: Based on the user portrait data and the skit fusion feature vector, a multi-agent reinforcement learning network is used to perform delivery environment modeling and delivery action prediction to obtain skit delivery strategy data; Step S104: segmenting and reorganizing the skit materials using the skit delivery strategy data, and generating multiple versions of creative materials using a conditional generative adversarial network to form a skit creative material library; Step S105: For the materials in the short play creative material library, a distributed computing framework is used to analyze the delivery effect, obtain delivery effect data, and generate a delivery resource allocation plan based on the delivery effect data.

[0014] It is understandable that the execution subject of this application can be an AI-based short play delivery optimization system, or a terminal or a server, and the specific implementation is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0015] Specifically, in the multimodal data collection stage, the multimodal data collection unit conducts all-round data collection on the content of the short play. The multimodal data collection unit includes a video processing module, an audio analysis module, and a user behavior tracking module, which are respectively responsible for collecting the short play picture data, audio data, and user interaction data. For the short play picture data, the system extracts key frames at fixed time intervals (such as every second), performs scene analysis on each key frame, and extracts visual features such as color distribution, texture features, and scene semantics. For audio data, the system divides the audio stream into several audio clips and extracts acoustic features such as tone, rhythm, and emotion. For user interaction data, the system records the user's likes, comments, sharing, and other behaviors and their timestamps. Through the cross-modal feature fusion network, these three types of feature data are time-series aligned and feature fused to generate a unified short play fusion feature vector.

[0016] During the user behavior analysis phase, the system uses a deep temporal network (DTN) to analyze user viewing behavior sequences based on the skit fusion feature vector. The DTN employs an attention mechanism to capture temporal dependencies in user viewing behavior. The system divides a user's viewing history into time windows, encodes the viewing behavior within each time window, and extracts both short-term and long-term interest characteristics. By analyzing user migration patterns between different types of skits, the system generates user interest migration features that characterize the dynamic evolution of user interests. Based on these user interest migration features, the system constructs multi-dimensional user profile data, including basic user attributes, interest preferences, and activity levels. During the delivery strategy generation phase, the system utilizes a multi-agent reinforcement learning network based on the user profile data and the skit fusion feature vector to make delivery decisions. The system models the delivery environment as a state space of multi-agent interactions, with each agent representing a delivery channel. Based on historical delivery data, the system trains the agents to learn optimal delivery strategies, including decisions regarding delivery timing, target demographics, and budget allocation. The system generates specific skit delivery strategy data through a delivery action prediction module.

[0017] During the creative material generation phase, the system intelligently segments and reorganizes the original skit materials based on the skit delivery strategy data. The system identifies key scenes and highlights in the skits, and then reorganizes them based on different creative themes to generate multiple versions of creative materials. Specifically, the system uses a conditional generative adversarial network to input the delivery strategy as a condition and generate creative variants that meet the preferences of the target audience. All generated creative materials are stored in the skit creative material library for subsequent delivery. During the delivery effect analysis phase, the system uses a distributed computing framework to process large-scale delivery data. The system groups the delivery data of creative materials according to the delivery channel and counts the performance indicators of each channel, such as exposure, click-through rate, conversion rate, etc. By calculating the efficiency indicators and delivery cost ratio of each channel, the system evaluates the delivery effect of different channels. Based on the delivery effect data, the system generates the optimal delivery resource allocation plan.

[0018] For example, for a popular short play, the system extracts its visual features (such as scene transitions, character expressions, and screen composition), audio features (such as the mood of the soundtrack and the content of the dialogue), and user interaction features (such as the sentiment of comments and the time distribution of forwarding). Through deep temporal network analysis, it was found that user viewing behavior has obvious time-of-day characteristics, and the preferences of user groups in different time periods vary significantly. Based on these analysis results, the system generates targeted delivery strategies, such as focusing on emotional short play content during prime time in the afternoon and evening. The system also intelligently edits the original short play according to the characteristics of different target groups, generating multiple versions that highlight different highlights. Through continuous effect monitoring and optimization, the system can dynamically adjust resource allocation to improve the overall delivery effect.

[0019] In the embodiment of the present application, short play picture data, audio data and user interaction data are collected by a multi-modal data acquisition unit, feature extraction and fusion operations are performed by a cross-modal feature fusion network, and the short play content features are captured in all directions, effectively overcoming the problem of insufficient feature expression caused by single data dimension in traditional methods; user interest transfer features are obtained and user portrait data are constructed by modeling and analyzing user viewing behavior sequences through a deep time sequence network, accurately depicting the dynamic evolution law of user interest and improving the timeliness and accuracy of the user portrait; the adaptive of the delivery decision is significantly improved by realizing the precise perception and strategy optimization of the complex delivery environment based on the multi-agent reinforcement learning network for delivery environment modeling and delivery action prediction; the content differentiation expression ability is enhanced by generating multiple versions of creative materials through intelligent segmentation and reorganization of short play materials by the conditional generative adversarial network; and the efficient allocation of delivery resources is realized by using a distributed computing framework for delivery effect analysis. The main innovations of the present application at the algorithm level are as follows: the cross-modal feature fusion network realizes the deep integration of multi-modal data through an attention mechanism, and extracts more expressive content features; the deep time sequence network introduces a long short-term memory unit, effectively modeling the time sequence dependence of user interest; the multi-agent reinforcement learning network adopts a hierarchical architecture design, improving the generalization ability of the delivery strategy; and the conditional generative adversarial network combines domain prior knowledge to optimize the generation quality of creative materials. These algorithm innovations are deeply combined with specific application scenarios, and play an important role in the short play content distribution process: accurate identification of user interest, intelligent optimization of delivery strategy, automatic generation of creative materials, and real-time monitoring of delivery effect, ultimately achieving the goal of improving the delivery effect of short plays. Through actual application verification, the present application significantly improves the accuracy of short play content reach, optimizes the use efficiency of delivery resources, enhances the expressiveness of creative materials, and provides effective technical support for the intelligent distribution of short play content.

[0020] In a specific embodiment, the process of step S101 can specifically include the following steps: (1) segmenting and sampling the short play picture data according to a time window to obtain a set of sampling segments, and performing multi-level feature decomposition on each sampling segment in the set of sampling segments to extract color distribution features, texture features and motion trajectory features, and combining the color distribution features, texture features and motion trajectory features to obtain a visual feature matrix; (2) dividing the audio data into multiple overlapping audio frames, performing Fourier transform on each audio frame to obtain a spectrum graph, extracting pitch features, rhythm features and emotion features from the spectrum graph, and combining the pitch features, rhythm features and emotion features in time sequence to obtain an audio feature matrix; (3) Sort user interaction data according to timestamps, construct interaction behavior sequences, calculate click density, dwell time, and interaction frequency in the interaction behavior sequences, and combine click density, dwell time, and interaction frequency into a user behavior feature matrix; (4) Using the cross-modal feature fusion network to calculate the temporal correlation between the visual feature matrix, the audio feature matrix, and the user behavior feature matrix, a temporal correlation matrix is ​​generated. Based on the temporal correlation matrix, the three feature matrices are resampled and interpolated to obtain aligned feature data; (5) By calculating the information entropy of each modal feature in the aligned feature data, the modal fusion weight is determined, and the aligned feature data is weighted summed using the modal fusion weight, and normalized to generate a short drama fusion feature vector.

[0021] Specifically, in the stage of processing the short drama screen data, a time window segmented sampling method is adopted. The window length is set to 3 seconds, and adjacent windows overlap by 1 second. This overlapping design ensures the continuity of feature extraction. For the video sequence within each window, the sampling formula is used: in, is the sampling result at time t, N is the number of sampling points in the window (for example, for a 30fps video, a 3-second window corresponds to 90 sampling points), is the weight coefficient of the i-th sampling point, F(t) is the original video frame sequence, is the sampling interval. Taking an action scene as an example, when a fast action change is detected, the weight coefficient of the corresponding position A larger value is assigned to highlight the features of key action frames. For the sampled frame sequence, color distribution features (histogram of HSV color space), texture features (gray-level co-occurrence matrix statistics), and motion trajectory features (optical flow field vectors) are extracted and combined into a visual feature matrix.

[0022] In the audio processing phase, the system first divides the audio data into frames according to a window length of 25ms and a frame shift of 10ms. Short-time Fourier transform is applied to each audio frame: here represents the time-frequency spectrum, is the Hamming window function, x(t) is the audio signal, is the angular frequency. Using this formula, the system converts the time-domain signal into a time-frequency representation, facilitating the extraction of audio features. For example, when processing an emotional dialogue scene, the system analyzes the spectrogram to extract pitch variations (fundamental frequency trajectory), rhythmic characteristics (energy envelope), and emotional features (harmonic structure). These features are organized into an audio feature matrix.

[0023] For the processing of user interaction data, various interaction behaviors are sorted and feature extracted based on timestamps. In each analysis window (usually 10 seconds), the cumulative value of interaction features is calculated: wherein is the comprehensive interaction index at time t, is the click density (times / second), is the average dwell time (seconds), and F(t) is the interaction frequency (times / minute), , , are the weight coefficients of each feature, respectively. For example, for a dramatic climax segment, when the click density reaches 10 times / second, the average dwell time exceeds 30 seconds, and the interaction frequency reaches 60 times / minute, the system will recognize it as a key moment of high user participation.

[0024] In the cross-modal feature fusion stage, the system establishes a feature alignment mechanism by calculating the temporal correlation between the three feature matrices. All features are resampled to a unified time scale (such as 100 ms as the basic unit), and then linear interpolation is used to fill in missing values. For a typical short drama scene, such as an emotional explosion point, visual features may exhibit dramatic changes in expressions and actions, audio features may exhibit significant changes in volume and tone, and user interactions may exhibit dense comments and likes. The system ensures the accurate correspondence of these features in the time dimension through feature alignment. By calculating the information entropy of each modal feature, the weights of different features in the fusion process are determined. For different types of scenes, these weights are dynamically adjusted. For example, in action scenes dominated by visual effects, the weight of visual features will be correspondingly increased; in emotional scenes dominated by dialogue, audio features will be given greater weight. This dynamic weight adjustment mechanism ensures the flexibility and accuracy of feature fusion.

[0025] In a specific embodiment, the process of performing step S102 can specifically include the following steps: (1) constructing a viewing feature sequence according to the short drama fusion feature vector, and dividing the viewing feature sequence into time windows to obtain viewing window data; (2) sequentially encoding the user viewing behavior sequence in each time window in the viewing window data to generate a behavior encoding sequence; (3) using a deep time series network to analyze the time series correlation of the behavior encoding sequence, extracting viewing duration, switching frequency, and interaction intensity, and generating a behavior feature matrix; (4) calculating the user interest change rate based on the behavior feature matrix, and performing time series accumulation on the user interest change rate to obtain user interest migration features; (5) Clustering user interest migration features according to interest topics and constructing an interest category table; Generate user portrait data based on the weight distribution of each category in the interest category table.

[0026] Specifically, if Figure 2 The figure below is a user behavior diagram of the user viewing behavior sequence in the embodiment of the present application. When constructing the viewing feature sequence based on the short drama fusion feature vector, a time window division method is adopted. The length of each window is set to 5 minutes, and adjacent windows overlap by 1 minute. For each viewing sequence, the following feature extraction formula is used: in represents the viewing behavior characteristics at time t, is the number of sampling points in the window, is the weight coefficient of the kth sampling point, Q(t) is the original viewing behavior sequence, For example, when processing a continuous plot segment, if the user repeatedly watches it at a certain time point, the weight coefficient of the corresponding position A larger value will be assigned to highlight the behavioral characteristics of that time point.

[0027] For the generation of user behavior coding sequence, the sequence coding function is used: here is the behavior encoding value at position r, Play status (1 means play, 0 means pause), is the viewing time (unit: seconds), is the number of jumps, 、 、 are the weight coefficients of each behavioral feature. Taking a typical scenario as an example, when a user is watching a plot, if they frequently pause or re-watch a certain segment, these features will be encoded as higher behavioral feature values.

[0028] When calculating the user interest change rate, the interest dynamic change function is used: in is the interest change rate at time t, is the change in content type (number of conversions between different types), To change the viewing speed (the number of times the speed is adjusted), is the intensity of interaction (the frequency of likes, comments, etc.), 、 、 is the corresponding weight coefficient. In practical applications, for example, when a user switches from a lighthearted plot to a suspenseful plot, the system will detect a significant change in content type and, based on the user's viewing speed adjustments and interactive behavior, calculate the trend of interest changes.

[0029] Based on the calculated interest change rate, the system accumulates the time series to generate user interest migration features. This records the evolution of user interests over time. For example, if a user initially prefers lighthearted and humorous content but gradually develops an interest in suspenseful plots, this interest migration will be accurately captured and quantified by the system.

[0030] During the interest clustering and grouping stage, the system clusters user interest migration features according to content themes. For example, users' interests are divided into different categories such as suspense and mystery, urban emotions, and lighthearted comedy. For each category, the system calculates its weight value to reflect the user's preference for that type of content. For example, if the user's viewing time on suspense short dramas accounts for 60% and the interactive behavior accounts for 70%, the suspense category will be given a higher weight. Based on the weight distribution of interest categories, the system generates complete user portrait data. This portrait contains information on multiple dimensions such as the user's interest preferences, viewing habits, and interaction patterns. In this way, the system can accurately portray the user's interest characteristics.

[0031] In a specific embodiment, the process of executing step S103 may specifically include the following steps: (1) User portrait data is stratified according to interest dimensions to construct a user stratification matrix, and the short drama fusion feature vector is mapped into the user stratification matrix to generate interest matching distribution; (2) Expand the interest matching degree distribution in the time and space dimensions, establish a delivery status space table, and extract the delivery time period, delivery area and delivery channel from the delivery status space table to form delivery environment characteristics; (3) Input the delivery environment characteristics into the multi-agent reinforcement learning network to model the competitive relationship between each delivery channel and obtain the channel competition matrix; (4) Calculate the reward value of the delivery action based on the channel competition matrix, use the reward value as the optimization target, and construct the action value table; (5) Select the optimal delivery action combination from the action value table to generate a delivery decision sequence; (6) Match the delivery decision sequence with the delivery constraints to obtain the short drama delivery strategy data.

[0032] Specifically, in the user stratification processing stage, user portrait data is stratified according to different interest dimensions. The user interest matching degree is calculated using the following formula: in Indicates interest level The matching value of is the number of interest feature dimensions, is the weight coefficient of the p-th dimension, G(x) is the interest feature function, For example, for a young user group, if their viewing history shows that they spend 70% of their viewing time on suspense short dramas and interact with them 5 times per hour, the weight coefficient of the suspense category will be increased accordingly, which will be reflected in the interest matching distribution.

[0033] For modeling of the delivery environment characteristics, the system uses the spatiotemporal state transfer function: here Spacetime point The environmental characteristic value of is the time characteristic (such as peak period, trough period), is a spatial feature (such as regional attributes), is the spatiotemporal coupling feature, 、 、 is the corresponding weight coefficient. In actual applications, for example, if the short drama viewing peak is between 8 and 10 pm in a certain city, the time feature weight of this period will be increased. At the same time, combined with the user activity data of the area, the delivery environment characteristics are formed.

[0034] The modeling of channel competition relationships uses the reward calculation formula: Where R(a,b) is the competitive relationship value between channels a and b, V(a,b) is the traffic competition degree, C(a,b) is the cost ratio, and F(a,b) is the conversion efficiency ratio. 、 、 is the corresponding weight coefficient. For example, between two mainstream short video platforms, if the unit delivery cost of platform A is 0.8 times that of platform B, but the conversion rate is 1.2 times that of platform B, these factors will be comprehensively calculated through the formula to obtain the competitive relationship value between channels. When generating the delivery action value table, the system comprehensively considers the competitive relationship and delivery effect of each channel. For example, when it is found that the delivery effect of a certain channel in a specific period of time significantly exceeds that of other channels, the system will increase the delivery weight of the channel in this period. In specific operations, the system will record the historical performance of each delivery action, including indicators such as click-through rate, conversion rate, and return on investment, and update the action value table accordingly.

[0035] In the generation process of the delivery decision sequence, the system selects the optimal delivery action combination from the action value table. Multiple factors are considered, such as the active period of the target audience, the traffic distribution of the channel, the delivery situation of competitors, etc. For example, for young user groups, the system will increase the delivery intensity during the evening rest period and preferentially select short video platforms with high user activity. The system matches the generated delivery decision sequence with actual delivery constraints. These constraints include budget limitations, delivery time period limitations, material specification requirements, etc. Through this matching process, the final short drama delivery strategy data not only meets the optimization goal but also meets the actual execution conditions. For example, when the daily budget is 10,000 yuan, the system will reasonably allocate the budget according to the expected effect of each time period to ensure that key time periods and channels receive sufficient resource input.

[0036] In a specific embodiment, the process of performing step S104 can specifically include the following steps: (1) Convert the short drama delivery strategy data into a scene description matrix, and perform key frame extraction on the short drama materials based on the scene description matrix to obtain a scene segment set; (2) Perform semantic analysis on the scene segment set to generate a scene semantic vector, and establish a segment association graph based on the scene semantic vector; (3) According to the scene connection relationship in the segment association graph, reorganize and sort the scene segment set to generate a reorganized scene sequence; (4) Input the reorganized scene sequence into the conditional generation adversarial network to generate multiple initial creative materials with different themes and styles; (5) Perform creative scoring on the initial creative materials, extract creative theme tags and style feature tags, and construct a creative material feature table; (6) Classify and organize the creative materials in the creative material feature table according to the theme tags and style feature tags to generate a short drama creative material library.

[0037] Specifically, it is converted into a scene description matrix. This conversion process takes into account information such as delivery time, target audience, and expected effect. For example, a late-night delivery strategy targeting young people will be converted into scene description information containing time characteristics, population characteristics, and effect targets. Based on this scene description matrix, key frames are extracted from the original sketch material, and the extraction rules include scene switching points, expression change points, and action climax points. Each key frame has a timestamp and a scene type label, forming a scene segment set. For the extracted scene segment set, the next step is semantic analysis. The semantic analysis process includes identifying the main elements in the scene (such as characters, actions, and environment), emotional tone, and plot development. Through these analysis results, a scene semantic vector is generated, which contains the core semantic information of the scene. Based on these semantic vectors, an association graph between segments is established, which reflects the logical relationship between different scene segments. For example, in an emotional short drama, scenes with similar emotional tones will have a strong association, and emotional turning points will be marked as key nodes.

[0038] According to the established segment association graph, the scene segments are reorganized and sorted. The reorganization process needs to consider the continuity between scenes, the emotional progression relationship, and the plot development rule. By analyzing the scene connection relationship in the association graph, the best scene combination order is determined, and a reorganized scene sequence is generated. This reorganization process is not a simple linear arrangement, but an intelligent scene arrangement according to creative needs. The reorganized scene sequence is input into the conditional generative adversarial network to generate multiple initial creative materials with different theme styles. During the generation process, the style characteristics of the generated materials are controlled by adjusting the conditional parameters. For example, for different target audience groups, generate creative versions with different styles such as vitality, warmth, and suspense. Each generated creative material retains the core content of the original scene while having unique style characteristics.

[0039] The next key step is to score the generated initial creative materials. The scoring dimensions include visual appeal, emotional appeal, and information transmission effect. At the same time, the theme tags (such as "youth", "inspiration", "warmth", etc.) and style feature tags (such as "bright", "warm", "tense", etc.) of each creative material are extracted. These scoring and tagging information are organized into a creative material feature table, which records the detailed feature description of each creative material. The materials in the creative material feature table are classified and organized. The classification standard is mainly based on theme tags and style feature tags, and materials with similar features are classified into the same category. This classification method facilitates the quick retrieval of creative materials that meet the needs in actual delivery. The sorted materials are stored in the short drama creative material library, forming a systematic creative resource library.

[0040] Take a typical short drama optimization case as an example: the original material of a short drama of the emotional type is 3 minutes long, and 12 key scene segments are extracted through scene analysis. After semantic analysis, core plot nodes such as “first meeting”, “complication”, and “reconciliation” are identified. According to the semantic correlation between segments, a complete scene correlation network is constructed. Based on this network, multiple creative versions are generated for different delivery scenarios: the version for young groups highlights the elements of youth and vitality, focusing on the interactive scenes of the characters; the version for mature groups focuses on emotional resonance, highlighting the inner monologue of the characters. After scoring and label extraction, these different versions of creative material are organized into a creative material library. When delivery is needed in a specific scenario, the most suitable creative version is retrieved from the material library to achieve precise delivery.

[0041] In a specific embodiment, the process of performing step S105 can specifically include the following steps: (1) Group the materials in the short drama creative material library according to the delivery channels, construct a channel delivery record table, and generate an effect statistics matrix by counting the exposure, click volume, and conversion volume from the channel delivery record table; (2) Calculate the click conversion ratio and delivery cost ratio of each delivery channel based on the effect statistics matrix, and combine the click conversion ratio and delivery cost ratio to form a channel efficiency indicator; (3) Time series expansion of the channel efficiency indicator, construction of the efficiency trend chart, extraction of the delivery time period weight from the efficiency trend chart; (4) Cross operation of the delivery time period weight and the channel efficiency indicator to obtain a space-time delivery scoring table; (5) Priority ranking of each delivery channel based on the space-time delivery scoring table to generate delivery effect data; (6) Space-time dimension allocation of delivery resources based on the delivery effect data to obtain a delivery resource allocation scheme.

[0042] Specifically, the 24-hour delivery data of six typical delivery channels (channels AA, BB, CC, DD, EE, and FF) is grouped and organized. The key indicators of channel AA in the evening peak period (20:00-21:00) are: total exposure 2.2 million times, effective click volume 13.2 million times, deep conversion volume 2.64 million times, benchmark cost 9000 yuan, quality interaction volume 4.62 million times, and benchmark interaction volume 13.2 million times. The data of channel BB at the same time shows: total exposure 180 million times, effective click volume 9 million times, deep conversion volume 1.8 million times, benchmark cost 7000 yuan, quality interaction volume 2.7 million times, and benchmark interaction volume 9 million times. The data of channel CC is: total exposure 150 million times, effective click volume 6 million times, deep conversion volume 1.2 million times, benchmark cost 5500 yuan, quality interaction volume 1.8 million times, and benchmark interaction volume 6 million times.

[0043] Channel effectiveness evaluation uses a multi-level calculation model: in is the performance index value of channel v, is the effective click volume (times / hour), is the total exposure (times / hour), is the depth of conversion (times / hour), is the total number of clicks (times / hour), is the conversion income (yuan / hour), is the benchmark cost (yuan / hour), is the high-quality interaction volume (times / hour), is the benchmark interaction volume (times / hour), and is the balance coefficient (the values ​​are 0.4 and 0.6).

[0044] Taking channel AA as an example, substitute the specific data: , indicating an effective click rate of 6%; Reflects the deep conversion effect; represents the logarithmic measure of benefit cost; Reflects the impact of the proportion of high-quality interactions.

[0045] Time series analysis introduces a progressive evaluation function: in is the efficiency value of channel v in period i, is the peak flow rate (times / minute), is the average flow rate (times / minute), is the actual income (yuan / hour), is the expected income (yuan / hour), is the cumulative number of interactions (times), is the target interaction amount (times), 、 、 are dynamic weight coefficients (0.3, 0.4, and 0.3 respectively).

[0046] Specifically for channel AA, the peak flow rate is 3600 times / minute, and the average flow rate is 1285 times / minute. ; Actual income is 22,500 yuan / hour, expected income is 17,300 yuan / hour, calculated ; The cumulative interaction volume is 125,000 times, the target interaction volume is 100,000 times, and the result is .

[0047] Cross-dimensional feature fusion model: in Spacetime point The comprehensive rating of is the real-time conversion volume (times / minute), is the benchmark conversion rate (times / minute), The user's stay time (seconds), is the average dwell time (seconds), To share the number of transmissions (times), is the number of impressions (times), 、 、 is the feature weight (values ​​are 0.35, 0.35, 0.3).

[0048] The specific performance of channel AA is: real-time conversion volume 29 times / minute, benchmark conversion volume 20 times / minute, ; The user stays for 96 seconds, with an average stay of 60 seconds. ; 110,000 shares and 2.2 million impressions, .

[0049] Based on the above comprehensive scoring results, resource allocation was optimized for the six channels. The specific allocation plan for the 200,000 yuan daily budget was as follows: Channel AA received 60,000 yuan, primarily for advertising during the evening peak period (20:00-22:00), as it exhibits the highest performance indicators and conversion rates during this period; Channel BB received 45,000 yuan, focusing on evening and weekend periods, leveraging its user activity during these times; Channel CC received 35,000 yuan, targeted at its strong performance periods; and the remaining 60,000 yuan was allocated to channels DD, EE, and FF according to their respective performance ratios.

[0050] In a specific embodiment, the process of executing step S106 may specifically include the following steps: (1) Normalize the scoring data in the time-space delivery scoring table, construct a standardized scoring matrix, and extract the time period performance value of each channel from the standardized scoring matrix; (2) Aggregate the performance values ​​of each period according to the channel dimension to obtain the channel performance index, and construct a channel ranking table based on the channel performance index; (3) Calculate the historical return rate for each channel in the channel ranking table, generate a return distribution curve, and extract the return advantage coefficient from the return distribution curve; (4) Weighted combination of the placement advantage coefficient and the channel performance index to form a comprehensive channel score; (5) Sort the channel comprehensive scores in descending order to generate a channel priority list; (6) Combine the channel priority list with the time period performance value to generate delivery effect data.

[0051] Specifically, we normalized the spatiotemporal scoring data for five delivery channels (channels KK, LL, MM, NN, and PP). The raw scoring data for channel KK during peak hours (20:00-21:00) was: 92 points, 2800 actual conversions / hour, 2000 target conversions / hour, 4200 effective interactions / hour, and 3000 baseline interactions / hour. These data were normalized and constructed into a standardized scoring matrix.

[0052] Normalized calculation uses a comprehensive evaluation model: Where H(x) is the standardized score value of channel x, I(x) is the original score value, and are the maximum and minimum score values, respectively. is the actual conversion volume (times / hour), is the target conversion volume (times / hour), is the effective interaction volume (times / hour), is the benchmark interaction volume (times / hour), and is the weight coefficient (with values ​​of 0.6 and 0.4). Taking channel KK as an example, substitute the actual data: original score 92 points, actual conversion volume 2800 times / hour, target conversion volume 2000 times / hour, effective interaction volume 4200 times / hour, and baseline interaction volume 3000 times / hour. After the standardization process is completed, the performance value of each time period is extracted from the scoring matrix. The performance data of channel KK in different time periods shows: the morning peak (8:00-10:00) has a standardized score of 0.75, the noon peak (12:00-14:00) has a score of 0.82, and the evening peak (20:00-22:00) has a score of 0.95. These time period performance values ​​are further aggregated according to the channel dimension to obtain the channel's overall performance index.

[0053] The calculation of channel performance index introduces a multi-level evaluation model: in is the performance index of channel x in period t, is the peak traffic (times / minute), P(t,x) is the average traffic (times / minute), Q(t,x) is the user's stay time (seconds), R(t,x) is the benchmark stay time (seconds), S(t,x) is the actual revenue (yuan), T(t,x) is the target revenue (yuan), 、 、 are weight coefficients (take 0.3, 0.4, and 0.3 respectively). Specific data of channel KK during peak hours: peak traffic 4,000 times / minute, average traffic 1,500 times / minute; user stay time 120 seconds, benchmark stay time 80 seconds; actual income 25,000 yuan, target income 20,000 yuan. Based on the calculated performance index, a channel ranking table is constructed. The ranking data shows that: channel KK has the highest comprehensive performance index, reaching 0.92; channel LL is second, at 0.85; channel MM is 0.78. At the same time, the historical delivery data of each channel in the past 30 days is analyzed, the delivery rate of return is calculated, and a rate of return distribution curve is generated. The historical return data of channel KK shows: the average daily rate of return is 185%, and the fluctuation range is controlled within 15%; the rate of return during peak hours can reach up to 230%.

[0054] The comprehensive scoring system uses a dynamic weighting model: in is the comprehensive score of channel x in period t, is the historical performance index, is the current income (yuan), is the historical average return (yuan), is the user activity (times / hour), V(t,x) is the baseline activity (times / hour), 、 、 is a dynamic weight coefficient (summing to 1). Data for channel KK shows: a historical performance index of 0.92, current revenue of 28,000 yuan, historical average revenue of 22,000 yuan, user activity of 8,500 times / hour, and baseline activity of 6,000 times / hour. After extracting the delivery advantage coefficient from the revenue distribution curve, it is weighted and combined with the channel performance index to form the final channel comprehensive score. Channel KK's advantage coefficient is 1.28, and combined with the performance index of 0.92, the final comprehensive score reaches 94 points. The comprehensive scores of each channel are sorted in descending order to generate a priority list: channel KK > LL > MM > NN > PP.

[0055] Combining the priority list with the time-of-day performance values ​​yields comprehensive campaign effectiveness data. The specific campaign plan is as follows: Channel KK receives a 54,000 RMB budget, focusing on the evening peak period, with an expected ROI of 220%; Channel LL receives 45,000 RMB, focusing on the midday and evening hours, with an expected ROI of 180%; the remaining funds are allocated to other channels based on their ratings. A week of operational data shows that after implementing this plan, overall campaign effectiveness increased by 35%, with ROI exceeding expectations for each channel.

[0056] The above describes the short play delivery optimization method based on artificial intelligence in the embodiment of the present application. The following describes the short play delivery optimization system based on artificial intelligence in the embodiment of the present application. Figure 3 In the embodiments of the present application, an embodiment of the short drama delivery optimization system based on artificial intelligence includes: an acquisition module for collecting skit image data, audio data, and user interaction data through a multimodal data acquisition unit, and performing feature extraction and fusion operations on the skit image data, audio data, and user interaction data using a cross-modal feature fusion network to generate a skit fusion feature vector; A modeling module is used to model and analyze the user viewing behavior sequence based on the short drama fusion feature vector using a deep temporal network to obtain user interest migration features and construct user portrait data based on the user interest migration features; A prediction module is used to perform delivery environment modeling and delivery action prediction through a multi-agent reinforcement learning network based on the user portrait data and the short play fusion feature vector to obtain short play delivery strategy data; A reorganization module is used to segment and reorganize the short play materials using the short play delivery strategy data, and generate multiple versions of creative materials through a conditional generative adversarial network to form a short play creative material library; The analysis module is used to analyze the delivery effect of the materials in the short drama creative material library using a distributed computing framework, obtain delivery effect data, and generate a delivery resource configuration plan based on the delivery effect data.

[0057] Through the cooperation of the above-mentioned components, the short drama picture data, audio data and user interaction data are collected through the multi-modal data acquisition unit, the cross-modal feature fusion network is used for feature extraction and fusion operation, the short drama content features are captured in all directions, and the problem of insufficient feature expression caused by single data dimension in the traditional method is effectively overcome; the user interest transfer features are obtained and the user portrait data are constructed by modeling and analyzing the user viewing behavior sequence through the deep time sequence network, the dynamic evolution law of the user interest is accurately described, and the timeliness and accuracy of the user portrait are improved; the multi-agent reinforcement learning network is used for modeling the delivery environment and predicting the delivery action, the complex delivery environment is accurately perceived and the strategy is optimized, and the adaptability of the delivery decision is significantly improved; the short drama materials are intelligently segmented and reorganized through the conditional generative adversarial network, and multiple versions of creative materials are generated, and the differentiated expression ability of the content is enhanced; the distributed computing framework is used for delivery effect analysis, and efficient allocation of delivery resources is realized. The main innovations of the present application at the algorithm level are as follows: the cross-modal feature fusion network realizes the deep integration of multi-modal data through the attention mechanism, and extracts more expressive content features; the deep time sequence network introduces the long short-term memory unit, and effectively models the time sequence dependence relationship of the user interest; the multi-agent reinforcement learning network adopts a hierarchical architecture design, which improves the generalization ability of the delivery strategy; the conditional generative adversarial network combines with the prior knowledge in the field, and optimizes the generation quality of the creative materials. These algorithm innovations and specific application scenarios are deeply combined, and play an important role in the short drama content distribution process: accurate identification of user interest, intelligent optimization of delivery strategy, automatic generation of creative materials, and real-time monitoring of delivery effect, ultimately achieving the goal of improving the delivery effect of short dramas. Through actual application verification, the present application significantly improves the accuracy of short drama content reach, optimizes the use efficiency of delivery resources, enhances the expressiveness of creative materials, and provides effective technical support for the intelligent distribution of short drama content.

[0058] Referring to Figure 4 , the computer device in the embodiment of the present application can be a server, and the internal structure thereof can be as shown in Figure 4 . The computer device comprises a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in the embodiment. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement the above-mentioned method.

[0059] Those skilled in the art will understand that Figure 4 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0060] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which implements the above-described method when executed by a processor. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0061] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing the relevant hardware using a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the above-described method embodiments. Any reference to memory, storage, database, or other media provided herein and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct RAMbus dynamic RAM (DRDRAM), and RAMbus dynamic RAM.

[0062] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0063] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0064] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A short drama delivery optimization method based on artificial intelligence, characterized in that: The AI-based short drama delivery optimization method includes: The skit image data, audio data and user interaction data are collected by a multimodal data collection unit, and the skit image data, audio data and user interaction data are subjected to feature extraction and fusion operation by a cross-modal feature fusion network to generate a skit fusion feature vector; Based on the short drama fusion feature vector, a deep temporal network is used to model and analyze the user viewing behavior sequence to obtain user interest migration features, and user portrait data is constructed based on the user interest migration features; Based on the user portrait data and the short play fusion feature vector, a multi-agent reinforcement learning network is used to perform delivery environment modeling and delivery action prediction to obtain short play delivery strategy data; Using the short play delivery strategy data, the short play materials are segmented and reorganized, and multiple versions of creative materials are generated through a conditional generative adversarial network to form a short play creative material library; For the materials in the short play creative material library, a distributed computing framework is used to analyze the delivery effect, obtain delivery effect data, and generate a delivery resource configuration plan based on the delivery effect data.

2. The method for optimizing short play delivery based on artificial intelligence according to claim 1, characterized in that: The method includes collecting skit screen data, audio data, and user interaction data through a multimodal data collection unit, performing feature extraction and fusion operations on the skit screen data, audio data, and user interaction data using a cross-modal feature fusion network, and generating a skit fusion feature vector, including: Segmented sampling of the skit screen data according to a time window is performed to obtain a set of sampled segments, and multi-level feature decomposition is performed on each sampled segment in the set of sampled segments to extract color distribution features, texture features, and motion trajectory features, and the color distribution features, texture features, and motion trajectory features are combined to obtain a visual feature matrix; Dividing the audio data into a plurality of overlapping audio frames, performing Fourier transform on each of the audio frames to obtain a spectrogram, extracting pitch features, rhythm features, and emotional features from the spectrogram, and combining the pitch features, rhythm features, and emotional features in time sequence to obtain an audio feature matrix; Sort the user interaction data according to timestamps, construct an interaction behavior sequence, calculate click density, dwell time, and interaction frequency in the interaction behavior sequence, and combine the click density, dwell time, and interaction frequency into a user behavior feature matrix; Utilizing the cross-modal feature fusion network to calculate the temporal correlation between the visual feature matrix, the audio feature matrix, and the user behavior feature matrix, generating a temporal correlation matrix, and resampling and interpolating the three feature matrices based on the temporal correlation matrix to obtain aligned feature data; The information entropy of each modal feature in the aligned feature data is calculated to determine the modal fusion weight, and the aligned feature data is weighted summed using the modal fusion weight and normalized to generate the skit fusion feature vector.

3. The method for optimizing short play delivery based on artificial intelligence according to claim 1, characterized in that: The method of modeling and analyzing the user viewing behavior sequence based on the short drama fusion feature vector using a deep temporal network to obtain user interest migration features, and constructing user portrait data based on the user interest migration features, includes: Constructing a viewing feature sequence according to the short play fusion feature vector, and dividing the viewing feature sequence into time windows to obtain viewing window data; Sequentially encoding a user viewing behavior sequence within each time window in the viewing window data to generate a behavior coding sequence; Using the deep temporal network to perform temporal correlation analysis on the behavior coding sequence, extract viewing duration, switching frequency and interaction intensity, and generate a behavior feature matrix; Calculating a user interest change rate based on the behavior feature matrix, and performing time series accumulation on the user interest change rate to obtain the user interest migration feature; Clustering and grouping the user interest migration features according to interest topics to construct an interest category table; The user portrait data is generated according to the weight distribution of each category in the interest category table.

4. The method for optimizing short play delivery based on artificial intelligence according to claim 1, characterized in that: The method of performing delivery environment modeling and delivery action prediction through a multi-agent reinforcement learning network based on the user portrait data and the short play fusion feature vector to obtain short play delivery strategy data includes: The user portrait data is layered according to the interest dimension to construct a user layer matrix, and the short play fusion feature vector is mapped to the user layer matrix to generate an interest matching degree distribution; Expanding the interest matching degree distribution in time and space dimensions to establish a delivery state space table, and extracting delivery time period, delivery area, and delivery channel from the delivery state space table to form delivery environment features; Inputting the delivery environment characteristics into the multi-agent reinforcement learning network, modeling the competition relationship between each delivery channel, and obtaining a channel competition matrix; Calculate the reward value of the delivery action based on the channel competition matrix, and use the reward value as the optimization target to construct an action value table; Selecting the optimal delivery action combination from the action value table to generate a delivery decision sequence; The delivery decision sequence is matched with delivery constraint conditions to obtain the short play delivery strategy data.

5. The method for optimizing short play delivery based on artificial intelligence according to claim 1, characterized in that: The short play delivery strategy data is used to segment and reorganize the short play materials, and multiple versions of creative materials are generated through a conditional generative adversarial network to form a short play creative material library, including: Converting the skit delivery strategy data into a scene description matrix, and extracting key frames from the skit material based on the scene description matrix to obtain a scene segment set; Performing semantic analysis on the scene segment set to generate a scene semantic vector, and establishing a segment association graph based on the scene semantic vector; Reorganizing and sorting the scene segment set according to the scene connection relationship in the segment association graph to generate a reorganized scene sequence; Inputting the reorganized scene sequence into the conditional generative adversarial network to generate a plurality of initial creative materials with different themes and styles; Performing a creativity score on the initial creative material, extracting creative theme tags and style feature tags, and constructing a creative material feature table; The creative materials in the creative material feature table are classified and sorted according to theme tags and style feature tags to generate the short play creative material library.

6. The method for optimizing short play delivery based on artificial intelligence according to claim 1, characterized in that: The method of analyzing the delivery effect of the materials in the short play creative material library using a distributed computing framework to obtain delivery effect data and generating a delivery resource allocation plan based on the delivery effect data includes: Grouping the materials in the short play creative material library according to the delivery channels, constructing a channel delivery record table, and counting the exposure, click volume, and conversion volume from the channel delivery record table to generate an effect statistical matrix; Calculating the click-through conversion ratio and delivery cost ratio of each delivery channel based on the effect statistical matrix, and combining the click-through conversion ratio and delivery cost ratio to form a channel effectiveness index; Expand the channel performance indicators in time series, construct a performance trend graph, and extract the delivery period weight from the performance trend graph; Perform cross calculation on the delivery period weight and the channel performance index to obtain a time-space delivery score table; Prioritize each delivery channel according to the spatiotemporal delivery scoring table to generate the delivery effect data; Based on the delivery effect data, the delivery resources are allocated in time and space dimensions to obtain the delivery resource configuration plan.

7. The method for optimizing short play delivery based on artificial intelligence according to claim 6, characterized in that: The step of prioritizing each delivery channel according to the spatiotemporal delivery scoring table to generate the delivery effect data includes: Normalizing the scoring data in the spatiotemporal delivery scoring table to construct a standardized scoring matrix, and extracting the time period performance value of each channel from the standardized scoring matrix; Aggregating the performance values ​​of the time periods according to the channel dimension to obtain a channel performance index, and constructing a channel ranking table based on the channel performance index; Calculating historical investment returns for each channel in the channel ranking table, generating a return distribution curve, and extracting an investment advantage coefficient from the return distribution curve; Performing a weighted combination of the delivery advantage coefficient and the channel performance index to form a comprehensive channel score; Sorting the comprehensive scores of the channels in descending order to generate a channel priority list; The channel priority list is combined with the time period performance value to generate the delivery effect data.

8. An AI-based short play delivery optimization system, used to implement the AI-based short play delivery optimization method according to any one of claims 1 to 7, characterized in that: The AI-based short drama delivery optimization system includes: an acquisition module for collecting skit image data, audio data, and user interaction data through a multimodal data acquisition unit, and performing feature extraction and fusion operations on the skit image data, audio data, and user interaction data using a cross-modal feature fusion network to generate a skit fusion feature vector; A modeling module is used to model and analyze the user viewing behavior sequence based on the short drama fusion feature vector using a deep temporal network to obtain user interest migration features and construct user portrait data based on the user interest migration features; A prediction module is used to perform delivery environment modeling and delivery action prediction through a multi-agent reinforcement learning network based on the user portrait data and the short play fusion feature vector to obtain short play delivery strategy data; A reorganization module is used to segment and reorganize the short play materials using the short play delivery strategy data, and generate multiple versions of creative materials through a conditional generative adversarial network to form a short play creative material library; The analysis module is used to analyze the delivery effect of the materials in the short drama creative material library using a distributed computing framework, obtain delivery effect data, and generate a delivery resource configuration plan based on the delivery effect data.

9. A computer device, characterized in that: It includes a memory and a processor, the memory stores a computer program that can be run on the processor, and is characterized in that when the processor executes the computer program, it implements the short drama delivery optimization method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the short drama delivery optimization method based on artificial intelligence as described in any one of claims 1 to 7.

Images