Internet e-commerce live broadcast dynamic management system

Abstract

The invention discloses an internet e-commerce live broadcast dynamic management system, and relates to the technical field of internet e-commerce, the system comprises an e-commerce platform, the e-commerce platform is in communication connection with an information integration analysis module, a knowledge graph construction module, a commodity quality monitoring module, a trust rating module and a quality guarantee module, and the modules are in electric signal connection; and the information integration and analysis module is used for collecting and integrating commodity information, merchant qualification information and user evaluation information of the live broadcasting room to obtain an e-commerce information sequence table. By combining the machine learning algorithm and the knowledge graph, the commodity quality condition is monitored in real time, the commodity quality supervision efficiency is effectively improved, potential quality problems can be quickly identified, early warning notifications are sent to merchants and consumers in time, the merchants are prompted to rectify and modify in time, the risk that unqualified commodities flow into the market is remarkably reduced, and the market competitiveness is improved. The rights and interests of the consumers are guaranteed, and the credibility of the consumers on the e-commerce live broadcast is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet e-commerce, and particularly relates to an Internet e-commerce live broadcast dynamic management system. Background Art

[0002] In recent years, the market scale of the e-commerce live broadcast industry has continued to expand. With the rapid development of Internet technology, live streaming with goods has become an important sales model in the e-commerce industry. Especially in recent years, the rise of short video platforms and social media has made e-commerce live broadcast an extremely popular marketing tool. Through live broadcast, merchants can directly interact with consumers, display products, answer questions, and even promote purchases through real-time discounts, etc., greatly improving the exposure rate and sales volume of products.

[0003] For example, the adaptive Internet e-commerce live broadcast management system disclosed in Chinese Patent Publication No.: CN117010994A includes: a content analysis device, which is used to adopt a deep neural network that has completed multiple learning processes to intelligently analyze the average number of viewers, the number of goods sold, and the sales amount of a set live broadcast room for goods at the current time segment based on each viewing record data corresponding to each historical time segment before the current time segment of the set live broadcast room for goods.

[0004] In the prior art, based on the history of the live broadcast room, the utilization rate of the limited service resources of the live broadcast room for goods is allocated to solve the problem of relatively high operating costs of the live broadcast room for goods. However, due to the variety of products in the live broadcast room, and some merchants taking advantage of consumers' psychology to sell products that do not match the description or have unqualified quality, resulting in the problem of decreased shopping satisfaction and trust. Therefore, how to integrate commodity information, merchant qualifications, and user evaluation data through a knowledge graph, monitor and warn the quality of commodities, and enhance consumers' trust in live e-commerce is the problem to be solved by the present invention. For this reason, an Internet e-commerce live broadcast dynamic management system is proposed. Summary of the Invention

[0005] The purpose of the present invention is to provide an Internet e-commerce live broadcast dynamic management system to solve the problems raised in the above background art.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is:

[0007] An Internet e-commerce live broadcast dynamic management system includes an e-commerce platform, and the e-commerce platform is communicatively connected to an information integration and analysis module, a knowledge graph construction module, a commodity quality monitoring module, a trust rating module, and a quality assurance module, wherein, the modules are electrically connected to each other;

[0008] The information integration and analysis module is used to collect and integrate the commodity information, merchant qualification information, and user evaluation information of the live broadcast room to obtain an e-commerce information sequence table;

[0009] The knowledge graph construction module is used to construct an e-commerce knowledge graph by using the relevant information including product information, merchant qualification information, and user evaluation information in the e-commerce information sequence list;

[0010] The product quality monitoring module is used to establish a product quality evaluation model based on the relevant information of the e-commerce information sequence list, combined with machine learning algorithms and knowledge graphs, to monitor the product quality status in real time, calculate the product quality score, and identify potential quality problems;

[0011] The trust rating module is used to comprehensively construct a knowledge graph in combination with the product quality evaluation results, conduct a risk quantification assessment of the live broadcast room and products, assign corresponding trust levels to merchants, and provide intuitive trust references for consumers;

[0012] The quality assurance module is used to provide feedback reports to merchants by combining product quality monitoring, trust rating, and user evaluation data, point out the problems existing in the products, urge merchants to make rectifications, and at the same time recommend high-quality products and successful marketing cases for merchants to refer to.

[0013] A further improvement of the technical solution of the present invention lies in that: the information integration and analysis module includes a product information integration unit, a merchant qualification review unit, and a user evaluation analysis unit;

[0014] Among them, the product information integration unit is used to collect and integrate multi-source information of all products in the live broadcast room, including the basic attributes, usage methods, and categories of the products, mark the production batches, shelf life and other key information of the products, realize the comprehensive management of product information, and ensure the accuracy and integrity of the information;

[0015] The merchant qualification review unit is used to review the qualifications of merchants entering the live broadcast platform, including business licenses, tax registration certificates, brand authorization letters, historical sales records, and after-sales service commitments, etc., but not limited to this, to ensure that merchants have legal business qualifications, and establish a merchant credit history database to monitor the business status of merchants;

[0016] The user evaluation analysis unit is used to analyze the evaluations of users on products and merchants by combining natural language processing and sentiment analysis algorithms, and mine user needs and feedback.

[0017] A further improvement of the technical solution of the present invention lies in that: the acquisition process of the e-commerce information sequence list includes:

[0018] The docking commodity information integration unit establishes connections with the commodity database, commodity display page, sales records, and inventory system in the live broadcast room, collects commodity information including basic attributes and sales records, and obtains merchant qualification information from the merchant qualification review unit, including information such as business licenses and brand authorizations. At the same time, it collects user evaluation information on commodities and merchants from the user evaluation analysis unit to ensure extensive and comprehensive data sources;

[0019] Preprocess the various types of information collected, remove irrelevant characters, special symbols, and duplicate data, correct obvious spelling and grammar errors, unify the text format, and convert data in different forms into a standardized format;

[0020] Associate and integrate the commodity information, merchant qualification information, and user evaluation information. According to the corresponding relationship between the commodity and the merchant, associate the merchant qualification information with the corresponding commodity information. At the same time, match the user evaluation information with the corresponding commodity and merchant, and then sort the integrated information to form an e-commerce information sequence list.

[0021] A further improvement of the technical solution of the present invention lies in: The knowledge graph construction module specifically includes:

[0022] Extract commodity information, merchant qualification information, and user evaluation information from the e-commerce information sequence list, and use natural language processing technology and pattern matching algorithms to identify commodity, merchant, and user entities. At the same time, define the relationships between entities, including the ownership relationship between the commodity and the merchant, the purchase relationship between the user and the commodity, and the evaluation relationship of the user to the merchant;

[0023] According to the identified entities and relationships, design the structural framework of the knowledge graph, determine the types of nodes and edges. Among them, the nodes represent entities, and the edges represent the relationships between entities, and define the attributes of each node and edge, including the name, price, and category of the commodity, the name and qualification of the merchant, as well as the strength and timestamp of the relationship. Furthermore, construct the schema of the graph database, including the labels, attributes, and their data types of the nodes, and the types and attributes of the edges;

[0024] Fill the specific data in the e-commerce information sequence list into the knowledge graph according to the designed structure. When creating nodes, assign a unique identifier to each entity and store its attribute values in the attribute fields of the corresponding nodes. When establishing edges, connect the corresponding nodes according to the relationship type and direction between entities, and record the attribute values of the relationships to ensure that all data is correctly integrated into the knowledge graph to form a comprehensive and accurate graph structure, which is the e-commerce knowledge graph.

[0025] A further improvement of the technical solution of the present invention lies in: The commodity quality monitoring module specifically includes:

[0026] Extract various types of information related to products from the e-commerce information sequence list, and combine the relationship data between products, merchants, and users in the knowledge graph to analyze the relevant information on product materials, production processes, and user feedback problems, and extract quality evaluation features related to product quality, including the user positive review rate, negative sentiment ratio, and quality complaint rate. The user positive review rate is used to measure the popularity of the product among consumers, the negative sentiment ratio is used to reflect the degree of dissatisfaction of consumers with the product, a higher negative sentiment ratio is a precursor to a product quality crisis, and the quality complaint rate is an important indicator directly reflecting the product quality status;

[0027] Based on the collected data and the extracted features, select a machine learning algorithm based on the decision tree model as the basic model architecture to build a product quality evaluation model. Among them, use historical data to repeatedly train the decision tree model, input the quality evaluation features including the user positive review rate, negative sentiment ratio, and quality complaint rate, and output the product quality score;

[0028] Input the product information collected in real time into the trained product quality evaluation model to monitor the product quality status in real time, output the product quality score, and intuitively reflect the product quality level according to the set abnormal quality threshold, identify potential quality problems. If the product quality score is lower than the set abnormal quality threshold, send a warning notice to the relevant merchant, requiring it to make timely rectification, and send a reminder message to consumers, informing them to pay attention to preventing possible product quality problems.

[0029] A further improvement of the technical solution of the present invention lies in that: the calculation process of the product quality score includes:

[0030] According to business requirements and historical data, determine the weights of the user positive review rate, negative sentiment ratio, and quality complaint rate, and the sum of the weights is 1;

[0031] Obtain the user positive review rate from the product information data, divide the user positive review rate by the maximum allowed value of the user positive review rate to obtain the normalized value of the positive review rate, and use the exponential function combined with the set constant to adjust the normalized value of the positive review rate to obtain the exponential adjustment value of the positive review rate, and then multiply the normalized value of the positive review rate by the exponential adjustment value of the positive review rate to obtain the processing function of the user positive review rate;

[0032] Obtain the negative sentiment ratio from the product information data, divide the negative sentiment ratio by the maximum allowed value of the negative sentiment ratio to obtain the normalized value of the sentiment ratio, and use the exponential function combined with the set constant to adjust the normalized value of the sentiment ratio to obtain the exponential adjustment value of the sentiment ratio, and then subtract the product of the normalized value of the sentiment ratio and the exponential adjustment value of the sentiment ratio from 1 to obtain the processing function of the negative sentiment ratio;

[0033] Obtain the quality complaint rate from the data of commodity information, divide the quality complaint rate by the maximum allowed value of the quality complaint rate to obtain the normalized value of the complaint rate, and use the exponential function combined with the set constant to adjust the normalized value of the complaint rate to obtain the exponentially adjusted value of the complaint rate. Then subtract the product of the normalized value of the complaint rate and the exponentially adjusted value of the complaint rate from 1 to obtain the processing function of the quality complaint rate;

[0034] Multiply the processing functions of the user positive review rate, the negative sentiment ratio, and the quality complaint rate by their corresponding weights respectively, and add the results to obtain the final commodity quality score. When the commodity quality score is close to 1, it indicates that the commodity quality is good; when the commodity quality score is close to 0, it indicates that the commodity quality is poor.

[0035] A further improvement of the technical solution of the present invention lies in that: the trust rating module specifically includes:

[0036] Obtain the quality score of the commodity from the commodity quality evaluation module, and extract multi-dimensional information related to the live broadcast room and the commodity from the knowledge graph, including merchant qualifications, user evaluations, sales data, etc. Then comprehensively analyze and extract the trust evaluation features for trust rating, including commodity quality score, after-sales response speed, merchant violation rate, and commodity return and exchange rate, and integrate them to obtain a trust rating feature set;

[0037] Combine with a trust risk assessment model pre-trained based on trust evaluation features in historical data, input the relevant features of the trust rating feature set, and output a trust risk score;

[0038] Based on the trust risk score, divide the merchants in the live broadcast room into different trust levels, namely high trust level, medium trust level, and low trust level, and assign corresponding trust risk thresholds to each trust level. Among them, merchants with a high trust level have a low trust risk, while merchants with a low trust level have a high trust risk;

[0039] Determine the trust level of the merchant through the trust risk score output by the trust risk assessment model and display it to consumers, guiding consumers to choose live broadcast rooms and commodities with good reputation and low risk, enhancing consumer trust. At the same time, feedback the trust rating result to the merchant to encourage the merchant to improve the commodity quality and service level, promote the healthy development of the e-commerce platform, and form a good market ecology.

[0040] A further improvement of the technical solution of the present invention lies in that: the expression of the trust risk score is:

[0041]

[0042] Wherein, T is the trust risk score. The closer T is to 1, the lower the trust risk and the better the merchant reputation. The closer T is to 0, the higher the trust risk and the worse the merchant reputation. Q is the product quality score, which reflects the overall quality level of the product. Q max is the maximum allowable value of the product quality score, which is used to normalize the product quality score. R is the merchant violation rate, which represents the degree of merchant violation. R max is the maximum allowable value of the merchant violation rate, which is used to normalize the merchant violation rate. X is the after-sales response speed, which represents the speed at which the merchant responds to after-sales issues and is expressed as the reciprocal of the average response time. X max is the maximum allowable value of the after-sales response speed, which is used to normalize the after-sales response speed. P is the product return and exchange rate, which represents the proportion of product returns and exchanges. P max is the maximum allowable value of the product return and exchange rate, which is set according to platform standards and is used to normalize the product return and exchange rate. The value range of T is between 0 and 1;

[0043] Multiple said trust levels correspond one-to-one with multiple said trust risk thresholds. The specific correspondence is as follows:

[0044] The trust risk threshold for the high trust level is: T h ≤T<1;

[0045] The trust risk threshold for the medium trust level is: T l ≤T<T h ;

[0046] The trust risk threshold for the low trust level is: 0<T<T l ;

[0047] Wherein, T is the trust risk score, and T h is the lower threshold of the high trust level and the upper threshold of the medium trust level. T l is the lower threshold of the medium trust level and the upper threshold of the low trust level. T h =0.8, T l =0.5.

[0048] A further improvement of the technical solution of the present invention lies in that: the quality assurance module specifically includes:

[0049] Collect product quality monitoring data, trust rating results, and user evaluation information, analyze the product quality, reputation, and user feedback of merchants, accurately locate problem products and deficiencies, and generate a feedback report to point out product problems and deficiencies to merchants. Combine market trends and successful experiences to provide merchants with rectification suggestions and recommendations for high-quality products, guide merchants to learn successful marketing strategies, and help them improve product quality and market competitiveness;

[0050] Screen out high-quality products and successful marketing cases from the products and marketing activities on the e-commerce platform, provide references for merchants, help them learn advanced practices, optimize their own products and marketing strategies, and then recommend feedback reports and high-quality cases to merchants and deliver them to merchants in a timely manner through the information push system of the e-commerce platform;

[0051] Establish a rectification tracking mechanism, regularly collect merchants' rectification information, analyze the rectification effect, and measure the improvement of merchants by comparing the quality data and user feedback before and after rectification.

[0052] Due to the adoption of the above technical solutions, the technical progress achieved by the present invention compared with the prior art is:

[0053] 1. The present invention provides an Internet e-commerce live broadcast dynamic management system. By combining machine learning algorithms and knowledge graphs, it can monitor the quality status of products in real time, effectively improve the efficiency of product quality supervision, quickly identify potential quality problems, and send early warning notifications to merchants and consumers in a timely manner, prompting merchants to make timely rectifications, thus significantly reducing the risk of unqualified products flowing into the market, protecting the rights and interests of consumers, and enhancing consumers' trust in e-commerce live broadcasts.

[0054] 2. The present invention provides an Internet e-commerce live broadcast dynamic management system. By comprehensively considering multi-dimensional information such as product quality and merchant reputation, it quantitatively evaluates the risks of live broadcast rooms and products, provides intuitive trust references for consumers. Consumers can select live broadcast rooms and products with good reputation and low risks according to the trust level, avoiding shopping risks that may be caused by information asymmetry, and greatly improving consumers' shopping experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0056] Figure 1 It is a schematic diagram of the system function modules of the present invention;

[0057] Figure 2 It is a schematic diagram of the working process of the trust rating module of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0058] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0059] Embodiment 1, as Figure 1 、 Figure 2 shown, the present invention provides an Internet e-commerce live broadcast dynamic management system, including an e-commerce platform, which is communicatively connected to an information integration and analysis module, a knowledge graph construction module, a commodity quality monitoring module, a trust rating module, and a quality assurance module. Among them, the modules are electrically connected to each other;

[0060] The information integration and analysis module is used to collect and integrate the commodity information, merchant qualification information, and user evaluation information in the live broadcast room to obtain an e-commerce information sequence table. The information integration and analysis module includes a commodity information integration unit, a merchant qualification review unit, and a user evaluation analysis unit;

[0061] Among them, the commodity information integration unit is used to collect and integrate the multi-source information of all commodities in the live broadcast room, including the basic attributes, usage methods, and categories of the commodities, and mark the key information of the production batch and shelf life of the commodities to achieve all-round management of the commodity information and ensure the accuracy and integrity of the information. The commodity information integration unit specifically includes: establishing connections with various information sources in the live broadcast room, including the commodity database, commodity display page, sales record, and inventory system, and automatically grabbing the basic attributes of the commodities, covering the commodity name, model, specification, color, size, as well as the usage method and category, and preliminarily screening the obtained commodity information to remove obvious errors or irrelevant content, standardizing and normalizing the collected various commodity information, unifying the format and coding rules of the commodity information, and marking the production batch and shelf life information of the commodities. Classify and organize the collected commodity information according to the attributes and categories of the commodities, associate the multi-source information of the same commodity scattered in each data source to form a structured information set, and comprehensively verify the classified commodity information. Through data comparison and logical judgment methods, check the accuracy and consistency of the information. Among them, for information with doubts or omissions, communicate and confirm with the merchant or relevant personnel to supplement and improve it. Finally, integrate the verified, supplemented, and improved commodity information to form a complete commodity information file, establish a commodity information database, store the integrated commodity information, and achieve all-round management of the commodity information. At the same time, set up a permission management and data backup mechanism to ensure the security and reliability of the information;

[0062] The merchant qualification review unit is used to review the qualifications of merchants entering the live streaming platform, including business licenses, tax registration certificates, brand authorization letters, historical sales records, and after-sales service commitments, but not limited to these, to ensure that merchants have legal business qualifications, establish a merchant credit history database, monitor the business status of merchants, effectively ensure that the products sold during the live stream come from legal and trustworthy merchants, improve consumers' trust in the platform, and prevent unqualified products from entering the live streaming room. The merchant qualification review unit specifically includes: connecting with the entering merchants, collecting their qualification documents such as business licenses, tax registration certificates, brand authorization letters, historical sales records, and after-sales service commitments, and conducting a preliminary format and integrity check on the collected qualification documents, rejecting applications that do not meet the basic requirements, and at the same time verifying the authenticity of the documents to ensure that the basic information provided by the merchants is true and valid; conducting a in-depth review of the qualification documents that have passed the preliminary verification, verifying whether the business scope of the business license covers the categories of products sold, whether the tax payment records of the tax registration certificate are normal, the validity period and authorization scope of the brand authorization letter, and analyzing the historical sales records of the merchants to understand their past sales scale, product categories, and sales trends, evaluating the specific content and feasibility of the after-sales service commitment to ensure that the merchants have a perfect after-sales service system; entering the information of the merchants that have passed the review into the merchant credit history database, establishing a credit file for the merchants, and regularly updating the merchant credit data in the database, including changes in the business status of the merchants, new sales records, and after-sales service performance;

[0063] A user evaluation analysis unit, which is used to analyze users' evaluations of products and merchants by combining natural language processing and sentiment analysis algorithms, mine users' needs and feedback. Specifically, the user evaluation analysis unit includes: collecting users' evaluation data of products and merchants from multiple channels of the live streaming platform, including bullet screen comments in the live streaming room, user evaluations on the product detail page, and after-sales service feedback, and preprocessing the collected original evaluation data, including removing irrelevant characters and special symbols, unifying the text format, and converting different forms of evaluations into standardized data convenient for analysis. Using natural language processing technology to perform word segmentation on the preprocessed evaluation data, splitting continuous text into independent words or phrases to form a vocabulary set, and based on the word segmentation results, extracting the key features of the text, including part-of-speech tagging, word frequency statistics, and keyword recognition of the words, identifying the core words and phrases in the evaluation text, using sentiment analysis algorithms to judge the sentiment tendency of the evaluation data text after feature extraction, analyzing the sentiment words and sentiment expression patterns contained in the text, judging whether the sentiment tendency of each evaluation is positive, negative or neutral, counting the word frequency of positive and negative sentiment words, and presetting the word frequency benchmark values of positive and negative sentiment words, and then calculating the sentiment score to quantify users' satisfaction and sentiment attitudes towards products and merchants, identifying the sentiment intensity and sentiment tendency degree in users' evaluations, based on the results of sentiment analysis, deeply mining users' needs and feedback, analyzing the preferences and expectations mentioned in positive evaluations, as well as the problems and dissatisfaction reflected in negative evaluations, classifying and sorting out the needs and feedback, summarizing the representative and targeted content, and timely feeding back the mined information to relevant departments and merchants;

[0064] The expression of the sentiment score is:

[0065]

[0066] In the formula, S is the sentiment score, W i is the weight of the i-th positive sentiment word. The higher the weight, the greater the contribution of this word to positive sentiment. F i is the word frequency of the i-th positive sentiment word, and the value is a non-negative integer, indicating the number of times this word appears in the text. M is the word frequency benchmark value of positive sentiment words, which is used to adjust the influence degree of word frequency on the sentiment score. According to the specific sentiment analysis task and domain characteristics, refer to the average word frequency of common positive sentiment words in this domain to determine the value of M. Analyze a large number of normal sentiment evaluation texts, count the appearance frequency of positive sentiment words in them, and take an average value as M. K j$f_j$ is the word frequency of the $j$-th negative sentiment word, taking a non-negative integer value, representing the number of times the word appears in the text. $B$ is the word frequency benchmark value of negative sentiment words, determined based on the weight and word frequency statistics of negative sentiment words. First, determine the weights of negative sentiment words. According to the word frequency statistics of negative sentiment words in a large number of texts, take the average value of the product of the weights and word frequencies of all negative sentiment words as $B$. $P$ is the total number of positive sentiment words, taking a non-negative integer value. $N$ is the total number of negative sentiment words, taking a non-negative integer value. The value range of $S$ is $[-\infty, +\infty]$, but in practical applications, it is normalized between $[-1, 1]$ or $[0, 1]$. If $S$ is positive, it indicates a positive sentiment tendency, and the larger the value, the stronger the positive sentiment. If $S$ is negative, it indicates a negative sentiment tendency, and the smaller the value, the stronger the negative sentiment. If $S$ is zero, it indicates a neutral sentiment tendency;

[0067] In addition, the process of obtaining the e-commerce information sequence list includes:

[0068] Connect to the commodity information integration unit, establish connections with the commodity database, commodity display page, sales records, and inventory system in the live broadcast room, collect commodity information including basic attributes and sales records, and connect to the merchant qualification review unit to obtain merchant qualification information, including information such as business licenses and brand authorizations. At the same time, collect user evaluation information of users on commodities and merchants from the user evaluation analysis unit to ensure that the data sources are extensive and comprehensive. Preprocess the various types of information collected, remove irrelevant characters, special symbols, and duplicate data, correct obvious spelling and grammar errors, unify the text format, and convert data in different forms into a standardized format. Among them, for commodity information, unify its attribute description format and coding rules. For merchant qualification information, convert various certificate files into a standardized data format. User evaluation information is subjected to text cleaning and format adjustment, removing irrelevant characters and special symbols. Through format standardization, various types of information can be processed and integrated under the same standard, improving the compatibility and usability of the information. Correlate and integrate commodity information, merchant qualification information, and user evaluation information. According to the corresponding relationship between commodities and merchants, associate merchant qualification information with the corresponding commodity information. At the same time, match user evaluation information with the corresponding commodities and merchants, and then sort the integrated information to form an e-commerce information sequence list;

[0069] A knowledge graph construction module is used to construct an e-commerce knowledge graph by leveraging relevant information including product information, merchant qualification information, and user evaluation information in the e-commerce information sequence list, for data association and integration. Through the knowledge graph, the relationships between products, merchants, and users are visually displayed. Product information, merchant qualification information, and user evaluation information are extracted from the e-commerce information sequence list, and natural language processing techniques and pattern matching algorithms are used to identify product, merchant, and user entities. At the same time, the relationships between entities are defined, including the ownership relationship between products and merchants, the purchase relationship between users and products, and the evaluation relationship of users towards merchants. According to the identified entities and relationships, the structural framework of the knowledge graph is designed, determining the types of nodes and edges. Here, nodes represent entities and edges represent the relationships between entities, and the attributes of each node and edge are defined, including the name, price, and category of products, the name and qualification of merchants, as well as the strength and timestamp of the relationship. Furthermore, the schema of the graph database is constructed, including the labels, attributes, and their data types of nodes, and the types and attributes of edges. The specific data in the e-commerce information sequence list is filled into the knowledge graph according to the designed structure. When creating nodes, a unique identifier is assigned to each entity, and its attribute values are stored in the attribute fields of the corresponding nodes. When establishing edges, according to the relationship type and direction between entities, the corresponding nodes are connected, and the attribute values of the relationship are recorded to ensure that all data is correctly integrated into the knowledge graph, forming a comprehensive and accurate graph structure, which is the e-commerce knowledge graph. The constructed knowledge graph is verified and optimized. The accuracy of entities and relationships is checked to ensure that it truly reflects the information in the original data. The integrity and consistency of the graph are verified to ensure that all relevant entities and relationships are correctly included and represented. Then, according to the verification results, the knowledge graph is adjusted and optimized, including correcting incorrect entities or relationships, supplementing missing data, etc., to improve its quality and usability;

[0070] The product quality monitoring module is used to establish a product quality assessment model based on the relevant information in the e-commerce information sequence list, combined with machine learning algorithms and knowledge graphs, to monitor the product quality status in real time, calculate the product quality score, identify potential quality problems, and send warning notifications to relevant merchants and consumers, reminding merchants to make timely rectifications and informing consumers to pay attention to prevention. It extracts various types of information related to products from the e-commerce information sequence list, and combines the relationship data between products, merchants, and users in the knowledge graph to analyze the relevant information on product materials, production processes, and user feedback problems, and extracts quality assessment features related to product quality, including the user praise rate, negative sentiment ratio, and quality complaint rate. The user praise rate is used to measure the popularity of the product among consumers, the negative sentiment ratio is used to reflect the degree of dissatisfaction of consumers with the product, and a higher negative sentiment ratio is a precursor to a product quality crisis. The quality complaint rate is an important indicator directly reflecting the product quality status. Based on the collected data and extracted features, a machine learning algorithm based on the decision tree model is selected as the basic model architecture to construct a product quality assessment model. Among them, the decision tree model is repeatedly trained using historical data, and the quality assessment features including the user praise rate, negative sentiment ratio, and quality complaint rate are input, and the product quality score is output. During the training process, the parameters are continuously adjusted according to the feedback of the model to continuously optimize the performance of the model. The accuracy and reliability of the model are verified through cross-validation methods to ensure that the model has the ability to effectively identify potential quality problems. Specifically, the relevant data of the quality assessment features in the historical data are divided into a training set, a validation set, and a test set in a ratio of 7:1.5:1.5. The training set is used to train the decision tree model and adjust the parameters of the model so that the model can learn the rules in the data. The validation set is used to evaluate the model, and the evaluation indicators include accuracy, recall rate, F1 value, etc. According to the evaluation results, the model is optimized, and the performance of the model is improved by adjusting the hyperparameters of the model, increasing the training data, improving the feature engineering, etc. The cross-validation method is used to further verify the stability and generalization ability of the model. The test set is used to finally verify the optimized model to obtain the final product quality assessment model. The product information collected in real time is input into the trained product quality assessment model to monitor the product quality status in real time, output the product quality score, and intuitively reflect the product quality level according to the set abnormal quality threshold, identify potential quality problems. If the product quality score is lower than the set abnormal quality threshold, a warning notification is sent to the relevant merchant, requiring it to make timely rectifications, and a reminder message is sent to the consumer, informing it to pay attention to preventing possible product quality problems;

[0071] The calculation process of the product quality score includes:

[0072] According to business requirements and historical data, determine the weights of the user positive review rate, negative sentiment ratio, and quality complaint rate. The sum of the weights is 1. Obtain the user positive review rate from the data of product information, divide the user positive review rate by the maximum allowed value of the user positive review rate to get the normalized value of the positive review rate, and use the exponential function combined with a set constant to adjust the normalized value of the positive review rate to obtain the adjusted value of the positive review rate index. When the user positive review rate is small, the value of the exponential function is small and the contribution to the score is small. Then multiply the normalized value of the positive review rate by the adjusted value of the positive review rate index to obtain the processing function of the user positive review rate. Obtain the negative sentiment ratio from the data of product information, divide the negative sentiment ratio by the maximum allowed value of the negative sentiment ratio to get the normalized value of the sentiment ratio, and use the exponential function combined with a set constant to adjust the normalized value of the sentiment ratio to obtain the adjusted value of the sentiment ratio index. When the negative sentiment ratio is large, the value of the exponential function is large and the negative impact on the score is strong. When the negative sentiment ratio is small, the value of the exponential function is small and the negative impact on the score is weak. Then subtract the product of the normalized value of the sentiment ratio and the adjusted value of the sentiment ratio index from 1 to obtain the processing function of the negative sentiment ratio. Obtain the quality complaint rate from the data of product information, divide the quality complaint rate by the maximum allowed value of the quality complaint rate to get the normalized value of the complaint rate, and use the exponential function combined with a set constant to adjust the normalized value of the complaint rate to obtain the adjusted value of the complaint rate index. When the quality complaint rate is large, the value of the exponential function is large and the negative impact on the score is strong. When the quality complaint rate is small, the value of the exponential function is small and the negative impact on the score is weak. Then subtract the product of the normalized value of the complaint rate and the adjusted value of the complaint rate index from 1 to obtain the processing function of the quality complaint rate. Multiply the processing function of the user positive review rate, the processing function of the negative sentiment ratio, and the processing function of the quality complaint rate by their corresponding weights respectively, and add the results to obtain the final product quality score. When the product quality score is close to 1, it indicates that the product quality is good. When the product quality score is close to 0, it indicates that the product quality is poor;

[0073] The expression of the product quality score is:

[0074] Q = ω1·f(H)+ω2·g(N)+ω3·h(C);

[0075]

[0076] Wherein, Q is the commodity quality score, with a value range between 0 and 1. When Q is closer to 1, the commodity quality is better; when Q is closer to 0, the commodity quality is worse. ω1, ω2, and ω3 are the weights of the user positive review rate, negative sentiment ratio, and quality complaint rate respectively, satisfying ω1 + ω2 + ω3 = 1, and the value range of each weight is between [0, 1]. The larger the weight, the greater the impact of the corresponding feature on the commodity quality score. H is the user positive review rate, indicating the degree of positive reviews of the commodity by users. N is the negative sentiment ratio, indicating the degree of negative evaluations of the commodity by users. C is the quality complaint rate, indicating the proportion of quality problems of the commodity complained by users. f(H) is the processing function of the user positive review rate, and H max is the maximum allowable value of the user positive review rate, used for normalizing the positive review rate. k is a constant, used to adjust the growth rate of the function. When H is larger, the value of f(H) is also larger, indicating that the positive impact of the user positive review rate on the commodity quality score is stronger. g(N) is the processing function of the negative sentiment ratio, and N max is the maximum allowable value of the negative sentiment ratio, used for normalizing the negative sentiment ratio. When N is larger, the value of g(N) is smaller, indicating that the negative impact of the negative sentiment ratio on the commodity quality score is stronger. h(C) is the processing function of the quality complaint rate, and C max is the maximum allowable value of the quality complaint rate, used for normalizing the quality complaint rate. When C is larger, the value of h(C) is smaller, indicating that the negative impact of the quality complaint rate on the commodity quality score is stronger;

[0077] The trust rating module is used to combine the commodity quality assessment results, comprehensively construct a knowledge graph, quantitatively evaluate the risks of the live broadcast room and the commodity, assign corresponding trust levels to the merchants, provide intuitive trust references for consumers, guide consumers to choose live broadcast rooms and commodities with good reputations and low risks, and enhance consumer trust;

[0078] The quality assurance module is used to combine commodity quality monitoring, trust rating, and user evaluation data, provide feedback reports to merchants, point out the problems existing in the commodities, urge merchants to make rectifications, and at the same time recommend high-quality commodities and successful marketing cases for merchants to refer to.

[0079] Example 2, as Figure 1 、 Figure 2 shown. On the basis of Example 1, the present invention provides a technical solution: Preferably, the trust rating module specifically includes:

[0080] Obtain the quality score of the product from the product quality evaluation module, and extract multi-dimensional information related to the live broadcast room and the product from the knowledge graph, including merchant qualifications, user evaluations, sales data, etc. Then, comprehensively analyze and extract trust evaluation features for trust rating, including product quality score, after-sales response speed, merchant violation rate, and product return and exchange rate. Integrate to obtain a trust rating feature set. Combine with a trust risk assessment model pre-trained based on trust evaluation features in historical data, input relevant features of the trust rating feature set, and output a trust risk score. Among them, comprehensively collect feature data related to trust evaluation features from the historical data of the e-commerce platform to form a trust rating feature set of historical data. Clean the collected data to remove duplicates, missing values, and outliers, and perform normalization processing on the data to map the data to the interval [0,1]. Select the random forest model as the basic architecture of the trust risk assessment model. Divide the trust rating feature set of historical data into a training set, a validation set, and a test set with a ratio of 7:1.5:1.5. The training set is used for the actual training of the model, the validation set is used to adjust the model parameters during the training process, and the test set is used to evaluate the final performance of the model. Use the training set to train the random forest model, and continuously adjust the model parameters through an iterative optimization algorithm to make the prediction results of the model as close as possible to the true trust risk score. Use the validation set to optimize the hyperparameters of the model, select evaluation indicators according to business requirements, and the evaluation indicators include accuracy, precision, recall rate, F1 value, etc. Use the test set to verify the trained model, calculate the values of the evaluation indicators, and then obtain the final trust risk assessment model. Divide the merchants in the live broadcast room into different trust levels based on the trust risk score, namely high trust level, medium trust level, and low trust level, and assign corresponding trust risk thresholds to each trust level. Among them, merchants with a high trust level have a low trust risk, while merchants with a low trust level have a high trust risk. Determine the trust level of the merchant through the trust risk score output by the trust risk assessment model and display it to consumers to guide consumers to choose live broadcast rooms and products with good reputation and low risk, enhance consumer trust, and at the same time, feedback the trust rating results to the merchants to encourage merchants to improve product quality and service level, promote the healthy development of the e-commerce platform, and form a good market ecosystem;

[0081] The expression for the trust risk score is:

[0082]

[0083] In the formula, T is the trust risk score. The closer T is to 1, the lower the trust risk and the better the merchant's reputation. The closer T is to 0, the higher the trust risk and the worse the merchant's reputation. Q is the product quality score, which reflects the overall quality level of the product, Q maxis the maximum allowed value for the product quality score, used for normalizing the product quality score. R is the merchant violation rate, indicating the degree of merchant violations, R max is the maximum allowed value for the merchant violation rate, used for normalizing the merchant violation rate. X is the after-sales response speed, indicating the speed at which the merchant responds to after-sales issues, represented by the reciprocal of the average response time, X max is the maximum allowed value for the after-sales response speed, used for normalizing the after-sales response speed. P is the product return and exchange rate, indicating the proportion of product returns and exchanges, P max is the maximum allowed value for the product return and exchange rate, set according to platform standards, used for normalizing the product return and exchange rate. The value range of T is between 0 and 1. When the product quality score is higher, the merchant violation rate is lower, the after-sales response speed is faster, and the product return and exchange rate is lower, the trust risk score is higher, indicating lower trust risk and better merchant reputation. When the product quality score is lower, the merchant violation rate is higher, the after-sales response speed is slower, and the product return and exchange rate is higher, the trust risk score is lower, indicating higher trust risk and worse merchant reputation;

[0084] Multiple trust levels correspond one-to-one with multiple trust risk thresholds. The specific correspondence is as follows:

[0085] The trust risk threshold for the high trust level is: T h ≤T<1, indicating that the merchant has a relatively low trust risk. Such merchants usually have a high product quality score, good after-sales response speed, low violation rate, and few return and exchange situations, with excellent overall performance and worthy of consumers' trust;

[0086] The trust risk threshold for the medium trust level is: T l ≤T<T h , indicating that the merchant's trust risk is at a medium level. The performance of such merchants is relatively average, and there may be certain room for improvement in some aspects, but overall, they can still meet the basic needs of consumers;

[0087] The trust risk threshold for the low trust level is: 0<T<T l , indicating that the merchant has a relatively high trust risk. Such merchants may have more problems in aspects such as product quality, after-sales service, or compliance operations, and consumers need to be cautious;

[0088] Among them, T is the trust risk score, T h is the lower threshold of the high trust level and the upper threshold of the medium trust level, T l is the lower threshold of the medium trust level and the upper threshold of the low trust level, T h =0.8, T l =0.5;

[0089] The quality assurance module specifically includes:

[0090] Collect quality monitoring data of goods, trust rating results and user evaluation information, analyze the quality, reputation of merchants' goods and user feedback, accurately locate problem goods and deficiencies, generate feedback reports, point out the problems and deficiencies of goods to merchants, and provide rectification suggestions and recommendations for high-quality goods to merchants in combination with market trends and successful experiences, guide merchants to learn successful marketing strategies, help them improve the quality of goods and market competitiveness. Screen high-quality goods and successful marketing cases from the goods and marketing activities of the e-commerce platform, provide references for merchants, help merchants learn advanced practices, optimize their own goods and marketing strategies, and then recommend the feedback reports and high-quality cases to merchants and deliver them to merchants in a timely manner through the information push system of the e-commerce platform. Establish a rectification tracking mechanism, regularly collect merchants' rectification information, analyze the rectification effect, and measure the improvement of merchants by comparing the quality data and user feedback before and after rectification.

[0091] As mentioned above, it is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of changes or substitutions, which should be covered within the protection scope of this application. Therefore, the protection scope of this application shall be subject to the protection scope of the claims.

Claims

1. An Internet e-commerce live broadcast dynamic management system, including an e-commerce platform, characterized in that: The e-commerce platform is connected to the information integration and analysis module, the knowledge graph construction module, the product quality monitoring module, the trust rating module and the quality assurance module, wherein the modules are connected by electrical signals; The information integration and analysis module is used to collect and integrate product information, merchant qualification information, and user evaluation information in the live broadcast room to obtain an e-commerce information sequence list; The knowledge graph construction module is used to construct an e-commerce knowledge graph using relevant information in the e-commerce information sequence table; The product quality monitoring module is used to combine machine learning algorithms with knowledge graphs to establish a product quality assessment model, monitor product quality in real time, calculate product quality scores, and identify potential quality issues; The trust rating module is used to combine the product quality assessment results with the knowledge graph to conduct a quantitative risk assessment of the live broadcast room and products, and assign the merchant a corresponding trust level; The quality assurance module is used to combine product quality monitoring, trust rating and user evaluation data to provide feedback reports to merchants, point out problems with the products, and urge merchants to make corrections.

2. The dynamic management system for Internet e-commerce live broadcasts according to claim 1, characterized in that: The information integration and analysis module includes a product information integration unit, a merchant qualification review unit, and a user evaluation analysis unit; The product information integration unit is used to collect and integrate multi-source information of all products in the live broadcast room, including the basic attributes, usage methods, categories of the products, and key information such as the production batch and shelf life of the products; The merchant qualification review unit is used to review the qualifications of merchants entering the live broadcast platform, including business licenses, tax registration certificates, brand authorization letters, historical sales records, and after-sales service commitments, and to establish a merchant credit history database to monitor the merchant's operating status; The user evaluation analysis unit is used to combine natural language processing and sentiment analysis algorithms to analyze users' evaluations of products and merchants and explore user needs and feedback.

3. The dynamic management system for Internet e-commerce live broadcast according to claim 2, wherein: The process of obtaining the e-commerce information sequence list includes: Connect to the product information integration unit, establish connections with the product database, product display page, sales records and inventory system in the live broadcast room, collect product information including basic attributes and sales records, and connect to the merchant qualification review unit to obtain merchant qualification information, including business license and brand authorization information. At the same time, collect user evaluation information on products and merchants from the user evaluation analysis unit; Pre-process all types of collected information, remove irrelevant characters, special symbols and duplicate data, and convert different forms of data into a standardized format; Product information, merchant qualification information and user evaluation information are associated and integrated. According to the corresponding relationship between products and merchants, the merchant qualification information is associated with the corresponding product information. At the same time, the user evaluation information is matched with the corresponding products and merchants, and then the integrated information is sorted to form an e-commerce information sequence list.

4. The dynamic management system for Internet e-commerce live broadcasts according to claim 1, wherein: The knowledge graph construction module specifically includes: Extract product information, merchant qualification information, and user evaluation information from the e-commerce information sequence list, and use natural language processing techniques and pattern matching algorithms to identify product, merchant, and user entities. At the same time, define the relationships between entities, including the ownership relationship between products and merchants, the purchase relationship between users and products, and the evaluation relationship of users to merchants; According to the identified entities and relationships, design the structural framework of the knowledge graph, determine the types of nodes and edges. Among them, nodes represent entities, and edges represent the relationships between entities, and define the attributes of each node and edge, including the name, price, and category of the product, the name and qualification of the merchant, and the strength and timestamp of the relationship. Furthermore, construct the schema of the graph database, including the labels, attributes, and their data types of the nodes, and the types and attributes of the edges; Fill the specific data in the e-commerce information sequence list into the knowledge graph according to the designed structure. When creating nodes, assign a unique identifier to each entity and store its attribute values in the corresponding attribute fields of the nodes. When establishing edges, connect the corresponding nodes according to the relationship type and direction between entities, and record the attribute values of the relationships to form an e-commerce knowledge graph.

5. The dynamic management system for Internet e-commerce live broadcast according to claim 1, wherein: The product quality monitoring module specifically includes: Extract various types of information related to products from the e-commerce information sequence list, and combine the relationship data between products, merchants, and users in the knowledge graph to analyze the relevant information of products, and extract quality evaluation features related to product quality, including user positive review rate, negative sentiment ratio, and quality complaint rate; Based on the collected data and extracted features, select a machine learning algorithm based on the decision tree model as the basic model architecture to construct a product quality evaluation model. Among them, use historical data to repeatedly train the decision tree model, input the quality evaluation features including user positive review rate, negative sentiment ratio, and quality complaint rate, and output the product quality score; Input the real-time collected product information into the trained product quality evaluation model to monitor the product quality status in real time, output the product quality score, and identify potential quality problems according to the set abnormal quality threshold.

6. The dynamic management system for Internet e-commerce live broadcasts according to claim 5, characterized in that: The calculation process of the product quality score includes: Determine the weights of the user positive review rate, negative sentiment ratio, and quality complaint rate according to business requirements and historical data; Obtain the user positive review rate from the product information data, divide the user positive review rate by the maximum allowed value of the user positive review rate to obtain the normalized value of the positive review rate, and use the exponential function combined with the set constant to adjust the normalized value of the positive review rate to obtain the adjusted value of the positive review rate index. Furthermore, multiply the normalized value of the positive review rate by the adjusted value of the positive review rate index to obtain the processing function of the user positive review rate; Obtain the negative sentiment ratio from the product information data, divide the negative sentiment ratio by the maximum allowed value of the negative sentiment ratio to obtain the normalized value of the sentiment ratio, and use the exponential function combined with the set constant to adjust the normalized value of the sentiment ratio to obtain the adjusted value of the sentiment ratio index. Furthermore, subtract the product of the normalized value of the sentiment ratio and the adjusted value of the sentiment ratio index from 1 to obtain the processing function of the negative sentiment ratio; Obtain the quality complaint rate from the data of product information, divide the quality complaint rate by the maximum allowed value of the quality complaint rate to obtain the normalized value of the complaint rate, and use the exponential function combined with the set constant to adjust the normalized value of the complaint rate to obtain the adjusted value of the complaint rate index. Furthermore, subtract the product of the normalized value of the complaint rate and the adjusted value of the complaint rate index from 1 to obtain the processing function of the quality complaint rate; Multiply the processing functions of the user positive review rate, the negative sentiment ratio, and the quality complaint rate by their corresponding weights respectively, and add the results to obtain the final product quality score.

7. The dynamic management system for Internet e-commerce live broadcast according to claim 1, characterized in that: The trust rating module specifically includes: Obtain the quality score of the product from the product quality assessment module, extract multi-dimensional information related to the live broadcast room and the product from the knowledge graph, and then comprehensively analyze and extract the trust evaluation features for trust rating, including the product quality score, after-sales response speed, merchant violation rate, and product return and exchange rate, and integrate them to obtain the trust rating feature set; Combine the trust risk assessment model pre-trained based on the trust evaluation features in historical data, input the relevant features of the trust rating feature set, and output the trust risk score; Based on the trust risk score, divide the merchants in the live broadcast room into different trust levels, namely high trust level, medium trust level, and low trust level, and assign corresponding trust risk thresholds to each trust level; Determine the trust level of the merchant through the trust risk score output by the trust risk assessment model, and display it to consumers to guide consumers to choose live broadcast rooms and products with good reputation and low risk. At the same time, feedback the trust rating result to the merchant.

8. An Internet e-commerce live broadcast dynamic management system according to claim 7, characterized in that: The expression of the trust risk score is: Wherein, T is the trust risk score, Q is the product quality score, Q max is the maximum allowable value of the product quality score, R is the merchant violation rate, R max is the maximum allowable value of the merchant violation rate, X is the after-sales response speed, X max is the maximum allowable value of the after-sales response speed, P is the product return and exchange rate, P max is the maximum allowable value of the product return and exchange rate, and the value range of T is between 0 and 1; Multiple said trust levels correspond one-to-one with multiple said trust risk thresholds, and the specific corresponding relationship is as follows: The trust risk threshold for the high trust level is: T h ≤T < 1; The trust risk threshold for the medium trust level is: T l ≤ T < T h ; The trust risk threshold for the low trust level is: 0 < T < T l ; Among them, T is the trust risk score, T h is the lower threshold of the high trust level and the upper threshold of the medium trust level, T l is the lower threshold of the medium trust level and the upper threshold of the low trust level, T h = 0.8, T l = 0.

5.

9. An Internet e-commerce live broadcast dynamic management system according to claim 1, characterized in that: The quality assurance module specifically includes: Collect product quality monitoring data, trust rating results, and user evaluation information, analyze the product quality, reputation, and user feedback of the merchant, accurately locate the problem products and deficiencies, and generate a feedback report to point out the product problems and deficiencies to the merchant. Combine market trends and successful experiences to provide the merchant with rectification suggestions and recommendations for high-quality products; Screen high-quality products and successful marketing cases from the products and marketing activities of the e-commerce platform, and then recommend the feedback report and high-quality cases to the merchant, and send them to the merchant through the information push system of the e-commerce platform; Establish a rectification tracking mechanism, regularly collect the rectification information of the merchant, analyze the rectification effect, and measure the improvement of the merchant by comparing the quality data and user feedback before and after rectification.

Images