Live broadcast room commodity analysis method and system based on big data

Abstract

The present invention discloses a method and system for analyzing commodities in a live broadcast room based on big data, which obtains live broadcast data of each commodity in the live broadcast room, wherein the live broadcast data includes anchor information, audience data, interactive information, sales data and commodity information; semantic coding and feature extraction are performed on the live broadcast data of each commodity to obtain multiple live broadcast room product query feature vectors; and, the commodities in the live broadcast room are arranged based on the multiple live broadcast room product query feature vectors. In this way, multiple data can be integrated to rank the live broadcast efficiency of commodities in the live broadcast room, thereby optimizing the selection of products.

Description

Technical Field

[0001] The present invention relates to the field of intelligent analysis technology, and in particular to a live broadcast room commodity analysis method and system based on big data. Background Art

[0002] With the development of the Internet, live streaming e-commerce has become a new marketing model, attracting more and more consumers and merchants. Live streaming product analysis is an important part of live streaming e-commerce, which can help merchants understand consumers' needs and preferences, optimize product display and recommendation, and improve conversion rate and sales.

[0003] However, due to the large number and variety of goods in the live broadcast room, as well as the complex and changeable behavior and feedback of consumers, traditional product analysis methods are difficult to adapt to the scenario of live e-commerce, and require the use of big data technology for effective analysis.

[0004] Therefore, we look forward to a live broadcast product analysis solution based on big data. Summary of the invention

[0005] The embodiment of the present invention provides a method and system for analyzing commodities in a live broadcast room based on big data, which obtains live broadcast data of each commodity in the live broadcast room, wherein the live broadcast data includes anchor information, audience data, interaction information, sales data and commodity information; semantic coding and feature extraction are performed on the live broadcast data of each commodity to obtain multiple live broadcast room product query feature vectors; and, the commodities in the live broadcast room are arranged based on the multiple live broadcast room product query feature vectors. In this way, multiple data can be combined to rank the live broadcast efficiency of commodities in the live broadcast room, thereby optimizing the selection of products.

[0006] The embodiment of the present invention also provides a method for analyzing commodities in a live broadcast room based on big data, which includes:

[0007] Obtaining live broadcast data of each product in the live broadcast room, wherein the live broadcast data includes anchor information, audience data, interaction information, sales data and product information;

[0008] Performing semantic coding and feature extraction on the live broadcast data of each of the commodities to obtain multiple live broadcast room product query feature vectors; and

[0009] The commodities in the live broadcast room are arranged based on the multiple live broadcast room product query feature vectors.

[0010] The embodiment of the present invention also provides a live broadcast room commodity analysis system based on big data, which includes:

[0011] A data acquisition module is used to acquire live broadcast data of each product in the live broadcast room, wherein the live broadcast data includes anchor information, audience data, interaction information, sales data and product information;

[0012] A semantic coding and feature extraction module, used for performing semantic coding and feature extraction on the live broadcast data of each commodity to obtain multiple live broadcast room product query feature vectors; and

[0013] The product arrangement module is used to arrange the products in the live broadcast room based on the product query feature vectors of the multiple live broadcast rooms.

[0014] Compared with the existing technology, the live broadcast room product analysis method and system provided by the present application based on big data can help merchants better understand consumer demand and market trends, optimize product selection and recommendation strategies, and improve the conversion rate and sales of the live broadcast room through live broadcast room product analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0016] Figure 1 The present invention provides a flowchart of a method for analyzing commodities in a live broadcast room based on big data in an embodiment of the present invention.

[0017] Figure 2 A schematic diagram of the system architecture of a live broadcast room commodity analysis method based on big data provided in an embodiment of the present invention.

[0018] Figure 3 A block diagram of a live broadcast room commodity analysis system based on big data provided in an embodiment of the present invention.

[0019] Figure 4 This is an application scenario diagram of a live broadcast room commodity analysis method based on big data provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0021] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present invention have the same meanings as those commonly understood by those skilled in the art of the present invention. The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the scope of the present invention.

[0022] In the description of the embodiments of the present invention, it should be noted that, unless otherwise specified and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection or a connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0023] It should be noted that the terms "first\second\third" involved in the embodiments of the present invention are only used to distinguish similar objects, and do not represent a specific order for the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein.

[0024] Live streaming e-commerce refers to a business model that sells and promotes goods through an Internet live streaming platform. It combines e-commerce and live streaming technology to allow merchants to display and promote products and interact and communicate with audiences through real-time video broadcasts.

[0025] In live streaming e-commerce, merchants create live streaming rooms through live streaming platforms to display and introduce various products, including clothing, beauty products, household products, food, etc. Viewers can learn about the characteristics, usage methods and discount information of the products by watching live streaming, and directly place orders to purchase the products through the purchase function of the live streaming platform. At the same time, viewers can interact with the anchor during the live streaming, ask questions, comment or like, which increases interactivity and participation.

[0026] Among them, the audience can interact with the anchor in real time, ask questions, understand product information, and increase confidence and satisfaction in purchasing. Through live video broadcasting, the characteristics and usage effects of the products can be more intuitively displayed, providing a more realistic shopping experience. Live e-commerce has a high conversion rate and sales. The audience can directly place orders during the live broadcast, which promotes the real-time nature of purchasing decisions. Through the live e-commerce platform, merchants can build their own brand image and enhance brand awareness and recognition. During the live broadcast, the audience can interact and communicate with other viewers, share shopping experiences and recommendations, and expand the scope of product dissemination.

[0027] In recent years, live streaming e-commerce has developed rapidly in China and other places and has achieved great success. Many well-known brands and merchants have actively participated in live streaming e-commerce, using live streaming platforms to sell products and promote brands. At the same time, live streaming e-commerce has also provided opportunities for individual anchors to start businesses and find employment.

[0028] Live broadcast commodity analysis refers to the data analysis and evaluation of commodities on live broadcast e-commerce platforms to understand consumer needs and preferences, and optimize commodity display and recommendation strategies to increase conversion rates and sales. The main purpose of live broadcast commodity analysis is to provide merchants with insights into commodity sales and consumer behavior through in-depth research and analysis of commodity data, so as to make more informed decisions and optimize strategies.

[0029] By analyzing the sales data of goods, including sales volume, sales amount, sales trend, etc., we can understand the popularity and sales of goods in order to adjust inventory and supply chain management. Through the audience's behavioral data, such as viewing time, clicks, number of interactions, etc., we can understand the audience's interests and preferences for different goods in order to optimize the display and recommendation strategies of goods. Analyze the audience's evaluation and feedback on the goods, including the content of comments, ratings, etc., to understand the advantages and disadvantages of the goods in order to improve product quality and provide better after-sales service. By analyzing the relationship between the audience's purchasing behavior and goods, we can dig out the potential associations and cross-selling opportunities between goods, and provide cross-category recommendations and matching suggestions. Through the audience's personal information and behavioral data, such as gender, age, region, etc., we can build user portraits to understand the needs and purchasing preferences of different user groups in order to personalize recommendations and customized marketing.

[0030] Through live broadcast commodity analysis, merchants can better understand consumer demand and market trends, optimize commodity selection and recommendation strategies, and improve conversion rates and sales in live broadcasts. At the same time, they can also continuously collect and analyze data, optimize analysis models and algorithms, and improve the accuracy and effectiveness of analysis. The number and variety of commodities on live broadcast e-commerce platforms are huge, and consumer behavior and feedback are also very complex and changeable. Traditional commodity analysis methods may not be able to effectively cope with such large-scale and high-dimensional data, so it is necessary to use big data technology for more accurate and comprehensive analysis.

[0031] Big data technology plays an important role in the analysis of live broadcast products. It can process a large amount of data, including product data, audience behavior data, user comment data, etc., and extract useful information and patterns from them. Through technical means, the product data and audience behavior data of the live broadcast room are collected in real time and stored in a scalable big data storage system, such as a distributed database or a data lake. Using big data analysis technologies, such as machine learning, data mining, and statistical analysis, the live broadcast product data is mined and analyzed to discover the patterns, trends, and associations hidden in the data. By analyzing the audience's personal information and behavior data, user portraits are constructed to understand the interests and preferences of different user groups, and personalized product recommendations and purchase suggestions are provided to the audience based on personalized recommendation algorithms. Using real-time data processing and analysis technology, the real-time data of the live broadcast room is processed and analyzed so that merchants can make decisions and adjust strategies in a timely manner to improve sales results and user satisfaction.

[0032] Through the application of big data technology, live broadcast commodity analysis can more comprehensively and accurately understand consumer needs and behaviors, optimize commodity display and recommendation strategies, and improve sales conversion rates and user experience. At the same time, big data technology also provides merchants with deeper insights and decision-making support, helping them gain an advantage in the fierce market competition.

[0033] In one embodiment of the present invention, Figure 1 The present invention provides a flowchart of a method for analyzing commodities in a live broadcast room based on big data in an embodiment of the present invention. Figure 2 Schematic diagram of the system architecture of a live broadcast commodity analysis method based on big data provided in an embodiment of the present invention. Figure 1 and Figure 2 As shown, a live broadcast room commodity analysis method based on big data according to an embodiment of the present invention includes: 110, acquiring live broadcast data of each commodity in the live broadcast room, wherein the live broadcast data includes anchor information, audience data, interactive information, sales data and commodity information; 120, performing semantic coding and feature extraction on the live broadcast data of each commodity to obtain a plurality of live broadcast room product query feature vectors; and, 130, arranging each commodity in the live broadcast room based on the plurality of live broadcast room product query feature vectors.

[0034] In step 110, the live broadcast data of each product in the live broadcast room is obtained, and these live broadcast data can be obtained through the API of the live broadcast e-commerce platform or other data collection methods. The anchor information includes the anchor's identity information, number of fans, live broadcast experience, etc., which can help understand the anchor's influence and appeal. The audience data includes the number of viewers, viewing time, interactive behavior, etc., which can help understand the audience's participation and interest. Interactive information includes the audience's comments, likes, sharing and other interactive behaviors, which can help understand the audience's feedback and preferences for the products. Sales data includes the sales quantity, sales volume, conversion rate, etc. of the product, which can help understand the popularity and sales of the product. Product information includes the name, description, price, inventory, etc. of the product, which can help understand the attributes and characteristics of the product.

[0035] In step 120, semantic coding and feature extraction of live broadcast data is performed to convert raw data into feature vectors that can be used for analysis and comparison. First, natural language processing technology is used to perform text analysis and coding on anchor information, audience data, interactive information, and product information to extract information such as keywords and sentiment tendencies. Then, statistical features such as sales, conversion rate, number of viewers, viewing time, etc., and other related features such as the number of fans of the anchor and the price of the product are extracted from the sales data and audience data. In this way, the live broadcast data of each product can be converted into a feature vector, which represents the characteristics and performance of the product in various aspects.

[0036] In step 130, based on multiple live broadcast room product query feature vectors, the products can be arranged and sorted to determine the most suitable order for product display. First, by calculating the similarity or distance between feature vectors, the similarity or difference between different products can be measured. Then, the products are sorted using a sorting algorithm, such as sorting based on similarity, sorting based on sales data, etc., to determine the most suitable order for displaying the products. By arranging the various products in the live broadcast room, it is possible to ensure that the most relevant and attractive products are displayed during the live broadcast, thereby increasing the audience's willingness to buy and conversion rate.

[0037] Through the above steps, the exposure rate and sales conversion rate of products can be improved by displaying the most attractive and potentially demanded products; the user experience can be optimized by displaying products that meet the needs of the audience and providing a better shopping experience; personalized recommendations can be provided by analyzing the audience's interests and behavior data to provide personalized product recommendations and purchase suggestions to the audience; resources and time can be saved by arranging and sorting to avoid wasting resources and time; it also helps merchants understand market demand and consumer preferences and gain competitive advantages.

[0038] In view of the above technical problems, the technical concept of the present invention is to comprehensively sort the live broadcast efficiency of the goods in the live broadcast room by combining multiple data, so as to optimize the selection of goods. It should be understood that by comprehensively sorting the goods by combining multiple data, the most attractive and potential goods can be displayed at the forefront of the live broadcast room, which can improve the exposure rate of the goods and the click-through rate of the audience, thereby increasing the conversion rate and sales. The sorting and selection of goods in the live broadcast room directly affect the shopping experience of the audience. By comprehensively considering factors such as the popularity, inventory, and user evaluation of the goods, the goods that better meet the needs of the audience can be presented to them, providing a better shopping experience. By comprehensively analyzing the audience's behavioral data and purchase history, their interests and preferences can be understood, so as to make personalized product recommendations, which can increase the audience's interest in and willingness to buy the recommended goods and improve the hit rate of the recommendation. The number of goods on the live broadcast e-commerce platform is huge, but the live broadcast time is limited. By sorting and optimizing the selection of goods by comprehensive data, it can ensure that the most potential and attractive goods are displayed within a limited time, avoiding the waste of resources and time. The live broadcast e-commerce market is highly competitive. Sorting and optimizing the selection of goods by comprehensive data can help merchants better understand market demand and consumer preferences, thereby gaining a competitive advantage in the market.

[0039] Combining multiple data to rank and optimize product selection for the live broadcast performance of the live broadcast room can improve conversion rate, optimize user experience, make accurate recommendations, save resources and costs, and gain market competitive advantage, which is very important for live broadcast e-commerce platforms and merchants.

[0040] Based on this, in the technical solution of the present invention, first, the live broadcast data of each commodity in the live broadcast room is obtained, wherein the live broadcast data includes anchor information, audience data, interactive information, sales data and commodity information. Here, since the anchor is the core of the live broadcast room, his personal charm, influence and professional level have a great influence on the live broadcast effect. Through the anchor information, the characteristics and style of the anchor can be understood, and the matching degree between the anchor and the commodity can be judged. The audience is the audience group of the live broadcast room, and their feedback and behavior directly affect the sales effect of the commodity. The audience data includes information such as the number of viewers, viewing time, and audience portraits. Through the statistics and analysis of audience data, the popularity of the commodity in different audience groups can be understood, so as to better locate the target audience and adjust the promotion strategy. The interaction in the live broadcast room is an important means to increase user participation and willingness to buy. The interactive information includes the audience's comments, likes, sharing and other behaviors, as well as the anchor's response and interaction. By analyzing the interactive information, the audience's interest and feedback on the commodity can be understood, and the interactive effect of the commodity can be evaluated. The ultimate goal of the live broadcast room is to promote the sales of the commodity. Sales data includes indicators such as the order volume, sales, and conversion rate during the live broadcast. By analyzing sales data, we can understand the sales of different products and judge the popularity and purchasing power of products. Products are the key content of the live broadcast room, and their quality, characteristics and adaptability are directly related to the live broadcast effect. Product information includes product attributes, prices, inventory, etc. By analyzing product information, we can understand the characteristics and selling points of products, judge their matching degree with the target audience, and provide a reference for product selection.

[0041] Then, semantic coding is performed on the live broadcast data of each commodity to obtain multiple semantic coding feature vectors of the live broadcast data of the commodity, that is, the unstructured text data is converted into a numerical representation by semantic coding.

[0042] In one embodiment of the present invention, semantic encoding and feature extraction are performed on the live broadcast data of each commodity to obtain multiple live broadcast room product query feature vectors, including: semantic encoding is performed on the live broadcast data of each commodity to obtain multiple commodity live broadcast data semantic encoding feature vectors; semantic association features are extracted between the multiple commodity live broadcast data semantic encoding feature vectors to obtain an optimized live broadcast room commodity semantic association matrix; and, using the semantic encoding feature vector of each commodity live broadcast data as a query feature vector, respectively calculating the matrix product between it and the optimized live broadcast room commodity semantic association matrix to obtain the multiple live broadcast room product query feature vectors.

[0043] First, semantic encoding is performed on the live broadcast data of each product to obtain multiple semantic encoding feature vectors of the live broadcast data of the product. By semantically encoding the live broadcast data of the product, the text information can be converted into a vector representation to capture the semantic information of the product. Among them, through semantic encoding, the text information such as the name and description of the product can be converted into a vector representation to capture the semantic information of the product, so as to better understand and compare the semantic similarity between different products.

[0044] Then, the semantic association features between the semantic encoding feature vectors of the multiple commodity live broadcast data are extracted to obtain an optimized semantic association matrix of commodities in the live broadcast room. By extracting the semantic association features between the semantic encoding feature vectors of the multiple commodity live broadcast data, a semantic association matrix between commodities can be constructed to measure the semantic similarity and degree of association between different commodities. Among them, by extracting the semantic association features, the semantic similarity between different commodities can be calculated, so as to understand the degree of association and similarity between them. Then, by extracting the semantic association features, an optimized commodity semantic association matrix can be constructed for subsequent commodity arrangement and sorting.

[0045] Finally, the semantically encoded feature vectors of the live broadcast data of each commodity are used as query feature vectors, and the matrix products between them and the optimized semantic association matrix of commodities in the live broadcast room are calculated respectively to obtain the multiple live broadcast room product query feature vectors. By using the semantically encoded feature vectors of the live broadcast data of each commodity as query feature vectors and performing matrix product calculations with the optimized semantic association matrix of commodities in the live broadcast room, multiple live broadcast room product query feature vectors can be obtained for the arrangement and sorting of commodities. Among them, by calculating the matrix product of the query feature vector and the semantic association matrix, the degree of semantic association between each commodity and the live broadcast room can be obtained, so as to make personalized commodity recommendations. By calculating the matrix product of the query feature vector and the semantic association matrix, the degree of semantic association between each commodity and the live broadcast room can be measured, so as to improve the matching degree and display effect of the commodities.

[0046] Next, the semantic encoding feature vectors of the multiple commodity live broadcast data are arranged into a feature matrix and then passed through a live broadcast room commodity semantic association feature extractor based on a convolutional neural network model to obtain a live broadcast room commodity semantic association matrix. That is, the live broadcast room commodity semantic association feature extractor is constructed using a convolutional neural network model to capture the local neighborhood semantic association feature distribution between the semantic encoding feature vectors of the multiple commodity live broadcast data, and characterize the overall sales effect and audience level of the live broadcast room.

[0047] The feature extractor based on the convolutional neural network model can learn higher-level semantic features and capture more complex semantic associations between products. Compared with traditional feature extraction methods, convolutional neural networks can automatically learn more representative and discriminative feature expressions. The convolutional neural network model has a strong nonlinear modeling ability and can capture the complex relationships and nonlinear dependencies between products, which helps to more accurately express the semantic associations between products and improve the quality and accuracy of the semantic association matrix. By using the convolutional neural network model for feature extraction, the expressive power of the semantic association matrix can be improved, which means that the semantic association matrix can better capture the semantic similarity and degree of association between products, and provide a more accurate basis for subsequent product arrangement and sorting. By obtaining a more accurate semantic association matrix of live broadcast products, the accuracy and personalization of product recommendations can be improved, which helps to improve the audience's purchase intention and conversion rate, and increase sales and user satisfaction.

[0048] Using a live broadcast room product semantic association feature extractor based on a convolutional neural network model can improve the quality and accuracy of the live broadcast room product semantic association matrix, thereby bringing better product recommendation effects and commercial value.

[0049] Furthermore, the semantic coding feature vectors of the live broadcast data of each commodity are used as query feature vectors, and the matrix products between them and the semantic association matrix of the commodities in the live broadcast room are calculated respectively to obtain multiple query feature vectors of the products in the live broadcast room. That is, the correlation between the live broadcast effect characteristics of each commodity and the overall sales effect characteristics of the live broadcast room is established to evaluate the live broadcast performance of each commodity in the live broadcast room.

[0050] In one embodiment of the present invention, the semantic association features between the multiple semantic encoding feature vectors of commodity live broadcast data are extracted to obtain an optimized semantic association matrix of commodities in the live broadcast room, including: extracting the semantic association features between the multiple semantic encoding feature vectors of commodity live broadcast data to obtain the semantic association matrix of commodities in the live broadcast room; and performing feature distribution optimization on the semantic association matrix of commodities in the live broadcast room to obtain the optimized semantic association matrix of commodities in the live broadcast room.

[0051] Among them, by extracting semantic association features, the semantic similarity between different products can be calculated, so as to understand the degree of association and similarity between them, which helps to cluster and classify similar products, making it easier for users to browse and purchase related products. By calculating the semantic association features between products, personalized product recommendations can be provided to each user. By analyzing the user's historical behavior and preferences, products related to their interests can be found and recommended to users, improving the user's purchase intention and satisfaction. By optimizing the feature distribution of the semantic association matrix of live broadcast products, the degree of association and weight between products can be adjusted, and then the display and recommendation strategy of products can be optimized. By reasonably sorting and recommending related products, the exposure rate and click-through rate of products can be increased, and the conversion rate and sales can be increased. By optimizing the semantic association matrix of live broadcast products, the matching degree and display effect of products can be improved. Taking into account the degree of semantic association between products and live broadcast rooms can more accurately match user needs and interests, and improve the click-through and purchase rates of products.

[0052] Extracting the semantic association features of the semantic association matrix of products in the live broadcast room and optimizing its feature distribution can improve the semantic understanding and matching of products, realize personalized product recommendations, and increase the conversion rate and sales of products in the live broadcast room.

[0053] In the technical solution of the present invention, each of the multiple commodity live data semantic coding feature vectors expresses the text semantic features of the live data of a single commodity. Therefore, after the multiple commodity live data semantic coding feature vectors are arranged into a feature matrix and passed through a live broadcast room commodity semantic association feature extractor based on a convolutional neural network model, the data text semantic association features of a single commodity and the data text semantic association features between each commodity can be further extracted. Therefore, compared with the data text semantic features of a single commodity as the foreground object features, while representing the sample semantic association features between each commodity, background distribution noise will also be introduced. Moreover, when the high-rank distribution representation of the text semantic association features under the sample-data cross-dimension is performed, the high-dimensional features of the multiple commodity live data semantic coding feature vectors will cause the text semantic probability density mapping error of the live broadcast room commodity semantic association matrix relative to the multiple commodity live data semantic coding feature vectors due to the heterogeneous distribution of the text semantic space, thereby affecting the accuracy of the classification results obtained by the classifier of the multiple live broadcast room product query feature vectors obtained based on the live broadcast room commodity semantic association matrix.

[0054] Based on this, the applicant of the present invention performs soft matching of the rank arrangement distribution of the live broadcast room commodity semantic association matrix, for example, denoted as M, with the feature scale as the imitation mask, which is specifically expressed as: the feature distribution of the live broadcast room commodity semantic association matrix is ​​optimized by the following optimization formula to obtain the optimized live broadcast room commodity semantic association matrix; wherein, the optimization formula is:

[0055]

[0056] Among them, M is the semantic association matrix of the commodities in the live broadcast room, m i,j is the eigenvalue of the (i, j)th position of the semantic association matrix M of the live broadcast room commodities, S is the scale of the semantic association matrix M of the live broadcast room commodities, that is, the width multiplied by the height, represents the square of the Frobenius norm of the semantic association matrix M of the commodities in the live broadcast room, ||M||2 represents the bi-norm of the semantic association matrix M of the commodities in the live broadcast room, and α is a weighted hyperparameter, m′ i,j is the eigenvalue of the (i, j)th position of the semantic association matrix of the optimized live broadcast room products, and exp(·) represents the calculation of the natural exponential function value with the numerical value as the power.

[0057] Here, the soft matching of the rank arrangement distribution of the feature scale as an imitation mask can focus on the foreground object features and ignore the background distribution noise when mapping the high-dimensional features to be regressed into the probability density space, and soft matching of the distribution of the pyramid rank arrangement distribution is performed through different norms of the live broadcast room product semantic association matrix M, so as to effectively capture the correlation between the central area and the tail area of ​​the probability density distribution, avoid the probability density mapping deviation caused by the heterogeneous distribution of the text semantic space of the high-dimensional features of the live broadcast room product semantic association matrix M, thereby improving the accuracy of the classification results obtained by the classifier of the multiple live broadcast room product query feature vectors.

[0058] In one embodiment of the present invention, the commodities in the live broadcast room are arranged based on the multiple live broadcast room product query feature vectors, including: passing the multiple live broadcast room product query feature vectors through a classifier to obtain multiple probability values; and arranging the commodities in the live broadcast room based on the multiple probability values.

[0059] Then, the multiple live broadcast room product query feature vectors are passed through a classifier to obtain multiple probability values, and the commodities in the live broadcast room are arranged based on the multiple probability values. Here, a classifier is used to quantitatively characterize the live broadcast efficiency of each commodity in the live broadcast room.

[0060] By using a classifier to classify the live broadcast product query feature vector, we can get the probability value of the correlation between each product and the query. These probability values ​​can be used to sort the products in the live broadcast room, put the most relevant products in the front, and improve the user's purchase intention and conversion rate. By classifying the live broadcast product query feature vector, we can provide users with personalized product recommendations based on their query intentions and preferences, match the probability value with the user's preferences, and recommend the products that best meet their needs to them, improving user experience and satisfaction.

[0061] By arranging the live broadcast products and displaying the more relevant products in front according to the probability value, the exposure and click-through rate of the products can be improved, which helps to optimize the product display strategy, increase the visibility and attractiveness of the products, and improve sales and conversion rates. By using a classifier to classify the live broadcast product query feature vector, the accuracy of the search can be improved. The classifier can determine whether the product is relevant to the query based on the semantic association between the product and the query, and give the corresponding probability value, which helps to filter out products that are not related to the query and improve the quality and accuracy of the search results.

[0062] By using a classifier to classify the live broadcast room product query feature vectors and arranging the live broadcast room products based on probability values, the relevance sorting and personalized recommendation effects of products can be improved, the product display strategy can be optimized, and the search accuracy and user experience can be improved.

[0063] In summary, the big data-based live broadcast room product analysis method based on the embodiment of the present invention is explained, which integrates multiple data to rank the live broadcast performance of the products in the live broadcast room, thereby optimizing the product selection.

[0064] Figure 3 FIG. 1 is a block diagram of a live broadcast commodity analysis system based on big data provided in an embodiment of the present invention. Figure 3 As shown, the big data-based live broadcast room product analysis system 200 includes: a data acquisition module 210, used to obtain the live broadcast data of each product in the live broadcast room, wherein the live broadcast data includes anchor information, audience data, interaction information, sales data and product information; a semantic coding and feature extraction module 220, used to perform semantic coding and feature extraction on the live broadcast data of each product to obtain multiple live broadcast room product query feature vectors; and a product arrangement module 230, used to arrange the various products in the live broadcast room based on the multiple live broadcast room product query feature vectors.

[0065] In the live broadcast room commodity analysis system based on big data, the semantic coding and feature extraction module includes: a semantic coding unit, which is used to semantically encode the live broadcast data of each commodity to obtain a plurality of commodity live broadcast data semantic coding feature vectors; a semantic association feature extraction unit, which is used to extract the semantic association features between the plurality of commodity live broadcast data semantic coding feature vectors to obtain an optimized live broadcast room commodity semantic association matrix; and a product calculation unit, which is used to use the semantic coding feature vector of each commodity live broadcast data as a query feature vector, and respectively calculate the matrix product between it and the optimized live broadcast room commodity semantic association matrix to obtain the plurality of live broadcast room product query feature vectors.

[0066] In the live broadcast room commodity analysis system based on big data, the semantic association feature extraction unit includes: an association feature extraction subunit, used to extract the semantic association features between the semantic encoding feature vectors of the multiple commodity live broadcast data to obtain the live broadcast room commodity semantic association matrix; and an optimization subunit, used to optimize the feature distribution of the live broadcast room commodity semantic association matrix to obtain the optimized live broadcast room commodity semantic association matrix.

[0067] In the live broadcast room commodity analysis system based on big data, the optimization subunit is used to optimize the feature distribution of the live broadcast room commodity semantic association matrix using the following optimization formula to obtain the optimized live broadcast room commodity semantic association matrix; wherein the optimization formula is:

[0068]

[0069] Among them, M is the semantic association matrix of the commodities in the live broadcast room, m i,j is the eigenvalue of the (i, j)th position of the semantic association matrix M of the live broadcast room commodities, S is the scale of the semantic association matrix M of the live broadcast room commodities, that is, the width multiplied by the height, represents the square of the Frobenius norm of the semantic association matrix M of the commodities in the live broadcast room, ||M||2 represents the bi-norm of the semantic association matrix M of the commodities in the live broadcast room, and α is a weighted hyperparameter, m′ i,j is the eigenvalue of the (i, j)th position of the semantic association matrix of the optimized live broadcast room products, and exp(·) represents the calculation of the natural exponential function value with the numerical value as the power.

[0070] In the live broadcast room commodity analysis system based on big data, the commodity arrangement module is used to: pass the multiple live broadcast room product query feature vectors through a classifier to obtain multiple probability values; and arrange the commodities in the live broadcast room based on the multiple probability values.

[0071] Those skilled in the art will appreciate that the specific operations of each step in the above-mentioned live broadcast commodity analysis system based on big data have been described in detail above. Figure 1 to Figure 2 It has been introduced in detail in the description of the live broadcast room commodity analysis method based on big data, and therefore, its repeated description will be omitted.

[0072] As described above, the live broadcast room commodity analysis system 200 based on big data according to an embodiment of the present invention can be implemented in various terminal devices, such as a server for live broadcast room commodity analysis based on big data. In one example, the live broadcast room commodity analysis system 200 based on big data according to an embodiment of the present invention can be integrated into the terminal device as a software module and / or a hardware module. For example, the live broadcast room commodity analysis system 200 based on big data can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the live broadcast room commodity analysis system 200 based on big data can also be one of the many hardware modules of the terminal device.

[0073] Alternatively, in another example, the big data-based live broadcast room product analysis system 200 and the terminal device may also be separate devices, and the big data-based live broadcast room product analysis system 200 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0074] Figure 4 FIG. 1 is an application scenario diagram of a live broadcast commodity analysis method based on big data provided in an embodiment of the present invention. Figure 4 As shown, in this application scenario, first, the live broadcast data of each product in the live broadcast room is obtained (for example, Figure 4 Then, the acquired live broadcast data is input to a server (for example, Figure 4 In S) shown in , the server is capable of processing the live broadcast data based on a live broadcast room commodity analysis algorithm based on big data to arrange the various commodities in the live broadcast room.

[0075] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for analyzing commodities in a live broadcast room based on big data, characterized in that: include: Obtaining live broadcast data of each product in the live broadcast room, wherein the live broadcast data includes anchor information, audience data, interaction information, sales data and product information; Performing semantic coding and feature extraction on the live broadcast data of each of the commodities to obtain multiple live broadcast room product query feature vectors; and Arranging the commodities in the live broadcast room based on the multiple live broadcast room product query feature vectors; The semantic coding and feature extraction of the live broadcast data of each commodity to obtain multiple live broadcast room product query feature vectors include: Performing semantic coding on the live broadcast data of each of the commodities to obtain a plurality of semantic coding feature vectors of the live broadcast data of the commodities; Extracting semantic association features between the semantic encoding feature vectors of the plurality of commodity live broadcast data to obtain an optimized commodity semantic association matrix for the live broadcast room; and Taking the semantic coding feature vectors of each commodity live broadcast data as query feature vectors, respectively calculating the matrix product between the semantic coding feature vectors and the optimized live broadcast room commodity semantic association matrix to obtain the multiple live broadcast room product query feature vectors; The process of extracting semantic association features between the semantic coding feature vectors of the plurality of commodity live broadcast data to obtain an optimized commodity semantic association matrix in the live broadcast room includes: Extracting semantic association features between the semantic encoding feature vectors of the plurality of commodity live broadcast data to obtain a semantic association matrix of commodities in the live broadcast room; and Performing feature distribution optimization on the live broadcast room commodity semantic association matrix to obtain the optimized live broadcast room commodity semantic association matrix; Among them, the feature distribution optimization of the semantic association matrix of commodities in the live broadcast room is performed to obtain the optimized semantic association matrix of commodities in the live broadcast room, including: the feature distribution optimization of the semantic association matrix of commodities in the live broadcast room is performed according to the following optimization formula to obtain the optimized semantic association matrix of commodities in the live broadcast room; Wherein, the optimization formula is: ;in, is the semantic association matrix of commodities in the live broadcast room, is the semantic association matrix of the commodities in the live broadcast room No. The eigenvalues ​​of the positions, is the semantic association matrix of the commodities in the live broadcast room The dimensions are width times height, Represents the semantic association matrix of the commodities in the live broadcast room The square of the Frobenius norm, Represents the semantic association matrix of the commodities in the live broadcast room The second norm of , and is a weighted hyperparameter, is the first The eigenvalues ​​of the positions, Represents calculation of the natural exponential function value raised to the power of value.

2. The method for analyzing commodities in a live broadcast room based on big data according to claim 1 is characterized in that: Arranging the commodities in the live broadcast room based on the multiple live broadcast room product query feature vectors includes: Passing the plurality of live broadcast room product query feature vectors through a classifier to obtain a plurality of probability values; and The commodities in the live broadcast room are arranged based on the multiple probability values.

3. A live broadcast commodity analysis system based on big data, characterized in that: include: A data acquisition module is used to acquire live broadcast data of each product in the live broadcast room, wherein the live broadcast data includes anchor information, audience data, interaction information, sales data and product information; A semantic coding and feature extraction module, used for performing semantic coding and feature extraction on the live broadcast data of each commodity to obtain multiple live broadcast room product query feature vectors; and A product arrangement module, used for arranging the products in the live broadcast room based on the multiple live broadcast room product query feature vectors; Wherein, the semantic coding and feature extraction module includes: A semantic coding unit, used for semantically coding the live broadcast data of each commodity to obtain a plurality of semantic coding feature vectors of the live broadcast data of the commodity; A semantic association feature extraction unit, used to extract semantic association features between the semantic encoding feature vectors of the plurality of commodity live broadcast data to obtain an optimized semantic association matrix of commodities in the live broadcast room; and A product calculation unit, used to use the semantic coding feature vectors of each commodity live broadcast data as query feature vectors, and respectively calculate the matrix product between the semantic coding feature vectors and the optimized live broadcast room commodity semantic association matrix to obtain the multiple live broadcast room product query feature vectors; Wherein, the semantic association feature extraction unit includes: A correlation feature extraction subunit, used to extract semantic correlation features between the semantic coding feature vectors of the plurality of commodity live broadcast data to obtain a semantic correlation matrix of commodities in the live broadcast room; and An optimization subunit, used for optimizing the feature distribution of the semantic association matrix of commodities in the live broadcast room to obtain the optimized semantic association matrix of commodities in the live broadcast room; The optimization subunit is used to optimize the feature distribution of the live broadcast room commodity semantic association matrix using the following optimization formula to obtain the optimized live broadcast room commodity semantic association matrix; Wherein, the optimization formula is: ;in, is the semantic association matrix of commodities in the live broadcast room, is the semantic association matrix of the commodities in the live broadcast room No. The eigenvalues ​​of the positions, is the semantic association matrix of the commodities in the live broadcast room The dimensions are width times height, Represents the semantic association matrix of the commodities in the live broadcast room The square of the Frobenius norm, Represents the semantic association matrix of the commodities in the live broadcast room The second norm of , and is a weighted hyperparameter, is the first The eigenvalues ​​of the positions, Represents calculation of the natural exponential function raised to a value.

4. The live broadcast room commodity analysis system based on big data according to claim 3 is characterized in that: The commodity arrangement module is used to: Passing the plurality of live broadcast room product query feature vectors through a classifier to obtain a plurality of probability values; and The commodities in the live broadcast room are arranged based on the multiple probability values.

Images