Explosive concept product creation application system and method based on intelligent supply and demand estimation simulation

Through the intelligent supply and demand valuation simulation application system, combined with large language models and multimodal data, the shortcomings of the existing technology in the evaluation of users' deep demand and market potential are solved, more accurate product design and market forecasting are achieved, and products are improved R&D success rate and market competitiveness.

CN118967196APending Publication Date: 2024-11-15李源 +3
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411108857.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The existing technology lacks the exploration of users' deep needs and psychological motivations in auxiliary product design and market forecasting, and it is difficult to provide systematic product solution analysis and market potential assessment, resulting in large market risks in product design.

Method used

Adopt an application system based on intelligent supply and demand valuation simulation, and through the crowd demand identification module, product solution identification module, supply and demand relationship graph module and valuation combination module, combining large language models and multimodal data, accurately analyze user needs and build product solutions. Knowledge graphs, evaluate market potential, and automatically generate marketing materials.

Benefits of technology

It achieves more accurate user needs insights, systematic product solution analysis, reduce market risks, improve product design efficiency and market competitiveness, and enhance marketing efficiency and user attractiveness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967196A_ABST
    Figure CN118967196A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of product scheme matching, and particularly discloses a popular concept product creation application system and method based on intelligent supply and demand estimation simulation, and the system comprises the following modules: a crowd demand recognition module, a product scheme recognition module, a supply and demand relation graph module, an estimation combination module, and a concept product generation module. The crowd demand identification module is used for identifying and predicting user demands from multi-source data, including but not limited to collection of basic information and behavior data, construction of user portraits, crowd division and psychological feature extraction, and accurately analyzing internal demands of users through learning and fine tuning of a large language model; and the product scheme identification module is used for receiving an analysis result of the crowd demand identification module, and can more accurately insight the user demand, more systematically analyze the product scheme, more effectively evaluate the product market potential and more efficiently generate a conceptual product design scheme, thereby helping an enterprise to make a real popular product.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of product solution matching, and specifically to an application system and method for creating explosive concept products based on intelligent supply and demand valuation simulation. Background Art

[0002] In today's highly competitive and rapidly changing market environment, consumer demand is becoming increasingly diverse and personalized. With the development of e-commerce and online shopping platforms, companies are facing unprecedented opportunities and challenges. Consumers' shopping habits and preferences are constantly changing. How to better meet these needs has become one of the keys to corporate success. The development of modern information technology has made data collection and processing more efficient. Through big data analysis and artificial intelligence technology, companies can extract valuable information from large amounts of consumer behavior data. Based on the analysis of this information, companies can formulate more accurate and effective sales strategies.

[0003] As market competition becomes increasingly fierce, companies that want to gain a foothold in the market must continue to launch popular products that meet consumer demand. At present, some technologies have attempted to use artificial intelligence technology to assist product design and market forecasting, such as: Product recommendation system based on market trend analysis: This type of system mainly uses natural language processing (NLP) technology to analyze massive market data, such as social media comments, e-commerce platform reviews, search engine hot words, etc., to identify current market trends and user preferences, and recommend potential hot product directions. However, this type of system can usually only provide relatively general product direction suggestions, lacking guidance on specific product forms and functional characteristics.

[0004] Product design assistance tools based on image recognition: These tools mainly use computer vision technology to analyze a large number of product images, learn the association rules between different product categories, styles, and elements, and automatically generate product appearance design solutions that meet the requirements based on the keywords or sketches entered by the user. However, these tools usually only focus on the appearance design of the product and lack comprehensive consideration of product functions, user needs, and market potential.

[0005] The above-mentioned prior art has the following defects and problems in assisting product design and market forecasting: Lack of in-depth exploration of user needs: Most existing technologies only focus on user surface behavioral data or market trends, and lack exploration of users' deep needs and psychological motivations, resulting in the recommended products being difficult to truly meet user needs and making it difficult to create hit products.

[0006] Lack of systematic analysis of product solutions: Existing technologies usually regard products as independent individuals and lack systematic analysis of product solutions, such as product components, functions, application scenarios, target users, etc., making it difficult to find the optimal combination of product elements.

[0007] Lack of effective assessment of product market potential: Most existing technologies lack effective assessment models for product market potential, making it difficult to predict the market acceptance of products, resulting in greater market risks in product design.

[0008] Lack of automated support for conceptual product design: Most existing technologies can only provide product design suggestions or partial design elements, and lack automated support for conceptual product design, such as automatically generating product promotional images and marketing copy, making it difficult to improve product design efficiency. Summary of the invention

[0009] In view of the shortcomings of the prior art, the present invention provides an application system and method for creating explosive concept products based on intelligent supply and demand valuation simulation, which can effectively solve the problems involved in the above-mentioned background technology and aims to solve the problems existing in the above-mentioned prior art. It provides an application system and method for creating explosive concept products based on intelligent supply and demand valuation simulation, which can more accurately understand user needs, more systematically analyze product solutions, more effectively evaluate product market potential, and more efficiently generate concept product design solutions, thereby helping companies create truly explosive products.

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: The present invention provides a product sales plan matching system based on product information identification, including the following modules: a crowd demand identification module, a product plan identification module, a supply and demand relationship map module, a valuation combination module and a concept product generation module; the crowd demand identification module is used to identify and predict user needs from multi-source data, including but not limited to the collection of basic information and behavior data, the construction of user portraits, crowd segmentation, and psychological feature extraction, and accurately analyze the user's inherent needs through the learning and fine-tuning of a large language model; the product plan identification module is used to receive the analysis results of the crowd demand identification module, and construct a product plan knowledge map involving multiple dimensions based on this, and further understand and identify the key elements in the product plan in detail through the collection of multimodal data and model fine-tuning. elements; the supply and demand relationship map module is used to build a relationship network between users and products based on the user demand analysis results of the user population demand identification module and the product solution details of the product solution identification module, and to create a bipartite graph through user purchase behavior data to map user needs and product elements; the valuation combination module is used to evaluate the market potential of various product solutions, simulate valuation preferences by building intelligent agents representing the characteristics of different populations, and judge the market acceptance of products based on actual market data to find the product combination solution with the greatest market potential; the concept product generation module automatically generates promotional materials for concept products based on the optimal product solution evaluated by the valuation combination module, and generates promotional pictures, copywriting and video scripts for specific markets and user groups through a customized and fine-tuned AIGC model to maximize market influence and user appeal.

[0011] The system innovatively builds user portraits and demand forecasting models by integrating information from multiple data sources (including behavioral data, psychological data, etc.). This approach is uncommon in traditional market analysis and user research. By using the product solution knowledge graph, the system can understand and analyze products at multiple levels. This approach provides more comprehensive data support for product design and development and enhances the market competitiveness of products. Using intelligent agents to simulate the valuation preferences of different groups of people and combining market data for product evaluation, this strategy innovatively uses artificial intelligence to simulate market dynamics and provides a new way to evaluate the market potential of products. Automatically generating targeted marketing materials such as pictures, copywriting and videos, and using AIGC models to customize content creation are innovative in improving marketing efficiency and effectiveness. Through advanced algorithms and big data analysis, the system can accurately identify and predict user needs, providing companies with a powerful tool to optimize products and services. Real-time adjustments to product solutions are based on comprehensive data analysis, which can quickly respond to market changes and enable companies to remain flexible and forward-looking in competition. By building a complex supply and demand relationship network, the system provides in-depth market insights to help companies understand the complex relationship between consumer behavior and product performance. Through automatically generated high-quality marketing materials, companies can more effectively attract target customers and increase product market coverage and user engagement. The data-driven insights provided by the system support fast and accurate business decisions, reduce decision-making risks, and improve the operational efficiency of the company.

[0012] Furthermore, the specific steps of collecting basic information and behavior data, constructing user portraits, dividing groups, and extracting psychological characteristics are: automatically collecting user data from multiple data sources; the data sources include but are not limited to the basic information of users and the behavior data of users; the data include: age, gender, geographic location, educational background, and occupation; the behavior data of users includes browsing history, purchase history, search queries, and social media interactions; The collected data is preprocessed and cleaned, including denoising, deduplication and missing value filling, to ensure the quality and integrity of the data and provide accurate input for subsequent analysis; based on the preprocessed data, a detailed user portrait is constructed, which is used to comprehensively reflect the user's personal attributes, preferences, purchasing habits and behavior patterns; the user portraits are grouped and divided into groups with similar needs and preferences based on the user's similar characteristics, including age, interests and hobbies, and spending power; the user's psychological characteristic data is collected through questionnaire surveys and online behavior data analysis methods, and the psychological characteristic data includes MBTI personality type and Big Five personality traits for further user analysis and product customization, and the collected psychological characteristics are converted into model-recognizable embedding vectors, recorded as embedding vectors Ep, where the embedding vectors are used to characterize the different demand characteristics of users for further data analysis and model training.

[0013] Furthermore, through the learning and fine-tuning of the large language model, the user's intrinsic needs can be accurately analyzed. The specific implementation steps are as follows: Input the user feature vector U into the large language model, which includes the user's basic information and psychological feature embedding vector , and the corresponding human needs labels , using large-scale survey data and user evaluation data on products to fine-tune the large language model to learn the relationship between user characteristics and human needs, based on the input user feature vector U and the psychological feature embedding vector , and the learned model parameters, calculate the different needs of different groups of users The probability of; choose the highest probability value As the model's demand prediction result for this type of user, the specific formula is: ;in, Indicates that user U has a demand The probability of represents the i-th human need, U represents the user feature vector, including the user's basic information and personality feature embedding, The embedding vector representing the psychological theory is used to characterize the characteristics of different needs. Represents model parameters, which are learned through training data.

[0014] Furthermore, the analysis results of the receiving crowd demand identification module are used to construct a multi-dimensional product solution knowledge graph, and further through the collection of multimodal data and model fine-tuning, the key elements in the product solution are carefully understood and identified, specifically: Systematically analyze and organize product solutions based on in-depth domain knowledge and market research; Build a detailed product solution knowledge graph, including product components, appearance design, functional characteristics, expected efficacy, applicable scenarios, target user groups, marketing strategies, price ranges and other multi-dimensional information; Systematically collect and organize various product-related data, including but not limited to product titles, detailed descriptions, high-quality pictures, product usage videos, and user feedback and evaluations, from a variety of channels, including online sales platforms, social media, and customer service databases; Use the collected multimodal data to adjust the specific parameters of the pre-trained multimodal large model. The goal of the adjustment is to enhance the model's ability to recognize key visual elements in product images and videos, and to improve the depth of understanding of key information in product description texts. The specific application algorithm is:

[0015] in Indicates that product P has product elements The probability of represents a text encoder, which is used to extract features of the product text description T. represents an image encoder, which is used to extract features of product image i, Represents a video encoder, which is used to extract features of the product video V. Concat: concatenates multiple feature vectors into a complete feature vector.

[0016] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention provides an application system for creating explosive concept products based on intelligent supply and demand valuation simulation, which can accurately understand user needs and improve the success rate of product research and development. Through modules such as crowd demand identification and supply and demand relationship maps, the potential needs of target users can be more accurately identified, and product elements that meet their needs can be recommended, thereby improving the success rate of product research and development, intelligently matching product solutions, and shortening the product development cycle. Through modules such as product solution identification and valuation combination, the market potential of different product solutions can be quickly evaluated, and the optimal product element combination solution can be found, thereby shortening the product development cycle and reducing product investment risks. Through the market acceptance model, the market potential of the product can be more objectively evaluated, thereby reducing product investment risks, automatically generating concept product promotional materials, and improving product design efficiency. Through the concept product generation module, promotional materials for concept products can be automatically generated, thereby improving product design efficiency and providing support for product promotion.

[0017] (2) By utilizing big data and AI models (such as large language models and graph convolutional networks), the system can accurately identify and predict the specific needs and preferences of different user groups and provide personalized user portraits. By analyzing market data and user feedback, the system dynamically adjusts and optimizes product solutions to ensure that product design is consistent with market demand and user expectations. Using multimodal data and valuation models, the system can evaluate the market acceptance and potential success rate of different product solutions, helping decision makers select the most promising product solutions to bring to market.

[0018] (3) Through AIGC technology, the system can automatically generate high-quality marketing and promotional materials, such as customized pictures, accurate marketing copy and videos, greatly improving marketing efficiency and the ability to attract users. Combined with genetic algorithms and simulated annealing algorithms, the system can explore and optimize possible product portfolio solutions to achieve the best market response and resource allocation. The comprehensive analysis and forecasting tools provided by the system can help business teams understand complex market dynamics and support strategic decisions with more information and data support. Through automated design and evaluation processes, the system shortens the time from product concept to market, enabling companies to quickly respond to market changes and user needs and enhance their competitiveness.

[0019] (4) The system improves the comprehensiveness of data processing and analysis by integrating multiple data types (such as text, images, videos, and time series data) and using multimodal learning technology. This integration can provide richer information to help the system understand products and user behavior more comprehensively. The system can process incoming data in real time and respond quickly to market changes and user behavior. Real-time processing capabilities enable companies to quickly adjust strategies and products and optimize user experience. Using data analysis and machine learning models, the system can provide data-based insights and predictions to help companies make more accurate and scientific decisions. This includes but is not limited to product development, market positioning, and user segmentation. The system's efficient data processing process reduces the storage of redundant data and invalid processing. Through intelligent data storage and management strategies, data utilization is improved and storage costs are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0021] Figure 1 It is a schematic diagram of system module connection of the present invention.

[0022] Figure 2 It is a detailed connection diagram of the system modules of the present invention.

[0023] Figure 3 This is a connection diagram of the system product solution identification module of the present invention.

[0024] Figure 4 It is a schematic diagram of the structural connection of the system market acceptance model of the present invention.

[0025] Figure 5 It is a connection diagram of the system neural network learning module of the present invention.

[0026] Figure 6 A schematic diagram of module connection is generated for the system concept product of the present invention. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0028] Reference Figure 1 As shown, the first aspect of the present invention provides an application system for creating hot-selling concept products based on intelligent supply and demand valuation simulation, including the following modules: a crowd demand identification module, a product solution identification module, a supply and demand relationship map module, a valuation combination module and a concept product generation module.

[0029] The crowd demand identification module is used to identify and predict user needs from multi-source data, including but not limited to the collection of basic information and behavioral data, the construction of user portraits, crowd segmentation, and the extraction of psychological characteristics. It also accurately analyzes the user's intrinsic needs through the learning and fine-tuning of large language models.

[0030] The product solution identification module is used to receive the analysis results of the crowd demand identification module, and build a multi-dimensional product solution knowledge graph based on it. It further understands and identifies the key elements in the product solution through the collection of multimodal data and model fine-tuning.

[0031] The supply and demand relationship map module is used to build a relationship network between users and products based on the user demand analysis results of the user population demand identification module and the product solution details of the product solution identification module, and to create a bipartite graph through user purchase behavior data to map user needs and product elements.

[0032] The valuation portfolio module is used to evaluate the market potential of various product solutions. It simulates valuation preferences by building intelligent agents representing the characteristics of different populations, and judges the market acceptance of products based on actual market data to find the product portfolio with the greatest market potential.

[0033] Concept product generation module, such as Figure 6 As shown, based on the optimal product solution evaluated by the valuation portfolio module, promotional materials for concept products are automatically generated. Through the customized and fine-tuned AIGC model, promotional pictures, copywriting and video scripts targeting specific markets and user groups are generated to maximize market influence and user appeal.

[0034] Specifically, Figure 2As shown, the specific steps of collecting basic information and behavior data, constructing user portraits, dividing groups, and extracting psychological characteristics are: automatically collecting user data from multiple data sources; the data sources include but are not limited to the user's basic information and the user's behavior data; the data include: age, gender, geographic location, educational background and occupation; the user's behavior data includes browsing history, purchase history, search queries and social media interactions; pre-processing and cleaning the collected data, including denoising, de-duplication and missing value filling, to ensure the quality and integrity of the data and provide accurate input for subsequent analysis; based on the pre-processed data, construct a detailed user portrait, which is used to Comprehensively reflect the user's personal attributes, preferences, purchasing habits and behavior patterns; group user portraits and divide the user portraits into groups with similar needs and preferences based on the user's similar characteristics, including age, interests and hobbies, and spending power; collect user psychological characteristic data through questionnaire surveys and online behavior data analysis methods, and the psychological characteristic data includes MBTI personality type and Big Five personality traits for further user analysis and product customization; convert the collected psychological characteristics into an embedding vector that can be recognized by the model, recorded as the embedding vector Ep, where the embedding vector is used to characterize the different demand characteristics of users for further data analysis and model training.

[0035] In a specific embodiment, a user portrait is a data model that integrates multiple attributes of a user, such as age, gender, occupation, hobbies, consumption habits, etc., to form a comprehensive description of the user's behavior and preferences. In machine learning, an embedding vector is a mathematical representation that is usually used to convert high-dimensional data (such as text descriptions) into dense vectors in a low-dimensional space. Psychological feature extraction is to analyze the user's behavioral data through psychological theories (such as the Big Five Personality, MBTI) to extract information reflecting personal personality traits. This information is often used to deepen the understanding of the driving factors of user behavior, and is particularly important for personalized marketing strategies and product design. MBTI is a widely used psychological testing tool for personality classification. It divides people into 16 different personality types, each based on four dimensions: extroversion and introversion, sensing and intuition, thinking and feeling, and judgment and perception. The Big Five personality model is the five basic dimensions that describe personality, including openness, conscientiousness, extroversion, agreeableness, and neuroticism. These traits are used to predict the behavior patterns of individuals in various situations. This conversion helps computers process and understand large amounts of user data more efficiently. For example, psychological characteristic data of users is used in recommendation systems or sentiment analysis. These portraits help companies better understand their customer base to provide more personalized services and products. The quality and integrity of the data are ensured by preprocessing the collected data, including denoising, deduplication and missing value processing. This provides accurate and reliable input for subsequent data analysis, reduces analysis errors and biases, and improves the accuracy of data-driven decision-making. Detailed user portraits are built based on comprehensive user data (including basic information and behavioral data), which can fully reflect the user's personal attributes, preferences, purchasing habits and behavior patterns. This enables companies to better understand customers and provide customers with more personalized services and product recommendations. By grouping users according to characteristics such as age, hobbies, and spending power, the system can create segmented user groups. This segmentation makes marketing and product development more targeted, improves the effectiveness of marketing activities and the efficiency of resource use, and collects users' psychological characteristics such as MBTI and Big Five personality traits using questionnaires and online behavior data to further enrich user portraits. The addition of this psychological dimension enables companies to have a deeper understanding of users' intrinsic needs and motivations, providing a basis for designing products and services that meet users' psychological expectations, and converting psychological feature data into embedding vectors that can be recognized by the model. These vectors are used to train machine learning models, and can take into account users' psychological factors in model predictions and decisions. This enhances the model's predictive accuracy and applicability, especially in user behavior prediction and personalized recommendation systems. With in-depth user understanding and refined market analysis, companies can develop innovative products that better meet market and user needs. At the same time, personalized data analysis supports companies to customize special product solutions and marketing strategies for different user groups, improving user satisfaction and brand loyalty.

[0036] Furthermore, through the learning and fine-tuning of the large language model, the user's intrinsic needs can be accurately analyzed. The specific implementation steps are as follows: Input the user feature vector U into the large language model, which includes the user's basic information and psychological feature embedding vector , and the corresponding human needs labels .

[0037] Large-scale survey data and user evaluation data on products are used to fine-tune the large language model to learn the correlation between user characteristics and human needs.

[0038] The core of the crowd demand recognition model based on the large language model is to use the powerful semantic understanding and representation capabilities of the large language model to learn the complex relationship between user characteristics and human needs. The following is a more detailed formula description of the large language model algorithm.

[0039] The large language model uses the Transformer architecture and uses massive text data for self-supervised learning, with the goal of predicting the next word to appear in a sentence. Its core formula can be expressed as: .

[0040] in, Indicates that the preceding word is known In the case of The probability of represents the hidden state vector output by the last layer of the Transformer model encoder, which represents the semantic information of the input sequence to time step t. Represents the vocabulary matrix, which is used to transform the hidden state vector Mapped to a probability distribution for each word in the vocabulary.

[0041] The specific process of fine-tuning is: embedding the user feature vector U with the psychological theory vector Splicing, while introducing human needs labels As training target and as model input.

[0042] The input vector is encoded using a pre-trained Transformer model to obtain a semantic vector containing user characteristics and psychological theory information.

[0043] Add a fully connected layer to map the semantic vector to the probability distribution of human needs.

[0044] The specific probability formula is: .

[0045] in, Indicates that user U has a demand The probability depends on the user characteristics U and psychological theory and model parameters, Indicates the user feature vector U and the psychological theory embedding vector Splicing, , denote the hidden layer weight matrix and bias vector respectively, represents a nonlinear activation function, and Represent the output layer weight matrix and bias vector respectively.

[0046] According to the input user feature vector U and psychological feature embedding vector , and the learned model parameters, calculate the different needs of different groups of users probability.

[0047] Select the one with the highest probability value As the model's demand prediction result for this type of user, the specific formula is: .

[0048] in, Indicates that user U has a demand The probability of represents the i-th human need, U represents the user feature vector, including the user's basic information and personality feature embedding, The embedding vector representing the psychological theory is used to characterize the characteristics of different needs. Represents model parameters, which are learned through training data.

[0049] In a specific embodiment, Figure 5As shown in the figure, large language models are usually trained on a large amount of text data and use self-supervised learning to predict the next word in the text. They can capture rich language patterns and contextual relationships. Common examples include GPT and BERT series models. Transformer is a deep learning model that is widely used to process sequence data, especially in natural language processing. It relies on the self-attention mechanism, which allows the model to consider all other parts of the sequence when processing any part of the sequence. In the absence of explicit labels, the model relies on the information of the input data itself to learn. For example, in the language model, the model predicts the next word after the given context, and the context itself is used as input data. The Softmax function is a commonly used activation function used in multi-class classification problems. It can convert a vector into a probability distribution in which the probability of each category is positive and the sum of all probabilities is 1. The ReLU (Rectified Linear Unit) activation function is defined as the positive part of the output input (i.e. max(0, input)). It is used to increase nonlinearity and help the model learn complex patterns. Embedded vectors are used in machine learning, especially in NLP, to convert high-dimensional category features (such as words) into low-dimensional continuous vectors. This helps the model process and learn these features. A fully connected layer is a layer in a neural network in which each neuron is connected to each neuron in the previous layer. It can learn nonlinear combinations between input features. In this embodiment, by combining the user's basic information, psychological characteristics and historical behavior data, the large language model can accurately predict the user's specific needs. This prediction is not only based on surface behavior, but also goes deep into the user's psychological motivations and personality traits. The prediction results can be used to personalize the user experience, such as customized recommendations, personalized marketing and product development, thereby improving user satisfaction and brand loyalty, providing in-depth insights based on data, and helping companies make more information-supported decisions, such as market positioning, product improvement and innovation strategies. By understanding the user's deep needs and preferences, marketing activities can be more accurate, effectively reducing ineffective advertising expenditures and improving conversion rates.

[0050] Specifically, the analysis results of the receiving crowd demand identification module are used to build a multi-dimensional product solution knowledge graph, and further through the collection of multimodal data and model fine-tuning, the key elements in the product solution are carefully understood and identified, specifically: Systematically analyze and organize product solutions based on in-depth domain knowledge and market research; Build a detailed product solution knowledge graph, including multi-dimensional information such as product components, appearance design, functional characteristics, expected efficacy, applicable scenarios, target user groups, marketing strategies and price ranges.

[0051] Systematically collect and organize various product-related data, including but not limited to product titles, detailed descriptions, high-quality pictures, product usage videos and user feedback and evaluations, and the data comes from various channels, including online sales platforms, social media and customer service databases.

[0052] The collected multimodal data is used to adjust the specific parameters of the pre-trained multimodal large model. The goal of the adjustment is to enhance the model's ability to recognize key visual elements in product images and videos, and to improve the depth of understanding of key information in product description texts.

[0053] The specific application algorithm is:

[0054] in Indicates that product P has product elements The probability of represents a text encoder, which is used to extract features of the product text description T. represents an image encoder, which is used to extract features of product image i, Represents a video encoder, which is used to extract features of the product video V. Concat: concatenates multiple feature vectors into a complete feature vector.

[0055] It should be explained that the above-mentioned knowledge graph is a structured knowledge representation method, which usually expresses entities (such as product elements) and their relationships in the form of a graph. The product solution knowledge graph specifically refers to the collection and organization of various information about the product (such as component elements, functions, appearance, etc.), and forms a network structure to support complex queries and decision support. Multimodal data refers to data involving multiple forms, such as text, images, videos, etc. In the product information system, these data types are used in combination to provide a comprehensive view of the product. The Softmax function is a function that converts any real number vector into a probability distribution. It is often used to predict the probability of each category in multi-class classification problems. Concat (splicing) In data processing, splicing is to merge multiple vectors end-to-end into a longer vector. In multimodal learning, by splicing feature vectors of different modes, information from different sources can be integrated for comprehensive analysis. In this embodiment, by constructing a detailed product solution knowledge graph, the system can fully understand all aspects of the product, from basic attributes to user feedback. This in-depth understanding supports more effective product improvement and innovation. Combined with information collected from multiple data sources, decision makers can gain a comprehensive view of product performance and market acceptance, allowing them to make more informed decisions. Models fine-tuned using multimodal data can accurately identify and parse key visual and textual data, which is essential for designing effective marketing strategies and promotional materials. The system can predict user preferences for product elements based on user characteristics and historical behavior, support personalized product recommendations, and improve user satisfaction and loyalty.

[0056] Specifically, Figure 3 As shown, the relationship network between users and products is constructed based on the user demand analysis results of the user population demand identification module and the product solution details of the product solution identification module, and a bipartite graph is created through user purchase behavior data to map user needs and product elements as follows: Receive detailed user demand analysis results from the crowd demand identification module, including but not limited to user preferences, behavior patterns and potential needs; receiving details about the product solution from the product solution identification module, including but not limited to the components, functions and target markets of the product; Based on the received data, a complex bipartite “user-product” graph is constructed, which links users to products that they may be interested in; Assign weights to each user and product pair in the graph. These weights are based on the user's purchase history and the product's relevance assessment to reflect the user's preference for different products. On the basis of the "user-product" bipartite graph, further develop and construct the "user-demand-product element" tripartite graph; The three-part diagram includes a direct mapping of user requirements to corresponding product elements.

[0057] It should be explained that the above-mentioned user nodes in the "user-product" bipartite graph are associated with their corresponding demand nodes, for example, users who have purchased "stress relief toys" are associated with the "stress relief" demand node.

[0058] Associate the product nodes in the "user-product" bipartite graph to their corresponding product element nodes, for example, associate "stress relief toys" to product element nodes such as "soft material", "cute shape", and "bright color". A bipartite graph is a special type of graph in which nodes are divided into two independent sets, and each edge in the graph connects nodes from two different sets. In your application, one set is users and the other set is products, and the edge represents the relationship between users and products (such as purchasing behavior). A tripartite graph is a concept in graph theory that contains three groups of nodes, each of which can have links between them. In your scenario, the three groups are users, user needs, and product elements. Such a graph is more suitable for describing more complex relationships, such as the match between user needs and product features. User needs analysis results This refers to insights into user preferences, behavior patterns, and potential needs obtained through data analysis, user behavior research, and psychological analysis. These analysis results help define how to promote and customize products to specific user groups. The details of the product solution involve all the specific information of the product, including its functions, components, applicable scenarios, target markets, etc. This information is crucial to determine how products meet the needs of specific markets or individual users. By constructing a "user-product" bipartite graph and a "user-demand-product element" tripartite graph, the system can more accurately understand and predict users' preferences for specific product elements, thereby providing more personalized recommendations. The system adjusts product features by analyzing user purchase history and needs, thereby better responding to market changes and user needs and improving the market competitiveness of products. By mapping user needs to specific product elements, companies can design products and formulate marketing strategies more targeted, for example, developing new products or adjusting the marketing information of existing products to meet specific user needs. The use of bipartite and tripartite graphs not only provides traditional direct association analysis between users and products, but also allows the system to comprehensively consider the interaction between user needs and product features. This multi-level relationship modeling greatly enhances the expressive power of the model, can more accurately capture complex market dynamics and user behavior, integrate multi-dimensional data into a unified analysis model, enable enterprises to more effectively use existing data resources, avoid data silos, and improve the use value of data assets. Automated data analysis and model prediction can quickly provide decision support and reduce the need and time cost of manual analysis. Fast and data-based decision support brings higher operational efficiency and market adaptability to enterprises.

[0059] Furthermore, the supply and demand relationship map module also includes: Use graph convolutional networks to further learn and reveal the deep connections between user needs and product elements The specific process is: Build a graph structure, including user nodes, demand nodes, and product element nodes; Define edge connections in the graph, including edges between users and requirements, edges between users and products, and edges between products and product elements; Calculate edge weights based on the user's purchase behavior, demand relevance, and product element composition; Initialize node features, where: User node features are initialized to embeddings of user portrait information and demand preferences, and the embeddings of the preferences are obtained through a large language model; The demand node features are embedded through the category and description information of the demand; Product element node features are embedded through the product element name and attribute information Apply multi-layer graph convolutional networks to aggregate the information of neighbor nodes layer by layer to learn high-level feature representations of nodes; The convolution operation of each layer can be expressed as: ; in, is a nonlinear activation function, Represents the node feature matrix of the next layer of the graph convolutional network, represents the negative half power of the degree matrix D, and is used to normalize the adjacency matrix. A is the adjacency matrix of the graph, describing the connection relationship between nodes. Is the current layer The node feature matrix, Represents the weight matrix of the current layer; Use the cross entropy loss function to supervise the learning of the purchase relationship between users and products, and design a specific loss function to capture the strength of the association between user needs and product elements; Use an optimizer to minimize the loss function and update the model parameters. Input the target user’s feature vector into the trained graph convolutional network model to obtain its feature representation in low-dimensional space; Calculate the similarity score between the target user feature vector and the feature vectors of all product element nodes, such as cosine similarity; Sort product elements according to similarity scores, and recommend the top N product elements with the highest scores to users.

[0060] It also includes: algorithm optimization. In the process of graph convolution, an attention mechanism is introduced to give different neighbor nodes different levels of importance, thereby improving the efficiency and accuracy of the model in extracting key information; Jointly train the tasks of user demand prediction and product element recommendation, and use the correlation between these tasks to improve the overall performance of the model, so as to optimize the results of all related tasks by sharing learned features; Use the trained and optimized graph convolutional network model to implement product element recommendations to ensure the effectiveness and accuracy of the model in practical applications.

[0061] It should be explained that the above GCN can effectively utilize the local connection patterns in graph structured data, improve the prediction accuracy of the relationship between user needs and product elements, and through the data representation of the graph structure, the model can better generalize to unseen users or products because it learns structured node relationships, not just the features of specific instances. The introduction of the attention mechanism enables the model to pay more attention to key information, which is especially important when processing large-scale and high-dimensional data, and helps to improve processing efficiency and the relevance of results. Through the use of joint training and optimizers, the model training is more efficient and can more accurately perform multiple tasks such as demand forecasting and product recommendations. The optimized GCN model can quickly process new user data and product information, provide real-time recommendations and analysis, and enhance user experience and satisfaction.

[0062] Specifically, Figure 4 As shown in the figure, the market potential of various product solutions is evaluated. By constructing agents representing the characteristics of different groups to simulate valuation preferences, and combining actual market data to judge the market acceptance of products, the product combination solution with the greatest market potential is found. Specifically: Construct multiple agents (LLMs), each of which represents a specific group of people and personality traits, such as "25-35 years old, urban white-collar women pursuing a refined life", "40-50 years old, middle-aged men paying attention to health preservation", etc.

[0063] Receive the output from the crowd demand identification module, including detailed user demand and preference data. This data provides the basis for the creation and personalized configuration of intelligent agents, ensuring that each intelligent agent can accurately represent a specific user group; Receive detailed product information from the product solution identification module, including product components, functional descriptions, and market positioning; Based on the user needs and product information received above, multiple agents are constructed, each of which simulates the valuation preferences of different users according to their specific population and personality characteristics; Each agent uses product information to generate a valuation preference vector. For example, the “urban white-collar female pursuing a refined life” agent may give a higher valuation to products with “exquisite appearance”, “practical functions” and “high-end quality”.

[0064] Consider the various characteristics of the product and the personalized needs of users; Use the formula to calculate the valuation preference for each product, The specific formula is: ; in, Representing an Agent The valuation preference vector for product P, Indicates user Agent, P represents product information, including product description text, pictures, and prices.

[0065] In a specific embodiment, an agent generally refers to a software entity that simulates the behavior and preferences of a specific user group through an algorithm. Each agent is designed to simulate a group of people with certain personality traits and lifestyles, such as "urban white-collar women who pursue a refined life." By simulating the preferences and evaluation responses of multiple user groups, companies can more accurately understand the potential acceptance of different product solutions in the market. This analysis helps to adjust product design to better meet market demand. Using the valuation preference vector generated by the agent, companies can more accurately locate the specific needs of each user group, thereby providing more personalized product recommendations. The solution supports the decision-making process by integrating data from different modules (such as user needs, product information) and the analysis results of the agent, ensuring that decisions are based on comprehensive data analysis and real user feedback. Based on the results of the agent evaluation, the marketing team can design marketing activities more effectively, promote products with high user preference scores in a targeted manner, increase market acceptance and sales potential, and by analyzing the responses of different agents to product solutions, product development can pay more attention to the actual needs of users and market trends, thereby developing more competitive products.

[0066] Specifically, we evaluate the market potential of various product solutions, simulate valuation preferences by building agents representing the characteristics of different groups of people, and judge the market acceptance of products based on actual market data to find the product combination with the greatest market potential. Specifically: Implement a valuation model based on multimodal data, using the Transformer model as the core processing unit; Configure the model to receive three types of input data: product information (encoded by the product encoder EpE_pEp ), user valuation preferences (encoded by the crowd valuation preference encoder EuE_uEu ), and time series data (encoded by the time series encoder EtE_tEt ); Using the encoded data, the model predicts the market popularity of a given product P for a specific user U and time T. The specific formula is: ; in, represents the market popularity of product P under specific user U and time T; represents the product encoder, which is used to encode product information into a vector. represents the crowd valuation preference encoder, which is used to encode the crowd valuation preference into a vector, represents a time series encoder, which is used to encode time series information into a vector; Genetic simulated annealing algorithm Optimize Using the potential product element set recommended from the supply and demand relationship map module, evaluate the market potential of various product element combination solutions; Genetic simulated annealing algorithm , the objective function is to maximize the market popularity of the product combination solution.

[0067] On the basis of genetic operation, the simulated annealing algorithm is introduced to accept solutions with poor fitness with a certain probability, helping the algorithm to escape from the local optimal solution.

[0068] The initial temperature and cooling rate are set to control the probability of accepting inferior solutions. As the number of iterations increases, the temperature gradually decreases, and the probability of accepting inferior solutions also decreases, eventually converging to the global optimal solution.

[0069] Repeat iteration: Repeat steps 2-4 until a stopping condition is met, such as reaching the maximum number of iterations or finding a solution that meets the preset conditions.

[0070] The specific optimization process is: Randomly generate multiple product element combination schemes as the initial population; Use the valuation judgment model to calculate the market popularity of each solution as a fitness assessment; Perform selection, crossover, and mutation steps to increase the genetic diversity of the population and explore the solution space; Combined with the simulated annealing strategy, solutions with lower fitness are accepted in a timely manner to prevent the algorithm from prematurely converging to the local optimum in a probabilistic manner; Repeat the above steps until the termination condition is met, such as reaching the set number of iterations or the quality of the solution In a specific embodiment, promotional material generation: using the fine-tuned AIGC big model, automatically generate promotional pictures of concept products, product introduction copy, marketing video scripts, and even complete marketing videos, etc., to enhance the visual presentation of product concepts and provide support for product promotion.

[0071] Algorithm: O = AIGC(P', C).

[0072] Among them: O: generated concept product promotional materials, AIGC: pre-trained AIGC model (large language model (selling point and marketing script generation, etc.), text image model, text video model).

[0073] P': Information about the optimal product solution, including product element description, target users, etc.

[0074] C: Generate material control information, such as image style, text length, etc.

[0075] By using the Transformer model combined with multimodal data, the system can more accurately evaluate the market potential and user needs of products, thereby providing data support for market strategies and product development. Through the genetic simulated annealing algorithm, the system can effectively explore and optimize product portfolios and find the product solutions with the greatest market potential. This not only improves the market adaptability of products, but also meets the personalized needs of users. Using the AIGC model to automatically generate promotional materials such as images and videos can greatly improve the visual appeal and user engagement of marketing activities, while reducing the cost and time of content creation. Combined with real-time market data and continuous optimization algorithms, the system can dynamically adjust its recommendations and predictions and respond quickly to market changes. The system provides a comprehensive scientific basis for decision-making by comprehensively analyzing multi-source data and multi-dimensional information, reducing uncertainty and risk in the decision-making process.

[0076] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, It is characterized in that Includes the following modules: Crowd demand identification module, product solution identification module, supply and demand relationship map module, valuation combination module and concept product generation module; The crowd demand identification module is used to identify and predict user needs from multi-source data, including but not limited to the collection of basic information and behavioral data, the construction of user portraits, crowd segmentation, and psychological feature extraction. It also accurately analyzes the user's intrinsic needs through the learning and fine-tuning of large language models. The product solution identification module is used to receive the analysis results of the crowd demand identification module and build a multi-dimensional product solution knowledge graph based on it. It further understands and identifies the key elements of the product solution through the collection of multimodal data and model fine-tuning. The supply-demand relationship graph module is used to build a relationship network between users and products based on the user demand analysis results of the user population demand identification module and the product solution details of the product solution identification module, and to create a bipartite graph through user purchase behavior data to map user needs and product elements; The valuation portfolio module is used to evaluate the market potential of various product solutions. It simulates valuation preferences by building agents representing the characteristics of different groups of people and judges the market acceptance of products based on actual market data to find the product portfolio with the greatest market potential. The concept product generation module automatically generates promotional materials for concept products based on the optimal product solution evaluated by the valuation portfolio module. Through the customized and fine-tuned AIGC model, it produces promotional pictures, copywriting and video scripts for specific markets and user groups to maximize market influence and user appeal.

2. According to claim 1, a hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, Features: The specific steps of collecting the basic information and behavior data, constructing user portraits, dividing the population, and extracting psychological characteristics are as follows: Automatically collect user data from multiple data sources; The data source includes but is not limited to the user's basic information and the user's behavior data; Said include: age, gender, geographic location, educational background and occupation; The user's behavioral data includes browsing history, purchase history, search queries, and social media interactions; Pre-process and clean the collected data, including denoising, deduplication and missing value filling, to ensure the quality and integrity of the data and provide accurate input for subsequent analysis; Based on the preprocessed data, a detailed user profile is constructed, which is used to comprehensively reflect the user's personal attributes, preferences, purchasing habits and behavior patterns; Grouping user portraits into groups with similar needs and preferences based on similar characteristics of users, including age, interests, and spending power; Collect user psychological characteristic data through questionnaire surveys and online behavior data analysis methods, including MBTI personality type and Big Five personality traits, for further user analysis and product customization; The collected psychological features are converted into embedding vectors recognizable by the model, which are recorded as embedding vectors Ep, where the embedding vectors are used to characterize different demand characteristics of users for further data analysis and model training.

3. According to claim 1, a hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, Features: Through the learning and fine-tuning of the large language model, the specific steps to accurately analyze the user's intrinsic needs are as follows: Input the user feature vector U into the large language model, which includes the user's basic information and psychological feature embedding vector , and the corresponding human needs labels ; Use large-scale survey data and user evaluation data on products to fine-tune the large language model to learn the correlation between user characteristics and human needs; According to the input user feature vector U and psychological feature embedding vector , and the learned model parameters, calculate the different needs of different groups of users The probability of Select the one with the highest probability As the model's demand prediction result for this type of user, the specific formula is: ; in, Indicates that user U has a demand The probability of represents the i-th human need, U represents the user feature vector, including the user's basic information and personality feature embedding, The embedding vector representing the psychological theory is used to characterize the characteristics of different needs. Represents model parameters, which are learned through training data.

4. According to claim 3, a hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, Features: The analysis results of the receiving crowd demand identification module are used to build a multi-dimensional product solution knowledge graph. Through the collection of multimodal data and model fine-tuning, the key elements in the product solution are carefully understood and identified, specifically: Systematically analyze and organize product solutions based on in-depth domain knowledge and market research; Build a detailed product solution knowledge graph, including product components, appearance design, functional characteristics, expected efficacy, applicable scenarios, target user groups, marketing strategies, price ranges and other multi-dimensional information; Systematically collect and organize various product-related data, including but not limited to product titles, detailed descriptions, high-quality pictures, product usage videos, and user feedback and evaluations, from a variety of channels, including online sales platforms, social media, and customer service databases; Use the collected multimodal data to adjust the specific parameters of the pre-trained multimodal large model. The goal of the adjustment is to enhance the model's ability to recognize key visual elements in product images and videos, and to improve the depth of understanding of key information in product description text; The specific application algorithm is: in Indicates that product P has product elements The probability of represents a text encoder, which is used to extract features of the product text description T. represents an image encoder, which is used to extract features of product image i, Represents a video encoder, which is used to extract features of the product video V. Concat: concatenates multiple feature vectors into a complete feature vector.

5. According to claim 1, a hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, Features: The relationship network between users and products is constructed based on the user demand analysis results of the user population demand identification module and the product solution details of the product solution identification module, and a bipartite graph is created through user purchase behavior data to map user needs and product elements as follows: Receive detailed user demand analysis results from the crowd demand identification module, including but not limited to user preferences, behavior patterns and potential needs; receiving details about the product solution from the product solution identification module, including but not limited to the components, functions and target markets of the product; Based on the received data, a complex "user-product" bipartite graph is constructed, which links users to products that they may be interested in; Assign weights to each user and product pair in the graph. These weights are based on the user's purchase history and the product's relevance assessment to reflect the user's preference for different products. On the basis of the "user-product" bipartite graph, further develop and construct the "user-demand-product element" tripartite graph; The three-part diagram includes a direct mapping of user requirements to corresponding product elements.

6. According to claim 5, a hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, Features: The supply and demand relationship graph module also includes: Use graph convolutional networks to further learn and reveal the deep connections between user needs and product elements The specific process is: Build a graph structure, including user nodes, demand nodes, and product element nodes; Define edge connections in the graph, including edges between users and requirements, edges between users and products, and edges between products and product elements; Calculate edge weights based on the user's purchase behavior, demand relevance, and product element composition; Initialize node features, where: User node features are initialized to embeddings of user portrait information and demand preferences, and the embeddings of the preferences are obtained through a large language model; The demand node features are embedded through the category and description information of the demand; Product element node features are embedded through the product element name and attribute information Apply multi-layer graph convolutional networks to aggregate the information of neighboring nodes layer by layer to learn high-level feature representations of nodes; The convolution operation of each layer can be expressed as: ; in, is a nonlinear activation function, Represents the node feature matrix of the next layer of the graph convolutional network, represents the negative half power of the degree matrix D, and is used to normalize the adjacency matrix. A is the adjacency matrix of the graph, describing the connection relationship between nodes. Is the current layer The node feature matrix, Represents the weight matrix of the current layer; Use the cross entropy loss function to supervise the learning of the purchase relationship between users and products, and design a specific loss function to capture the strength of the association between user needs and product elements; Use an optimizer to minimize the loss function and update the model parameters; Input the target user’s feature vector into the trained graph convolutional network model to obtain its feature representation in low-dimensional space; Calculate the similarity score between the target user feature vector and the feature vectors of all product element nodes; Sort product elements according to similarity scores, and recommend the top N product elements with the highest scores to users.

7. According to claim 1, a hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, Features: The above-mentioned evaluation of the market potential of various product solutions is carried out by constructing intelligent agents representing the characteristics of different groups to simulate valuation preferences, and judging the market acceptance of products in combination with actual market data, so as to find the product combination solution with the greatest market potential. Specifically, it is as follows: Receive the output from the crowd demand identification module, including detailed user demand and preference data. This data provides the basis for the creation and personalized configuration of intelligent agents, ensuring that each intelligent agent can accurately represent a specific user group; Receive detailed product information from the product solution identification module, including product components, functional descriptions, and market positioning; Based on the user needs and product information received above, multiple agents are constructed, each of which simulates the valuation preferences of different users according to their specific population and personality characteristics; Each agent uses product information to generate a valuation preference vector, taking into account various product characteristics and the user's personalized needs; Use the formula to calculate the valuation preference for each product, The specific formula is: ; in, Representing an Agent The valuation preference vector for product P, Indicates user Agent, P represents product information, including product description text, pictures, and prices.

8. According to claim 7, a hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, Features: The above-mentioned evaluation of the market potential of various product solutions is carried out by constructing intelligent agents representing the characteristics of different groups to simulate valuation preferences, and judging the market acceptance of products in combination with actual market data, so as to find the product combination solution with the greatest market potential. Specifically, it is as follows: Implement a valuation model based on multimodal data, using the Transformer model as the core processing unit; Configure the model to receive three types of input data: product information, user valuation preferences, and time series data; Using the encoded data, the model predicts the market popularity of a given product P for a specific user U and time T. The specific formula is: ; in, represents the market popularity of product P under specific user U and time T; represents the product encoder, which is used to encode product information into a vector. represents the crowd valuation preference encoder, which is used to encode the crowd valuation preference into a vector, represents a time series encoder, which is used to encode time series information into a vector; Genetic simulated annealing algorithm Optimize Using the potential product element set recommended from the supply and demand relationship map module, evaluate the market potential of various product element combination solutions; Genetic simulated annealing algorithm , the objective function is to maximize the market popularity of the product combination solution.

9. According to claim 1, a hot-selling concept product creation application system based on intelligent supply and demand valuation simulation, Features: The optimal product solution evaluated by the valuation combination module automatically generates promotional materials for concept products. Through the customized and fine-tuned AIGC model, promotional pictures, copywriting and video scripts for specific markets and user groups are generated to maximize market influence and user appeal. Specifically: Receive optimal product solution data from the valuation portfolio module, the data including detailed product element description, target user group and market positioning; Customize and fine-tune the pre-trained AIGC model to suit specific product target users, application scenarios, and style positioning; Fine-tuning is applied to different generative models; Using the fine-tuned AIGC model, we automatically generate promotional content for concept products, including but not limited to product introduction copy, marketing video scripts, product images and videos; The generated promotional materials are designed to enhance the visual and textual presentation of the product concept and support the product's marketing activities.

10. A method for creating and applying hot-selling concept products based on intelligent supply and demand valuation simulation. Features: Identify and predict user needs from multi-source data, including but not limited to the collection of basic information and behavioral data, the construction of user portraits, the segmentation of populations, the extraction of psychological characteristics, and accurately analyze the inherent needs of users through the learning and fine-tuning of large language models; Receive the analysis results of the crowd demand identification module and build a multi-dimensional product solution knowledge graph based on it. Further understand and identify the key elements of the product solution through the collection of multimodal data and model fine-tuning. Build a relationship network between users and products based on the user demand analysis results of the user population demand identification module and the product solution details of the product solution identification module, and create a bipartite graph through user purchase behavior data to map user needs and product elements; Evaluate the market potential of various product solutions by building agents that represent the characteristics of different groups to simulate valuation preferences and judge the market acceptance of products based on actual market data to find the product combination with the greatest market potential; Based on the optimal product solution evaluated by the valuation portfolio module, promotional materials for concept products are automatically generated. Through the customized and fine-tuned AIGC model, promotional pictures, copywriting and video scripts targeting specific markets and user groups are generated to maximize market influence and user appeal.

Citation Information

Cited By

  • Model output optimization control method and device based on entropy and score

    CN119670898A

  • User portrait construction method based on knowledge graph and psychological knowledge pushing method

    CN120144862A