A digital management system and method for measuring the attention of commodities

Abstract

The application discloses a kind of digital management system and method for measuring commodity attention, belong to digital management technical field.The application inputs the e-commerce website needing to be crawled and the data item collected, sends HTTP request to obtain web page content, parses HTML, processes page and dynamic content, sets request interval;According to the data item, the click rate and purchase rate of the commodity are calculated, and the heat index of the commodity is generated according to the click rate and purchase rate;According to user evaluation, generate user evaluation comprehensive score, generate social media evaluation comprehensive score according to social media evaluation, generate user feedback index of the commodity;Generate the attention index of the commodity, obtain the same plate commodity dataset, the same category commodity dataset and the same concept commodity dataset, generate the attention index of the same plate, the same category and the same concept commodity;Use artificial intelligence model to form the attention index standard of the commodity, generate commodity attention.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital management, and particularly relates to a digital management system and method for measuring attention of commodities. BACKGROUND

[0002] With the rapid development of e-commerce, the types and quantities of commodities on e-commerce platforms are increasing, and the attention and feedback data of users on commodities gradually become important reference for e-commerce platform commodity management and marketing decision-making. In order to improve the operation efficiency and user experience of the platform, how to effectively measure the attention of commodities has become a key problem.

[0003] Traditional commodity management methods usually rely on single data items such as page views, click volumes or purchase volumes, and cannot fully reflect the actual interest and feedback of users on commodities. In addition, the popularity of social media makes the evaluation and discussion of users on commodities no longer limited to the e-commerce platform itself, but widely spread on the network through various social activities such as likes, comments and shares. However, existing technologies often fail to effectively integrate the data of e-commerce platforms and social media, and cannot fully evaluate the performance of commodities in multiple dimensions. SUMMARY

[0004] The purpose of the present application is to provide a digital management system and method for measuring the attention of commodities to solve the problems existing in the prior art.

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] A digital management method for measuring the attention of commodities, the method comprising the following steps:

[0007] S100, input the e-commerce website to be crawled and the collected data items, the data items including the page views, click volumes, purchase volumes, social media evaluations and user evaluations of commodities, use the Requests library to send HTTP requests to obtain web page contents, use the BeautifulSoup library to parse HTML, process pagination and dynamic contents, and set request intervals;

[0008] S200, according to the data items, calculate the click rate and purchase rate of commodities, generate a heat index of commodities according to the click rate and purchase rate; the user evaluation includes user scores, user comment texts, sentiment tendencies, user identities, user labels and picture / video feedbacks, generate a user evaluation comprehensive score according to the user evaluation, the social media evaluation includes like numbers, share numbers, mention frequencies, topic labels and sentiment tendencies, generate a social media evaluation comprehensive score according to the social media evaluation, and generate a user feedback index of commodities according to the user evaluation comprehensive score and the social media evaluation comprehensive score;

[0009] S300, generating a focus index of the commodity according to the heat index of the commodity and the user feedback index of the commodity, obtaining a commodity board, a commodity category and a commodity concept of the commodity, obtaining a same-board commodity dataset according to the commodity board, obtaining a same-category commodity dataset according to the commodity category, obtaining a same-concept commodity dataset according to the commodity concept, generating a focus index of a same-board commodity according to the same-board commodity dataset, generating a focus index of a same-category commodity according to the same-category commodity dataset, and generating a focus index of a same-concept commodity according to the same-concept commodity dataset;

[0010] S400, using an artificial intelligence model to integrate the focus indexes of the same-board commodity, the same-category commodity and the same-concept commodity, forming a focus index standard of the commodity, comparing the focus index standard of the commodity with the focus index of the commodity, and generating a focus of the commodity.

[0011] According to step S100, the user specifies the URL of the target e-commerce website, and specifies the data items to be collected, including the browsing volume, click volume, purchase volume, social media evaluation and user evaluation of the commodity; an HTTP request is sent using a web crawler, and the web page content is obtained from the server, the request header simulates browser access to avoid being identified as a robot program; the crawler controls the frequency when sending requests, sets the request interval, and prevents frequent requests in a short time from causing a burden on the target website or causing IP to be banned; after obtaining the web page content, the HTML page is parsed, and according to the different structures of the e-commerce website, the crawler program uses the HTML parsing tool BeautifulSoup to extract the data items, which are embedded in specific tags of the web page; the crawler finds the specific tags according to the page structure and extracts the text content therein;

[0012] The e-commerce website divides the commodity list into multiple pages for display, and the crawler program can automatically identify the page links and grab the data of each page; the URL of the page button or the dynamic request link is parsed to achieve this, and the content of each page is obtained by sending a request to it in turn.

[0013] For e-commerce websites that use JavaScript to asynchronously load and dynamically present commodity data, the crawler program needs to handle dynamic content, simulate browser behavior, load JavaScript, or send API requests to data interfaces to obtain dynamically generated data; if the e-commerce website integrates user interaction data from social media, the crawler program parses the embedded social media information to collect social media evaluation data of the commodity; if the e-commerce website does not integrate user interaction data from social media, the crawler accesses the social media platform and collects social media evaluations related to the commodity through the corresponding API or page scraping;

[0014] The e-commerce website displays user reviews, the crawler locates the display position of the user review data, extracts the user rating, user comment text, sentiment tendency, user identity, user label and picture / video feedback provided by the user, and the crawler program classifies the user review data;

[0015] The crawler sets a random time interval after each request to simulate human user operation behavior and dynamically changes the IP address using an IP proxy pool.

[0016] According to step S200, based on the obtained browsing volume, click volume and purchase volume of the commodity, the click rate and purchase rate are calculated, the click rate is the ratio of the click volume to the browsing volume, and the purchase rate is the ratio of the purchase volume to the click volume, the click rate reflects the preliminary interest of the user, and the purchase rate measures the effect of interest conversion into actual purchase, and the heat index of the commodity is calculated based on the weighted calculation of the click rate and the purchase rate.

[0017] According to step S200, in the user review, the user rating represents the score given by the user for the commodity, and is presented in the form of a positive integer; the user comment text is obtained, the sentiment tendency of the comment text is analyzed through natural language processing, it is judged whether the user's comment is positive, neutral or negative, and quantified as a score as the sentiment tendency; the user identity is the authentication information of the user on the e-commerce platform; the user label is the characteristic label for evaluating the commodity attached by the user when evaluating the commodity; the picture / video feedback shows the use effect of the commodity; each element in the user review is weighted and calculated to generate a comprehensive score of the user review;

[0018] In order to analyze the sentiment tendency of the user comment text through natural language processing, and judge whether the comment is positive, neutral or negative, the following steps are used for detailed processing:

[0019] Firstly, the user's comment text needs to be preprocessed so that the subsequent sentiment analysis can be more accurate. The main steps include the following: removing stop words, punctuation marks, URL links, numbers and other irrelevant content in the comment. The comment text is split into single words or phrases for subsequent sentiment analysis. For example, Chinese comments can use jieba and other word segmentation tools, and English comments can use spaCy and others. The words are restored to their original form, such as restoring "run" to "run", and removing meaningless stop words such as "of", "is" and others.

[0020] The sentiment dictionary is a database containing a large number of words or phrases with sentiment tendency, each word is associated with a sentiment polarity score, such as "good" corresponding to a positive score and "bad" corresponding to a negative score. By matching the words in the comment with the entries in the sentiment dictionary, the number and intensity of positive and negative words in the comment are counted. Finally, the sentiment score of the comment is obtained by weighted summation to determine whether the comment is positive, neutral or negative.

[0021] According to the analysis result, the comment is divided into positive, neutral or negative. For example, if most of the words are positive words, the comment is classified as positive; if negative words are dominant, the comment is classified as negative; if positive and negative words are comparable, the comment is classified as neutral.

[0022] In order to further convert the result of sentiment classification into quantifiable score, a weight score can be assigned to each sentiment tendency. For example: positive comment: the score ranges from +1 to +5, which is determined by the intensity of positive sentiment words in the comment. Neutral comment: the score is 0, indicating that the comment has no strong sentiment tendency. Negative comment: the score ranges from -1 to -5, which is determined by the intensity of negative sentiment words in the comment. The final sentiment tendency score can be calculated by weighting and averaging the proportion of positive, neutral and negative comments and their corresponding scores to generate an overall sentiment tendency score as the quantified sentiment score.

[0023] In the social media evaluation, the number of likes reflects the popularity of the product on social media, the number of shares measures the influence of the product, the frequency of mentions represents the degree to which the product is discussed and focused on by users, the topic label is the label and topic used by users when discussing the product to reflect the positioning and popularity of the product on social media, and the sentiment tendency is obtained by natural language processing; Weighted calculation of each element in the social media evaluation generates a comprehensive evaluation score of social media evaluation; According to the user evaluation comprehensive score and the social media evaluation comprehensive score, the user feedback index of the product is generated.

[0024] According to step S300, according to the heat index of the product and the user feedback index of the product, the comprehensive attention index of the product is calculated, and according to the demand of the platform and the merchant, different weights are given to the two indexes; The calculation is carried out through the weighted formula, and the formula is as follows:

[0025] A i =w h *H i +w u *F i ;

[0026] Among them, A i is the attention index of product i, H i is the heat index of product i,

[0027] F i is the user feedback index of product i, w h and w u are the heat index weight and the user feedback index weight, respectively, satisfying w h +w u =1;

[0028] According to the plate to which the commodity belongs, a data set of commodities in the same plate is obtained, and an attention index of commodities in the same plate is generated; let the set of commodities in the same plate be S p The average value of the attention index is calculated by the following formula: Wherein, A p is the attention index of commodities in the same plate, and |S p | is the total number of commodities in the plate.

[0029] According to the category of the commodity, a data set of commodities in the same category is obtained, and an attention index of commodities in the same category is generated; let the set of commodities in the same category be S c The average value of the attention index is calculated by the following formula: Wherein, A c is the attention index of commodities in the same category,

[0030] |S c | is the total number of commodities in the category.

[0031] According to the concept label of the commodity, a data set of commodities in the same concept is obtained, and an attention index of commodities in the same concept is generated; let the set of commodities in the same concept be S g The average value of the attention index can be calculated by the following formula: Wherein, A g is the attention index of commodities in the same concept, and |S g | is the total number of commodities in the concept.

[0032] For comparison of the relative performance of commodities in various dimensions, the relative attention ranking of commodities in their respective plates, categories and concepts is further calculated, wherein the calculation formula of the plate ranking is as follows:

[0033] Wherein, R p (i) represents the relative attention performance of commodity i in its plate, if R p (i) > 1, it means that the attention of commodity i is higher than the average level of the plate, otherwise it is lower than the average level; the calculation formula of the category ranking is as follows: Wherein,

[0034] R c (i) represents the relative attention performance of commodity i in its category, if R c (i) > 1, it means that the attention of commodity i is higher than the average level of the category, otherwise it is lower than the average level; the calculation formula of the concept ranking is as follows: Wherein, R g (i) represents the relative attention performance of commodity i in its concept, if R g (i) > 1, it means that the attention of commodity i is higher than the average level of the concept, otherwise it is lower than the average level.

[0035] According to step S400, a deep neural network is used as an artificial intelligence model for synthesizing attention indexes of different dimensions; the input layer includes three inputs, specifically, the attention index of the same-plateau commodity, the attention index of the same-category commodity, and the attention index of the same-concept commodity; the hidden layer uses multiple levels of neurons to capture the relationship between each dimension, and the neurons of each layer gradually synthesize and extract deep features of the data, so as to compress multi-dimensional information into a more compact representation; the output layer generates a comprehensive attention index standard, representing the overall market performance of the commodity under the plateaus, categories, and concepts.

[0036] The attention indexes of the same-plateau commodity, the same-category commodity, and the same-concept commodity are input into the input layer of the deep neural network, and historical commodity data are collected, including commodity attention indexes under different plateaus, categories, and concepts, for training the model to enable the model to identify the nonlinear relationship between each dimension; the target attention value is generated by labeling the sales performance and market reaction of the historical commodities; the model adjusts the weight parameters by learning the relationship between the target value and the input data, so as to make a prediction on new commodity data; through the hidden layer of the neural network, the model can extract deep features of each dimension, and different hidden layer neurons capture the mutual influence of commodities among plateaus, categories, and concepts by weighted combination of attention indexes of different dimensions; finally, the model generates an attention index standard in the output layer;

[0037] If the actual attention index is higher than the attention index standard, it means that the commodity performs outstandingly, exceeding market expectations; if the actual attention index is lower than the attention index standard, it means that the commodity performs poorly, below market expectations; according to the difference between the standard value of the model and the actual attention index, the commodity attention is generated.

[0038] A digital management system for measuring commodity attention, comprising:

[0039] A data crawling module, comprising: a data input unit, an HTTP request unit, an HTML parsing unit, a content processing unit, and a request frequency control unit; wherein the data input unit inputs the e-commerce website to be crawled and the collected data items, the data items including the browsing volume, the click volume, the purchase volume, the social media evaluation, and the user evaluation of the commodity, the HTTP request unit sends an HTTP request to obtain web page content using the Requests library, the HTML parsing unit parses HTML using the BeautifulSoup library, the content processing unit processes pagination and dynamic content, and the request frequency control unit sets the request interval;

[0040] The index generation module comprises: a click rate and purchase rate calculation unit, a heat index generation unit, a user evaluation processing unit, a social media evaluation processing unit, and a user feedback index generation unit; wherein the click rate and purchase rate calculation unit calculates the click rate and purchase rate of the product according to the data items, the heat index generation unit generates the heat index of the product according to the click rate and the purchase rate, the user evaluation comprises user scores, user comment texts, sentiment tendencies, user identities, user tags, and picture / video feedbacks, the user evaluation processing unit generates a user evaluation comprehensive score according to the user evaluation, the social media evaluation comprises like numbers, share numbers, mention frequencies, topic tags, and sentiment tendencies, the social media evaluation processing unit generates a social media evaluation comprehensive score according to the social media evaluation, and the user feedback index generation unit generates a user feedback index of the product according to the user evaluation comprehensive score and the social media evaluation comprehensive score;

[0041] The product attention index generation module comprises: a data set generation unit and an attention generation unit; wherein the data set generation unit generates the attention index of the product according to the heat index of the product and the user feedback index of the product, obtains the product plate, the product category, and the product concept of the product, obtains the same-plate product data set according to the product plate, obtains the same-category product data set according to the product category, and obtains the same-concept product data set according to the product concept, and the attention generation unit generates the attention index of the same-plate product according to the same-plate product data set, generates the attention index of the same-category product according to the same-category product data set, and generates the attention index of the same-concept product according to the same-concept product data set;

[0042] The artificial intelligence model integration and standardization module comprises: an artificial intelligence model training unit and an attention generation unit; wherein the artificial intelligence model training unit integrates the attention index of the same-plate product, the attention index of the same-category product, and the attention index of the same-concept product by using an artificial intelligence model to form an attention index standard of the product, and the attention generation unit compares the attention index standard of the product with the attention index of the product to generate the product attention.

[0043] Compared with the prior art, the present application has the following beneficial effects:

[0044] 1、The present application performs detailed sentiment analysis and multi-dimensional evaluation processing on user evaluation, not only including scores, but also combining sentiment tendencies of comments, user identities, tags, picture / video feedbacks and other information, so as to more comprehensively and accurately reflect the real feelings and feedbacks of users on products.

[0045] 2、The application can not only evaluate the performance of a single commodity, but also make horizontal comparison of commodities of the same kind, same plate and same concept, and generate their respective attention indexes, by classifying and analyzing the plate, category and concept to which the commodity belongs. This way provides more comprehensive market analysis for merchants, facilitating them to develop more accurate operation strategies.

[0046] 3、The application introduces artificial intelligence technology, integrates the attention indexes of commodities of the same plate, same category and same concept through a deep learning model, forms a standard of commodity attention, and compares it with the actual commodity attention, so as to more accurately evaluate the market performance of the commodity. This dynamic standardization process effectively solves the limitations in the traditional system evaluation process, making the commodity attention evaluation more intelligent and dynamic. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 is a step schematic diagram of a digital management method for measuring commodity attention of the application;

[0048] Figure 2 is a system structure diagram of a digital management system for measuring commodity attention of the application. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0050] Embodiment: As shown in the figure, the application provides a technical solution, Figures 1-2

[0051] According to one embodiment of the application, as shown in the figure, Figure 1 A step schematic diagram of a digital management method for measuring commodity attention shows that a digital management method for measuring commodity attention includes the following steps:

[0052] S100, input the e-commerce website to be crawled and the collected data items, the data items including the browsing volume, click volume, purchase volume, social media evaluation and user evaluation of the commodity, use the Requests library to send HTTP request to obtain web page content, use the BeautifulSoup library to parse HTML, process pagination and dynamic content, and set request interval;

[0053] ​S200: Calculate the click-through rate and purchase rate of the product based on the data items, and generate a popularity index for the product based on the click-through rate and purchase rate; generate a comprehensive user evaluation score based on the user evaluations, including user ratings, user comment text, sentiment, user identity, user tags, and image and video feedback; generate a comprehensive social media evaluation score based on the social media evaluations, including the number of likes, number of shares, frequency of mentions, topic tags, and sentiment; and generate a user feedback index for the product based on the comprehensive user evaluation score and the comprehensive social media evaluation score;

[0054] S300: Generate a product attention index based on the product popularity index and the product user feedback index, obtain the product section, product category, and product concept of the product, obtain a product dataset of the same section based on the product section, obtain a product dataset of the same category based on the product category, and obtain a product dataset of the same concept based on the product concept, generate an attention index for products of the same section based on the product dataset of the same section, generate an attention index for products of the same category based on the product dataset of the same category, and generate an attention index for products of the same concept based on the product dataset of the same concept;

[0055] S400. Use an artificial intelligence model to integrate the attention index of products in the same section, the attention index of products in the same category, and the attention index of products with the same concept to form a product attention index standard, compare the product attention index standard with the product attention index, and generate product attention.

[0056] According to step S100, the user specifies the URL of the target e-commerce website and specifies the data items to be collected, including the number of views, clicks, purchases, social media reviews, and user reviews of the product; a web crawler is used to send HTTP requests to obtain web page content from the server, and the request header simulates browser access to avoid being identified as a robot program; the crawler controls the frequency of requests when sending requests and sets the request interval to prevent frequent requests in a short period of time from burdening the target website or causing the IP address to be banned; after obtaining the web page content, the HTML page is parsed. According to the different structures of the e-commerce website, the crawler program uses the HTML parsing tool BeautifulSoup to extract data items, which are embedded in specific tags of the web page. The crawler finds the specific tags based on the page structure and extracts the text content therein;

[0057] E-commerce websites divide product lists into multiple pages for display. Crawler programs can automatically identify paging links and grab data from each page. This is achieved by parsing the URL of the paging button or dynamic request link, and sending requests to each page in turn to obtain its content.

[0058] E-commerce websites that dynamically render product data using JavaScript asynchronously, the crawler needs to handle dynamic content, simulate browser behavior, load JavaScript, or send API requests to data interfaces to obtain dynamically generated data; if the e-commerce website integrates user interaction data on social media, the crawler collects social media review data for the product by parsing embedded social media information; if the e-commerce website does not integrate user interaction data on social media, the crawler accesses social media platforms to collect social media reviews related to the product through the corresponding API or page scraping;

[0059] E-commerce websites display user reviews, and the crawler locates the display position of user review data, extracts user-provided user ratings, user comment text, sentiment orientation, user identity, user tags, and image / video feedback, and the crawler program categorizes user review data;

[0060] The crawler sets a random time interval after each request to simulate human user behavior and dynamically changes IP addresses using an IP proxy pool.

[0061] According to step S200, based on the obtained browsing volume, click volume, and purchase volume of the product, the click rate and purchase rate are calculated, the click rate is the ratio of click volume to browsing volume, and the purchase rate is the ratio of purchase volume to click volume, the click rate reflects the user's initial interest, and the purchase rate measures the effect of interest conversion into actual purchase, and the heat index of the product is calculated based on the weighted calculation of the click rate and the purchase rate.

[0062] In this embodiment, a smart watch is being sold on an e-commerce platform. To conduct an experiment, we obtained the browsing volume, click volume, purchase volume, and related data of user reviews and social media reviews for the product from the platform. The goal is to calculate the heat index of the product, the comprehensive score of user reviews, and the comprehensive score of social media reviews based on these data, and ultimately generate the user feedback index of the product. The browsing volume is 10,000, the click volume is 2,000, and the purchase volume is 300.

[0063] According to step S200, in the user evaluation, the user score represents the user's rating for the product, presented as a positive integer; the user comment text is obtained, the sentiment tendency of the comment text is analyzed through natural language processing, and it is judged whether the user's comment is positive, neutral or negative, quantified as a score as the sentiment tendency; the user identity is the authentication information of the user on the e-commerce platform; the user label is the characteristic label attached by the user when evaluating the product; the picture / video feedback shows the use effect of the product; the weighted calculation of each element in the user evaluation generates a comprehensive score of the user evaluation; user score: average 4.2 / 5 (based on 500 ratings), user comment text: 500 comments are analyzed through NLP processing, sentiment tendency: positive comments: 60% (300), neutral comments: 20% (100), negative comments: 20% (100), user identity: 80% of users are authenticated users, user label: users often mention "long battery life" and "reasonable price", picture / video feedback: 40% of users upload pictures / videos to show the use effect of the product.

[0064] The click rate and purchase rate of the product are calculated, click rate = clicks / views = 2,000 / 10,000 = 0.2 (20%), purchase rate = purchases / clicks = 300 / 2,000 = 0.15 (15%). The heat index of the product is calculated based on the weighted calculation of the click rate and the purchase rate, and the weights of the click rate and the purchase rate are 0.6 and 0.4 respectively: heat index = (click rate*0.6)+(purchase rate*0.4) = (0.2*0.6)+(0.15*0.4) = 0.12+0.06 = 0.18. The user score is 4.2, which is directly involved in the comprehensive score of the user evaluation. Sentiment tendency: divide the sentiment into positive, neutral and negative, sentiment score: positive: +1, neutral: 0, negative: -1, sentiment score = (300*1+100*0+100*-1) / 500 = (300-100) / 500 = 0.4. User identity: 80% of users are authenticated, the weight of authenticated users is 0.8, the weight of non-authenticated users is 0.2, and the user identity score is: user identity score = 0.8*80%+0.2*20% = 0.64+0.04 = 0.68. Picture / video feedback: 40% of users upload pictures / videos, the weight of picture / video feedback is 0.5, and the picture feedback score is: picture feedback score = 0.5*40% = 0.2.

[0065] The final user evaluation comprehensive score = user score * weight + sentiment score * weight + user identity score * weight + picture feedback score * weight. The weights are 0.4, 0.2, 0.2 and 0.2 respectively, then the user evaluation comprehensive score = (4.2 * 0.4) + (0.4 * 0.2) + (0.68 * 0.2) + (0.2 * 0.2) = 1.68 + 0.08 + 0.136 + 0.04 = 1.936.

[0066] In the social media evaluation, the number of likes reflects the popularity of the product on social media, the number of shares is used to measure the influence of the product, the mention frequency represents the degree to which the product is discussed and focused on by users, the topic label is the label and topic used by users when discussing the product, to reflect the positioning and popularity of the product in social media, and the sentiment tendency is obtained by natural language processing; each element in the social media evaluation is weighted and calculated to generate a social media evaluation comprehensive score; and a user feedback index of the product is generated according to the user evaluation comprehensive score and the social media evaluation comprehensive score. The number of likes: 3,000, the number of shares: 500, the mention frequency: 2,000 times, the topic label: #smart watch, #health monitoring, #exercise tracking, and the social media sentiment analysis: the sentiment tendency is analyzed by NLP as: positive sentiment: 65%, neutral sentiment: 25%, and negative sentiment: 10%. Among them, the weight of likes is 0.4, the weight of shares is 0.3, the weight of mention frequency is 0.2, and the weight of sentiment tendency is 0.1, and the sentiment tendency score = 65% * 1 + 25% * 0 + 10% * (-1) = 0.55.

[0067] The final social media evaluation comprehensive score = (like weight + share weight + mention frequency weight + sentiment analysis weight); then the social media evaluation comprehensive score = (3000 * 0.4) + (500 * 0.3) + (2000 * 0.2) + (0.55 * 0.1) = 1200 + 150 + 400 + 0.055 = 1750.055.

[0068] The user feedback index = user evaluation comprehensive score + social media evaluation comprehensive score, then the user feedback index = 1.936 + 1750.055 = 1751.991.

[0069] According to step S300, according to the heat index of the product and the user feedback index of the product, the comprehensive attention index of the product is calculated, and different weights are given to the two indexes according to the needs of the platform and the merchant; the smart watch belongs to the "electronic product" board, the "smart device" category, and is related to the "health monitoring" concept. In this embodiment, the heat index and user feedback index data of the product, as well as the related data of other products in the board, category and concept are obtained.

[0070] The calculation is performed by a weighting formula, and the formula is as follows:

[0071] A i = w h *H i + w u *F i = 0.4 * 0.18 + 0.6 * 1751.991 = 1051.2666;

[0072] Wherein, A i is the attention index of the commodity i, H i is the heat index of the commodity i,

[0073] F i is the user feedback index of the commodity i, w h and w u are the heat index weight and the user feedback index weight, respectively, and satisfy w h +w u = 1;

[0074] According to the data set of the same plate commodity obtained according to the plate to which the commodity belongs, the attention index of the same plate commodity is generated; let the same plate commodity set be S p , and the average value of the attention index is calculated by the following formula: Wherein, A p is the attention index of the same plate commodity, and |S p | is the total number of commodities in the plate;

[0075] According to the data set of the same category commodity obtained according to the category of the commodity, the attention index of the same category commodity is generated; let the same category commodity set be S c , and the average value of the attention index is calculated by the following formula: Wherein, A c is the attention index of the same category commodity, and |S c | is the total number of commodities in the category;

[0076] According to the data set of the same concept commodity obtained according to the concept label of the commodity, the attention index of the same concept commodity is generated; let the same concept commodity set be S g , and the average value of the attention index can be calculated by the following formula: Wherein, A g is the attention index of the same concept commodity, and |S g | is the total number of commodities in the concept.

[0077] For comparison of the relative performance of commodities in various dimensions, the relative attention ranking of commodities in their respective plates, categories and concepts is further calculated, wherein the calculation formula of the plate ranking is as follows:

[0078] wherein R p (i) represents the relative attention performance of product i in its plate, if R p (i) > 1, it means that the attention of product i is higher than the average level of the plate, otherwise it is lower than the average level, which means that the attention index of smart watch is higher than the average level of the plate; the calculation formula of the ranking in the category is as follows: wherein R c (i) represents the relative attention performance of product i in its category, if R c (i) > 1, it means that the attention of product i is higher than the average level of the category, otherwise it is lower than the average level, which means that the attention index of smart watch is higher than the average level of the category; the calculation formula of the ranking in the concept is as follows: wherein R g (i) represents the relative attention performance of product i in its concept, if R g (i) > 1, it means that the attention of product i is higher than the average level of the concept, otherwise it is lower than the average level, which means that the attention index of smart watch is higher than the average level of the concept.

[0079] According to step S400, a deep neural network is used as an artificial intelligence model for comprehensive attention index in different dimensions; the input layer includes three inputs, specifically the attention index of the same plate product, the attention index of the same category product and the attention index of the same concept product; the hidden layer uses multiple levels of neurons to capture the relationship between each dimension, and the neurons of each layer gradually integrate and extract the deep features of the data, so as to compress the multi-dimensional information into a more compact representation; the output layer generates a comprehensive attention index standard, representing the overall market performance of the product in the plate, category and concept.

[0080] The attention index of the same plate product, the attention index of the same category product and the attention index of the same concept product are input into the input layer of the deep neural network, and historical product data is collected, including product attention index in different plates, categories and concepts, for training the model to enable the model to identify the nonlinear relationship between each dimension; by labeling the sales performance and market reaction of historical products, a target attention value is generated; the model learns the relationship between the target value and the input data to adjust the weight parameters, so as to make predictions on new product data; through the hidden layer of the neural network, the model can extract the deep features of each dimension, and different hidden layer neurons can capture the mutual influence of products among plates, categories and concepts by weighting and combining the attention index of different dimensions; finally, the model generates an attention index standard in the output layer;

[0081] If the actual attention index is higher than the attention index standard, it means that the product performs well, exceeding market expectations; if the actual attention index is lower than the attention index standard, it means that the product performs poorly, lower than market expectations; according to the difference between the standard value of the model and the actual attention index, the product attention is generated.

[0082] A digital management system for measuring product attention, comprising:

[0083] A data crawling module: comprising a data input unit, an HTTP request unit, an HTML parsing unit, a content processing unit and a request frequency control unit; wherein the data input unit inputs the e-commerce website to be crawled and the collected data items, the data items including the browsing volume, click volume, purchase volume, social media evaluation and user evaluation of the product, the HTTP request unit sends HTTP request using the Requests library to obtain web page content, the HTML parsing unit parses HTML using the BeautifulSoup library, the content processing unit processes pagination and dynamic content, and the request frequency control unit sets the request interval;

[0084] An index generation module: comprising a click rate and purchase rate calculation unit, a heat index generation unit, a user evaluation processing unit, a social media evaluation processing unit and a user feedback index generation unit; wherein the click rate and purchase rate calculation unit calculates the click rate and purchase rate of the product according to the data items, the heat index generation unit generates the heat index of the product according to the click rate and purchase rate, the user evaluation includes user rating, user comment text, sentiment orientation, user identity, user label and picture / video feedback, the user evaluation processing unit generates a user evaluation comprehensive score according to the user evaluation, the social media evaluation includes likes, shares, mention frequency, topic labels and sentiment orientation, the social media evaluation processing unit generates a social media evaluation comprehensive score according to the social media evaluation, and the user feedback index generation unit generates the user feedback index of the product according to the user evaluation comprehensive score and the social media evaluation comprehensive score;

[0085] A product attention index generation module: comprising a data set generation unit and an attention generation unit; wherein the data set generation unit generates the attention index of the product according to the heat index of the product and the user feedback index of the product, obtains the product board, product category and product concept of the product, obtains the same-board product data set according to the product board, obtains the same-category product data set according to the product category, and obtains the same-concept product data set according to the product concept, and the attention generation unit generates the attention index of the same-board product according to the same-board product data set, generates the attention index of the same-category product according to the same-category product data set, and generates the attention index of the same-concept product according to the same-concept product data set;

[0086] The artificial intelligence model integration and standardization module comprises an artificial intelligence model training unit and an attention generation unit; wherein the artificial intelligence model training unit uses the artificial intelligence model to integrate the attention index of the same board commodity, the attention index of the same category commodity and the attention index of the same concept commodity to form the attention index standard of the commodity, and the attention generation unit compares the attention index standard of the commodity with the attention index of the commodity to generate the attention degree of the commodity.

[0087] In the embodiment, the model training unit aggregates all the attention indexes into a training data set. The attention indexes of the smart watch and its same board, same category and same concept commodities are included. The attention index of each commodity is extracted as a feature for training, and the target variable is the attention ranking of each commodity. The training data set is input into the model for training to generate an attention index standard model.

[0088] The attention generation unit uses the trained model to predict all commodities to form an attention standard. The attention index of the smart watch is compared with the generated attention standard to generate the attention degree of the commodity. The calculation of the attention degree of the commodity = 1051.27 / 1100 = 0.956, which indicates that the attention degree of the smart watch is slightly lower than the standard generated by the model.

[0089] It is apparent for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, but can be implemented in other concrete forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all aspects as illustrative and not restrictive, and the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and range of the equivalent elements of the claims are intended to be embraced in the present application. Any reference signs in the claims should not be considered as limiting the claims involved.

Claims

1. A digital management method for measuring product attention, characterized in that: The method comprises the following steps: S100: Input the e-commerce website to be crawled and the collected data items, including the number of product views, clicks, purchases, social media reviews, and user reviews. Use the Requests library to send HTTP requests to obtain web page content. Use the BeautifulSoup library to parse HTML, handle paging and dynamic content, and set the request interval. S200: Calculate the click-through rate and purchase rate of the product based on the data items, and generate a popularity index for the product based on the click-through rate and purchase rate; generate a comprehensive user evaluation score based on the user evaluations, including user ratings, user comment text, sentiment, user identity, user tags, and image and video feedback; generate a comprehensive social media evaluation score based on the social media evaluations, including the number of likes, number of shares, frequency of mentions, topic tags, and sentiment; and generate a user feedback index for the product based on the comprehensive user evaluation score and the comprehensive social media evaluation score; S300: Generate a product attention index based on the product popularity index and the product user feedback index, obtain the product section, product category, and product concept of the product, obtain a product dataset of the same section based on the product section, obtain a product dataset of the same category based on the product category, and obtain a product dataset of the same concept based on the product concept, generate an attention index for products of the same section based on the product dataset of the same section, generate an attention index for products of the same category based on the product dataset of the same category, and generate an attention index for products of the same concept based on the product dataset of the same concept; According to step S300, the comprehensive attention index of the product is calculated based on the popularity index of the product and the user feedback index of the product. Different weights are assigned to the two indices based on the needs of the platform and merchants. The calculation is performed using a weighted formula, which is as follows: ； in, is the attention index of product i, is the popularity index of product i, is the user feedback index of product i, and They are the heat index weight and the user feedback index weight, satisfying ; According to the section to which the product belongs, a data set of products in the same section is obtained, and the attention index of products in the same section is generated; let the set of products in the same section be , the average value of its attention index is calculated by the following formula: ,in, It is the attention index of products in the same sector. is the total number of products in the sector; According to the concept labels of the products, a data set of products with the same concept is obtained, and the attention index of products with the same concept is generated; let the set of products with the same concept be , the average value of its attention index can be calculated by the following formula: ,in, It is the attention index of products with the same concept. is the total number of goods within the concept; S400. Use an artificial intelligence model to integrate the attention index of products in the same section, the attention index of products in the same category, and the attention index of products with the same concept to form a product attention index standard, compare the product attention index standard with the product attention index, and generate product attention.

2. A digital management method for measuring product attention according to claim 1, characterized in that: According to step S100, the user specifies the URL of the target e-commerce website and specifies the data items to be collected, including the number of views, clicks, purchases, social media reviews, and user reviews of the product; A web crawler sends HTTP requests to retrieve web page content from a server. The request header simulates a browser access to avoid being identified as a robot. The crawler controls the frequency of requests and sets a request interval to prevent frequent requests in a short period of time from burdening the target website or causing an IP ban. After obtaining the web page content, the crawler parses the HTML page. Based on the different structures of e-commerce websites, the crawler uses the HTML parsing tool BeautifulSoup to extract data items embedded in specific tags on the web page. The crawler finds the specific tags based on the page structure and extracts the text content within them. E-commerce websites divide product lists into multiple pages for display. Crawler programs can automatically identify paging links and grab data from each page. This is achieved by parsing the URL of the paging button or dynamic request link, and sending requests to each page in turn to obtain its content.

3. A digital management method for measuring product attention according to claim 2, characterized in that: For e-commerce websites that use JavaScript to asynchronously load and dynamically present product data, crawlers need to process dynamic content, simulate browser behavior, load JavaScript, or send API requests to data interfaces to obtain dynamically generated data. If the e-commerce website integrates user interaction data from social media, the crawler program collects social media evaluation data for the products by parsing the embedded social media information. If the e-commerce website does not integrate user interaction data from social media, the crawler accesses the social media platform and crawls social media evaluations related to the product through the corresponding API or page. E-commerce websites display user reviews. The crawler locates the display location of the user review data and extracts the user rating, user comment text, sentiment, user identity, user tags, and image and video feedback provided by the user. The crawler program categorizes the user review data. The crawler sets a random time interval after each request to simulate the operation behavior of human users and uses the IP proxy pool to dynamically change the IP address.

4. The digital management method for measuring product attention according to claim 1, characterized in that: According to step S200, based on the obtained number of views, clicks and purchases of the product, the click-through rate and purchase rate are calculated. The click-through rate is the ratio of the number of clicks to the number of views, and the purchase rate is the ratio of the number of purchases to the number of clicks. The click-through rate reflects the user's initial interest, and the purchase rate measures the effect of converting interest into actual purchases. The popularity index of the product is calculated based on the weighted click-through rate and the purchase rate.

5. A digital management method for measuring product attention according to claim 4, characterized in that: According to step S200, in the user evaluation, the user rating represents the user's score for the product and is presented in the form of a positive integer; the user comment text is obtained, and the sentiment tendency of the comment text is analyzed through natural language processing to determine whether the user's comment is positive, neutral or negative, and quantified into a score as the sentiment tendency; the user identity is the user's authentication information on the e-commerce platform; The user tag is a characteristic tag that the user adds when evaluating a product; the picture and video feedback demonstrates the effect of using the product; and the various elements in the user evaluation are weighted to generate a comprehensive user evaluation score. In social media reviews, the number of likes reflects the popularity of the product on social media, the number of shares is used to measure the influence of the product, the frequency of mentions represents the extent to which the product is discussed and paid attention to by users, the topic tags are the tags and topics used by users when discussing the product to reflect the positioning and popularity of the product on social media, and the sentiment tendency is obtained through natural language processing; each element in the social media review is weighted and calculated to generate a comprehensive social media review score; and a user feedback index for the product is generated based on the comprehensive user review score and the comprehensive social media review score.

6. The digital management method for measuring product attention according to claim 1, characterized in that: It is used to compare the relative performance of products in various dimensions and further calculate the relative popularity ranking of products in their respective sectors, categories, and concepts. The calculation formula for the ranking within a sector is as follows: ,in, Represents the relative attention of commodity i in its sector. If , indicating that the attention level of product i is higher than the average level of the sector, and vice versa. The calculation formula for the ranking within the category is as follows: ,in, Represents the relative attention of product i in its category. If , indicating that the attention level of product i is higher than the category average, and vice versa. The calculation formula for the ranking within a concept is as follows: ,in, Represents the relative attention of product i in its concept. If , indicating that the attention level of product i is higher than the average level of the concept, and vice versa.

7. The digital management method for measuring product attention according to claim 1, characterized in that: According to step S400, a deep neural network is used as an artificial intelligence model to integrate the attention index of different dimensions; the input layer includes three inputs, specifically the attention index of products in the same section, the attention index of products in the same category, and the attention index of products with the same concept; the hidden layer uses multiple layers of neurons to capture the relationship between the various dimensions. The neurons in each layer gradually integrate and extract the deep features of the data, thereby compressing the multi-dimensional information into a more compact representation; The output layer generates a comprehensive attention index standard, which represents the overall market performance of the product under the sector, category and concept.

8. The digital management method for measuring product attention according to claim 7, characterized in that: The attention index of products in the same section, the attention index of products in the same category, and the attention index of products with the same concept are input into the input layer of the deep neural network. Historical product data, including the attention index of products in different sections, categories, and concepts, is collected for model training, enabling the model to identify nonlinear relationships between various dimensions. The target attention value is generated by annotating the sales performance and market response of historical products. The model adjusts the weight parameters by learning the relationship between the target value and the input data, and thus makes predictions on new product data. Through the hidden layer of the neural network, the model can extract deep features of each dimension. Different hidden layer neurons gradually capture the mutual influence of products between sections, categories, and concepts by weighted combination of attention indexes of different dimensions. Finally, the model generates an attention index standard at the output layer. If the actual attention index is higher than the attention index standard, it means that the product performs well and exceeds market expectations; if the actual attention index is lower than the attention index standard, it means that the product performs poorly and is lower than market expectations; the product attention is generated based on the difference between the standard value of the model and the actual attention index.

9. A digital management system for measuring product attention, using a digital management method for measuring product attention according to any one of claims 1 to 8, characterized in that: include: The data crawling module includes a data input unit, an HTTP request unit, an HTML parsing unit, a content processing unit, and a request frequency control unit. The data input unit inputs the e-commerce website to be crawled and the collected data items, such as the number of views, clicks, purchases, social media reviews, and user reviews of the product. The HTTP request unit uses the Requests library to send HTTP requests to obtain web page content. The HTML parsing unit uses the BeautifulSoup library to parse HTML. The content processing unit handles paging and dynamic content. The request frequency control unit sets the request interval. Index generation module: including: a click-through rate and purchase rate calculation unit, a popularity index generation unit, a user evaluation processing unit, a social media evaluation processing unit and a user feedback index generation unit; wherein the click-through rate and purchase rate calculation unit calculates the click-through rate and purchase rate of the product based on the data items, the popularity index generation unit generates the popularity index of the product based on the click-through rate and purchase rate, the user evaluation includes user ratings, user comment text, emotional tendency, user identity, user tags and picture and video feedback, the user evaluation processing unit generates a user evaluation comprehensive score based on the user evaluation, the social media evaluation includes the number of likes, the number of shares, the mention frequency, the topic tag and the emotional tendency, the social media evaluation processing unit generates a social media evaluation comprehensive score based on the social media evaluation, and the user feedback index generation unit generates a user feedback index for the product based on the user evaluation comprehensive score and the social media evaluation comprehensive score; The product attention index generation module includes: a data set generation unit and an attention generation unit; wherein the data set generation unit generates the product attention index based on the product's popularity index and the product's user feedback index, obtains the product section, product category and product concept of the product, obtains a data set of products in the same section based on the product section, obtains a data set of products in the same category based on the product category, and obtains a data set of products in the same concept based on the product concept; the attention generation unit generates the attention index of products in the same section based on the product data set of products in the same section, generates the attention index of products in the same category based on the product data set of products in the same category, and generates the attention index of products in the same concept based on the product data set of products in the same concept; Artificial intelligence model integration and standardization module: includes: artificial intelligence model training unit and attention generation unit; among them, the artificial intelligence model training unit uses the artificial intelligence model to integrate the attention index of products in the same section, the attention index of products in the same category, and the attention index of products with the same concept to form the attention index standard of the products. The attention generation unit compares the attention index standard of the products with the attention index of the products to generate the attention of the products.

Images