A data analysis method and system for e-commerce live shopping

Abstract

The present invention discloses a data analysis method and system for live shopping of an e-commerce company. In the present invention, the method comprises: receiving a verification request and an initial bonus from a target anchor sent by a willing seller, comparing the request with a pre-stored blacklist, performing a potential risk assessment on the target anchor according to live record data, establishing a relationship map for the target anchor with a potential risk tag, judging whether there is an abnormality according to the relationship map, establishing a monitoring model for the authenticity of live traffic, and establishing a first judgment standard and a second judgment standard for anchors with potential risk tags and dangerous tags to monitor live broadcasts, and if false traffic is monitored, adding the target anchor to the blacklist, and sending a warning report to the willing seller. The present invention realizes efficient and accurate identification and response to false traffic, and improves the credibility and user satisfaction of the live shopping platform.

Description

[Technical field]

[0001] The present invention relates to the field of e-commerce live broadcast technology, and in particular to a data analysis method and system for e-commerce live broadcast shopping. [Background technology]

[0002] The rapid development of the e-commerce sector has changed the traditional way of selling and marketing goods. With the rise of short videos and live broadcast platforms, more and more manufacturers and sellers choose to use the power of the Internet and social media to promote and sell their products. Among them, cooperating with live broadcast anchors and product bloggers has become a common way. These anchors attract a large number of viewers' attention and purchases by showing and promoting various products to the audience during live broadcasts.

[0003] However, some dishonest anchors resort to false means, such as using teams, purchasing followers or adopting automated scripts to increase the number of live broadcast viewers and buyers, in order to create false attention and purchasing booms during the live broadcast, misleading sellers into believing that there are a large number of real purchases, falsifying purchase transactions, and creating false purchase records, thereby showing false sales data to manufacturers and sellers in order to obtain higher commissions or cooperation fees. After the live broadcast, the false viewers will initiate large-scale return requests to cancel previous purchases, causing sellers to not only lose sales but also bear the cost of returns, which seriously damages the sellers' experience interests, as well as their reputation and credibility, and reduces their competitiveness on e-commerce platforms. [Summary of the invention]

[0004] In view of this, an embodiment of the present invention provides a data analysis method and system for e-commerce live shopping.

[0005] In a first aspect, an embodiment of the present invention provides a data analysis method for e-commerce live shopping, the method comprising:

[0006] Receive verification requests and initial bonuses from willing sellers, and compare them with the pre-stored blacklist. If there is no match, obtain live broadcast record data of the target anchor shared by other sellers in response to the request;

[0007] The target anchor is evaluated for potential risks based on the live broadcast record data. If the evaluation shows that there is no potential risk, the target anchor is monitored for public opinion in the recent preset time period. If negative public opinion related to the target anchor is monitored, the target anchor is given a potential risk label, otherwise the target anchor is given a potential safety label; if the evaluation shows that there is a potential risk, the target anchor is given a potential risk label, and a bonus of a preset proportion of the initial bonus is paid to the seller who shared the target anchor's live broadcast record data according to the preset smart contract;

[0008] A relationship map is established for the target anchor with the potential risk tag, and whether there is an anomaly is determined based on the relationship map. If an anomaly is determined, the target anchor's potential risk tag is replaced with a dangerous tag, and the judgment result is sent to the willing seller;

[0009] A live broadcast traffic authenticity monitoring model is established, and the first and second judgment criteria are respectively established for potential risk label and dangerous label anchors for live broadcast monitoring. If false traffic is monitored, the target anchor will be added to the blacklist and a warning report will be sent to the willing sellers.

[0010] According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein the method further includes:

[0011] Establishing a bonus pool that is bound to the seller's account and receives the initial bonus, wherein the bonus pool is set with read-only permissions, allowing only authenticated sellers to view the bonus pool balance and transaction records;

[0012] Compare the target anchor's identity information with the pre-stored blacklist;

[0013] If the target anchor's identity information does not match any record in the blacklist, the initial bonus will be pre-deposited into the bonus pool, and a response request for the target anchor's live broadcast record data will be initiated to other sellers;

[0014] If the identity information of the target anchor matches the record in the blacklist, the initial bonus will be deducted by the first preset proportion of system service points, the remaining bonus will be returned to the willing seller, and the blacklist data of the target anchor will be sent.

[0015] According to the above aspects and any possible implementation, there is further provided an implementation, wherein the live broadcast record data includes audience interaction data, product sales data, audience quality data, and interaction difference data between buyers and non-buyers;

[0016] Conduct potential risk assessment on the target anchor based on live broadcast record data, including:

[0017] Calculate the average potential risk assessment value The average potential risk assessment value The calculation formula is as follows:

[0018]

[0019] Where N represents the number of live recorded data, R i represents the potential risk assessment value of the i-th live broadcast record data;

[0020]

[0021] Among them, R represents the potential risk assessment value, I A An indicator representing product sales data, I B Indicator representing audience interaction records, I C An indicator of audience quality, I D An indicator of the difference in interactions between buyers and non-buyers, w A 、w B 、w C and w D Respectively represent weights;

[0022]

[0023] I B =α B C+β B ·L+γ B ·S,

[0024]

[0025]

[0026] Among them, M * represents the total number of products purchased by viewers watching the live broadcast, M0 represents the number of repeat purchases by viewers, Q represents product sales, Z represents the return rate, C represents the number of comments during the live broadcast, L represents the number of likes during the live broadcast, S represents the number of shares during the live broadcast, n represents the number of viewers, and x represents the number of i represents the longitude or latitude of the i-th viewer coordinates, represents the mean of the longitude or latitude of the viewer’s coordinates, Md represents the median of the viewer’s interactions, M1 represents the number of interactions by buyers, M2 represents the number of interactions by non-buyers, and α A , β A , α B , β B , γ B , α C and β C represent the regulating factors respectively;

[0027] Comparison of average potential risk assessment values Compared with the preset risk threshold R0, when the average potential risk assessment value When it is greater than the preset risk threshold R0, the target anchor is deemed to be at potential risk.

[0028] According to the above aspects and any possible implementation, an implementation is further provided, wherein the public opinion monitoring of the target anchor in the recent preset time period is performed, and if negative public opinion related to the target anchor is monitored, the target anchor is assigned a potential risk label, otherwise the target anchor is assigned a potential safety label, specifically including:

[0029] Collect the target anchor’s name and keywords, and set a monitoring time window;

[0030] Use the name and keywords of the target anchor to conduct real-time searches on news media and social media to obtain public opinions related to the target anchor within the monitoring time window, including news articles, social media posts, comments, and discussions;

[0031] Conduct sentiment analysis on public opinion content and classify it into three types of sentiment polarity: positive, negative, or neutral.

[0032] The sentiment analysis method is as follows: converting the acquired public opinions into text data and preprocessing the text, splitting the text into words or word sequences, using a predefined sentiment vocabulary, assigning a sentiment score to each word in the text, increasing the score for positive words and decreasing the score for negative words, and then calculating the summary score to classify the sentiment polarity of the text;

[0033] If the public opinion is marked as negative sentiment and is directly related to the target anchor, it will be recorded as negative public opinion;

[0034] Conduct content analysis on negative public opinions to determine whether they are related to the target anchor’s behavior, products, or speeches. If so, include the content of the public opinions in the risk assessment;

[0035] Count the number of negative public opinions within a preset time period. If the number exceeds the preset threshold, it is judged that there is a potential risk.

[0036] The target anchor is assigned a potential risk label, otherwise the target anchor is assigned a potential safety label.

[0037] According to the above aspects and any possible implementation, an implementation is further provided, wherein the smart contract preset method is as follows:

[0038] Deploy a smart contract to enable all sellers to participate;

[0039] Generate a unique contract address, which is used for interaction;

[0040] The execution conditions set in the smart contract are: if there is no potential risk in the assessment, the initial bonus will be deducted by a second preset proportion of system service points, and the remaining bonus will be returned to the willing seller;

[0041] If the assessment shows potential risks, a bonus equal to a preset proportion of the initial bonus will be paid to the seller who shared the live broadcast record data of the target anchor, and a third preset proportion of system service points will be deducted from the initial bonus.

[0042] According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein the third preset ratio is greater than the second preset ratio, and the second preset ratio is greater than the first preset ratio.

[0043] According to the above aspects and any possible implementation, an implementation is further provided, wherein a relationship map is established for the target anchor with the potential risk tag, and whether there is an abnormality is determined according to the relationship map. If an abnormality is determined, the potential risk tag of the target anchor is replaced with a dangerous tag, and the judgment result is sent to the willing seller, specifically including:

[0044] Collect the target anchor's fan attention data, and obtain the attention relationship between the target anchor and fans, and between fans and fans;

[0045] According to the attention relationship, the target anchor is used as the root node and the fans are used as child nodes to connect in a tree structure, so as to obtain a relationship map represented by a tree structure;

[0046] Use anomaly detection algorithms to detect whether there are anomalies in the relationship graph, and / or,

[0047] Collect anchor relationship graph data for training, label and classify them as normal or abnormal, pre-process the relationship graph, convert the relationship graph into an adjacency matrix or node feature matrix, build a convolutional neural network model for abnormality judgment, train the convolutional neural network model with labeled training data, and use the trained convolutional neural network to conduct in-depth analysis of the anchor relationship graph after evaluation;

[0048] If an abnormality is found, the target anchor's potential risk label will be replaced with a dangerous label, and the judgment result will be sent to the willing seller.

[0049] According to the above aspects and any possible implementation, an implementation is further provided, wherein the establishment of a live broadcast traffic authenticity monitoring model comprises:

[0050] Collect live broadcast data, clean and preprocess it, and handle missing values ​​and outliers;

[0051] Construct the characteristic variables of the aggregation model and calculate the aggregation degree G, where the calculation formula of the aggregation degree G is as follows:

[0052]

[0053] A1=N1·T1,

[0054] A2=H2·ΔP2,

[0055]

[0056] Among them, A1 represents the anchor's attractiveness, A2 represents the product's attractiveness, D1 represents the audience dispersion, F1 represents the number of audience interactions per minute, S1 represents the number of new viewers entering per minute, S2 represents the number of product orders placed per minute, w1 and w2 represent the adjustment coefficients respectively; N1 represents the number of fans of the anchor, T1 represents the cumulative live broadcast time of the anchor, H2 represents the weighted comprehensive value of the product's views, collections, and purchases, ΔP2 represents the preferential price of the product, and K i represents the proportion of the audience in the ith region to the total audience, and n represents the number of regions;

[0057] When the live broadcast reaches the preset time T1, the first live broadcast traffic authenticity monitoring is started for the first preset time period ΔT2 thereafter, and a concentration G growth rate comparison model for the first preset time period ΔT2 is constructed, the first comparison average growth rate R1 is calculated, and the real average growth rate R2 of the concentration G for the first preset time period ΔT2 is calculated, and the real average growth rate R2 is compared with the first comparison average growth rate R1, wherein the concentration G growth rate comparison model for the first preset time period ΔT2 is:

[0058] Wherein, α1 and b1 represent the set growth rate curve parameters, and t1 represents the time of the first preset time period ΔT2;

[0059] After the first preset time period ΔT2, when it is detected that the concentration G continues to decrease for a preset time period ΔT3, the second live broadcast traffic authenticity monitoring is started for the second preset time period ΔT4 thereafter, and a concentration G reduction rate comparison model for the second preset time period ΔT4 is constructed, and the second comparison average reduction rate R3 is calculated. The real average reduction rate R4 of the concentration G in the second preset time period ΔT2 is calculated, and the real average reduction rate R4 is compared with the second comparison average reduction rate R3. The concentration G reduction rate comparison model for the second preset time period ΔT4 is:

[0060] Among them, α2 and b2 represent the set reduction rate curve parameters, t2 represents the time of the second preset time period ΔT4;

[0061] When R2>R1 and R4<R3, the monitored live traffic is deemed to be false traffic.

[0062] According to the above aspects and any possible implementation, an implementation is further provided, wherein the first judgment standard and the second judgment standard are respectively established for the anchors with potential risk labels and dangerous labels to monitor the live broadcast, specifically including:

[0063] A first judgment standard is established for potential risk label anchors. The first judgment standard adopts a first preset time period ΔT2 aggregation degree G growth rate comparison model as follows: The comparison model of the reduction rate of the aggregation degree G in the second preset time period ΔT4 is:

[0064] A second judgment standard is established for the dangerous label anchor, and the second judgment standard adopts the adjusted first preset time period ΔT2 aggregation degree G growth rate comparison model as follows: The comparison model of the reduction rate of the aggregation degree G in the second preset time period ΔT4 is:

[0065] Set α1>α1 * , b1>b1 * , α2<α2 * , b2>b2 * .

[0066] In a second aspect, an embodiment of the present invention provides a data analysis system for e-commerce live shopping using the data analysis method for e-commerce live shopping, the system comprising:

[0067] The request management module is used to receive the verification request and initial bonus of the target anchor sent by the willing seller, and compare them with the pre-stored blacklist. If there is no match, the live broadcast record data of the target anchor shared by other sellers in response to the request is obtained;

[0068] The risk assessment module is used to conduct a potential risk assessment on the target anchor based on the live broadcast record data. If the assessment shows that there is no potential risk, the target anchor is monitored for public opinion in the recent preset time period. If negative public opinion related to the target anchor is monitored, the target anchor is assigned a potential risk label, otherwise the target anchor is assigned a potential safety label; if the assessment shows that there is a potential risk, the target anchor is assigned a potential risk label, and a bonus of a preset proportion of the initial bonus is paid to the seller who shared the target anchor's live broadcast record data according to the preset smart contract;

[0069] The anomaly monitoring module is used to establish a relationship map for the target anchors with potential risk labels, and use a convolutional neural network to conduct in-depth analysis of the anchors' relationship map to detect whether there are anomalies. If an anomaly is detected, the target anchor's potential risk label is replaced with a dangerous label, and the detection result is sent to the willing sellers;

[0070] The live broadcast monitoring module is used to establish a live broadcast traffic authenticity monitoring model, and establish the first judgment standard and the second judgment standard for potential risk label and dangerous label anchors for live broadcast monitoring. If false traffic is monitored, the target anchor will be added to the blacklist and a warning report will be sent to the willing sellers.

[0071] One of the above technical solutions has the following beneficial effects:

[0072] The method of the embodiment of the present invention proposes a data analysis method for e-commerce live shopping, which has significant technical effects in improving the risk prevention ability of the live broadcast platform, improving user experience, and increasing the credibility of the platform.

Brief Description of the Drawings

[0073] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. 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.

[0074] Figure 1 It is a flow chart of a data analysis method for live e-commerce shopping provided by an embodiment of the present invention;

[0075] Figure 2 This is a functional block diagram of the data analysis system for e-commerce live shopping provided by an embodiment of the present invention. [Specific implementation method]

[0076] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0077] Please refer to Figure 1 , which is a flow chart of a data analysis method for e-commerce live shopping provided by an embodiment of the present invention. Figure 1 As shown, the method includes:

[0078] Receive verification requests and initial bonuses from willing sellers, and compare them with the pre-stored blacklist. If there is no match, obtain live broadcast record data of the target anchor shared by other sellers in response to the request;

[0079] The target anchor is evaluated for potential risks based on the live broadcast record data. If the evaluation shows that there is no potential risk, the target anchor is monitored for public opinion in the recent preset time period. If negative public opinion related to the target anchor is monitored, the target anchor is given a potential risk label, otherwise the target anchor is given a potential safety label; if the evaluation shows that there is a potential risk, the target anchor is given a potential risk label, and a bonus of a preset proportion of the initial bonus is paid to the seller who shared the target anchor's live broadcast record data according to the preset smart contract;

[0080] A relationship map is established for the target anchor with the potential risk tag, and whether there is an anomaly is determined based on the relationship map. If an anomaly is determined, the target anchor's potential risk tag is replaced with a dangerous tag, and the judgment result is sent to the willing seller;

[0081] A live broadcast traffic authenticity monitoring model is established, and the first and second judgment criteria are respectively established for potential risk label and dangerous label anchors for live broadcast monitoring. If false traffic is monitored, the target anchor will be added to the blacklist and a warning report will be sent to the willing sellers.

[0082] The data analysis method for e-commerce live shopping of the present invention can accurately identify false traffic by comprehensively analyzing the live broadcast record data of the target anchor and evaluating the potential risks, and combining public opinion monitoring, including using public opinion data to detect negative public opinion to avoid being misled by false information; adopts a multi-dimensional evaluation system, including potential risk labels, relationship map abnormality judgment, live broadcast traffic authenticity monitoring model, etc., to evaluate the anchor's reputation from different angles, thereby improving accuracy and comprehensiveness; with the help of smart contract technology, automatically execute corresponding operations according to the evaluation results, including bonus payment, label replacement, blacklist addition, etc., to realize intelligent risk response strategies; when false traffic is monitored, not only the target anchor is added to the blacklist, but also a warning report can be sent to the willing seller in time to protect the interests of the willing seller and improve the credibility and user satisfaction of the platform; by establishing a relationship map and analyzing the association between the anchors, it can more comprehensively judge whether there is abnormal behavior, thereby enhancing the monitoring ability of false traffic; establish a live broadcast traffic authenticity monitoring model, which can monitor false traffic, so that a warning can be sent to the willing seller.

[0083] Therefore, the e-commerce live shopping data analysis method provided by the present invention has significant technical effects in improving the risk prevention ability of the live broadcast platform, improving user experience, and increasing the credibility of the platform.

[0084] In a preferred embodiment of the present invention, the method of the present invention further comprises:

[0085] Establishing a bonus pool that is bound to the seller's account and receives the initial bonus, wherein the bonus pool is set with read-only permissions, allowing only authenticated sellers to view the bonus pool balance and transaction records;

[0086] Compare the target anchor's identity information with the pre-stored blacklist;

[0087] If the target anchor's identity information does not match any record in the blacklist, the initial bonus will be pre-deposited into the bonus pool, and a response request for the target anchor's live broadcast record data will be initiated to other sellers;

[0088] If the identity information of the target anchor matches the record in the blacklist, the initial bonus will be deducted by the first preset proportion of system service points, the remaining bonus will be returned to the willing seller, and the blacklist data of the target anchor will be sent.

[0089] The present invention ensures the security of the bonus pool by binding with the seller's account and setting read-only permissions. Only sellers who have passed identity authentication can view the balance and transaction records of the bonus pool, thereby ensuring the safe storage and circulation of the bonus; the identity information of the target anchor is compared with the pre-stored blacklist to realize intelligent identity authentication. When the identity information of the target anchor matches the record in the blacklist, the system can automatically process the refund and send the blacklist data, thereby improving the security and stability of the platform; when the identity information of the target anchor does not match the blacklist, the initial bonus is pre-stored in the bonus pool, and a response request is initiated to other sellers. This mechanism ensures the safe storage of the seller's initial bonus, effectively avoids transactions with potential risk anchors, and improves the security of the platform; during the identity authentication process of the target anchor, the system can notify the relevant sellers in real time, thereby ensuring the timeliness of information transmission. At the same time, for the case of matching the blacklist, the blacklist data of the target anchor is sent in time, thereby improving the platform's ability to respond quickly to potential risks.

[0090] In a preferred embodiment of the present invention, the live broadcast record data of the present invention includes audience interaction data, product sales data, audience quality data, and interaction difference data between buyers and non-buyers;

[0091] Conduct potential risk assessment on the target anchor based on live broadcast record data, including:

[0092] Calculate the average potential risk assessment value The average potential risk assessment value The calculation formula is as follows:

[0093]

[0094] Where N represents the number of live recorded data, R i represents the potential risk assessment value of the i-th live broadcast record data;

[0095]

[0096] Among them, R represents the potential risk assessment value, I A An indicator representing product sales data, I B Indicator representing audience interaction records, I C An indicator of audience quality, I D An indicator of the difference in interactions between buyers and non-buyers, w A 、w B 、w Cand w D Respectively represent weights;

[0097]

[0098] I B =α B C+β B ·L+γ B ·S,

[0099]

[0100]

[0101] Among them, M * represents the total number of products purchased by viewers watching the live broadcast, M0 represents the number of repeat purchases by viewers, Q represents product sales, Z represents the return rate, C represents the number of comments during the live broadcast, L represents the number of likes during the live broadcast, S represents the number of shares during the live broadcast, n represents the number of viewers, and x represents the number of i represents the longitude or latitude of the i-th viewer coordinates, represents the mean of the longitude or latitude of the viewer’s coordinates, Md represents the median of the viewer’s interactions, M1 represents the number of interactions by buyers, M2 represents the number of interactions by non-buyers, and α A , β A , α B , β B , γ B , α C and β C represent the regulating factors respectively;

[0102] Comparison of average potential risk assessment values Compared with the preset risk threshold R0, when the average potential risk assessment value When it is greater than the preset risk threshold R0, the target anchor is deemed to be at potential risk.

[0103] It should be noted that the calculation method of Md is as follows: collect the data of audience interaction, including the number of comments, the number of likes, and the number of shares, store the number of interactions of all audiences in a list, sort the list of interaction numbers in ascending or descending order to find the middle value. If the length of the list is an odd number, the median is the value in the middle of the list. If the length of the list is an even number, the median is the average of the two middle values. It should be noted that the median of audience interaction is used to represent the overall interaction frequency of the audience. The median is usually more robust than the average and can avoid the influence of extreme values ​​on the results.

[0104] By comparing the average potential risk assessment value with the preset risk threshold, the present invention can judge the potential risk of the target anchor in real time. The real-time nature ensures timely monitoring and processing of anchor risks, enables the platform to respond quickly, and ensures the security and stability of transactions.

[0105] In a preferred embodiment of the present invention, the present invention monitors the public opinion of the target anchor in the recent preset time period. If negative public opinion related to the target anchor is monitored, the target anchor is assigned a potential risk label, otherwise the target anchor is assigned a potential safety label, specifically including:

[0106] Collect the target anchor’s name and keywords, and set a monitoring time window;

[0107] Use the name and keywords of the target anchor to conduct real-time searches on news media and social media to obtain public opinions related to the target anchor within the monitoring time window, including news articles, social media posts, comments, and discussions;

[0108] Conduct sentiment analysis on public opinion content and classify it into three types of sentiment polarity: positive, negative, or neutral.

[0109] The sentiment analysis method is as follows: converting the acquired public opinions into text data and preprocessing the text, splitting the text into words or word sequences, using a predefined sentiment vocabulary, assigning a sentiment score to each word in the text, increasing the score for positive words and decreasing the score for negative words, and then calculating the summary score to classify the sentiment polarity of the text;

[0110] If the public opinion is marked as negative sentiment and is directly related to the target anchor, it will be recorded as negative public opinion;

[0111] Conduct content analysis on negative public opinions to determine whether they are related to the target anchor’s behavior, products, or speeches. If so, include the content of the public opinions in the risk assessment;

[0112] Count the number of negative public opinions within a preset time period. If the number exceeds the preset threshold, it is judged that there is a potential risk.

[0113] The target anchor is assigned a potential risk label, otherwise the target anchor is assigned a potential safety label.

[0114] It should be noted that content analysis of negative public opinion can adopt existing methods or the following method: first, it is necessary to collect text data such as news articles, social media posts, comments, etc. that contain negative public opinion. These data can be obtained through web crawlers, API queries or data providers; then, the collected text data is cleaned and preprocessed, including removing special characters, punctuation marks, HTML tags, stop words, etc.; the text is split into sequences of words or terms; through keyword extraction technology, keywords or phrases related to the target anchor in the text are identified and extracted. These keywords can include the name of the target anchor, product name, live event name, etc.; text mining and natural language processing technology are used to analyze the text content to determine whether there is information related to the target anchor; a relevance score is assigned to each text record to indicate its relevance to the target anchor, which can be a binary classification or a score based on a score; the relevance score of the text record is output as relevant or irrelevant, and recorded for further analysis or visualization.

[0115] The present invention covers multiple information sources, including news articles, social media posts, comments and discussions, by real-time searching for public opinions related to the target anchor on news media and social media. Multi-source public opinion collection ensures the comprehensiveness and accuracy of information; sentiment analysis method is used to classify public opinions by sentiment polarity, text data is preprocessed, and a predefined sentiment vocabulary is used to assign sentiment scores to each word in the text. The sentiment analysis method based on the vocabulary can accurately judge the sentiment polarity of public opinions, thereby improving the accuracy of analysis; when determining whether public opinions are related to the target anchor, content analysis is performed to determine whether the public opinion content is directly related to the anchor's behavior, product or speech; correlation judgment ensures that only negative public opinions related to the anchor are included in the risk assessment, thereby avoiding the influence of irrelevant factors on the assessment results; a preset quantity threshold is set, and when the number of negative public opinions exceeds the threshold within a preset time period, it is judged that the anchor has potential risks. This threshold setting provides a clear standard for evaluation, making the evaluation more quantitative and operational; real-time search and monitoring methods are used to quickly obtain relevant information after public opinions are generated, and timeliness ensures the timely discovery and processing of potential risks of the anchor, thereby ensuring the security of platform transactions. The public opinion monitoring method of the present invention combines multi-source public opinion collection, sentiment analysis, correlation judgment and quantity threshold setting, and has the characteristics of high accuracy, real-time and timeliness. It provides reliable data support for the potential risk assessment of the target anchor and ensures the security of the trading platform.

[0116] In a preferred embodiment of the present invention, the smart contract preset method of the present invention is as follows:

[0117] Deploy a smart contract to enable all sellers to participate;

[0118] Generate a unique contract address, which is used for interaction;

[0119] The execution conditions set in the smart contract are: if there is no potential risk in the assessment, the initial bonus will be deducted by a second preset proportion of system service points, and the remaining bonus will be returned to the willing seller;

[0120] If the assessment shows potential risks, a bonus equal to a preset proportion of the initial bonus will be paid to the seller who shared the live broadcast record data of the target anchor, and a third preset proportion of system service points will be deducted from the initial bonus.

[0121] In a preferred embodiment of the present invention, the third preset ratio of the present invention is greater than the second preset ratio, and the second preset ratio is greater than the first preset ratio.

[0122] In a preferred embodiment of the present invention, the present invention establishes a relationship map for the target anchor with a potential risk tag, and determines whether there is an abnormality based on the relationship map. If an abnormality is determined, the target anchor's potential risk tag is replaced with a dangerous tag, and the determination result is sent to the willing seller, specifically including:

[0123] Collect the target anchor's fan attention data, and obtain the attention relationship between the target anchor and fans, and between fans and fans;

[0124] According to the attention relationship, the target anchor is used as the root node and the fans are used as child nodes to connect in a tree structure, so as to obtain a relationship map represented by a tree structure;

[0125] Use anomaly detection algorithms to detect whether there are anomalies in the relationship graph, and / or,

[0126] Collect anchor relationship graph data for training, label and classify them as normal or abnormal, pre-process the relationship graph, convert the relationship graph into an adjacency matrix or node feature matrix, build a convolutional neural network model for abnormality judgment, train the convolutional neural network model with labeled training data, and use the trained convolutional neural network to conduct in-depth analysis of the anchor relationship graph after evaluation;

[0127] If an abnormality is found, the target anchor's potential risk label will be replaced with a dangerous label, and the judgment result will be sent to the willing seller.

[0128] The present invention collects the fan attention data of the target anchor, including the attention relationship between the anchor and the fans, to form a comprehensive data set, which ensures that the data source of the relationship map is sufficient and improves the accuracy of the relationship analysis; based on the attention relationship, a relationship map with a tree structure is constructed, so that the structure of the relationship map is clear, which is convenient for subsequent analysis and judgment, and the representation of the tree structure helps to intuitively observe the relationship between the anchor and the fans; using anomaly detection algorithm and convolutional neural network model, the relationship map is deeply analyzed, and deep learning technology is used to train and evaluate the anchor relationship map, which improves the precision and accuracy of the relationship analysis. The combination of anomaly detection and deep analysis makes the judgment of relationship map anomalies more accurate; once the relationship map anomaly is detected, the potential risk label of the target anchor is replaced with a dangerous label, and the judgment result is sent to the willing seller in time. This real-time anomaly monitoring and response measure ensures the immediate security of the platform and reduces the possibility of development of potential risks.

[0129] The invented method for establishing potential risk labels and analyzing relationship graphs combines comprehensive data collection, tree-structured relationship graph construction, anomaly detection and in-depth analysis, ensuring the accuracy and real-time nature of relationship graph analysis, providing strong support for potential risk assessment of target anchors and ensuring the security of the trading platform.

[0130] In a preferred embodiment of the present invention, the present invention establishes a live broadcast traffic authenticity monitoring model, which includes:

[0131] Collect live broadcast data, clean and preprocess it, and handle missing values ​​and outliers;

[0132] Construct the characteristic variables of the aggregation model and calculate the aggregation degree G, where the calculation formula of the aggregation degree G is as follows:

[0133]

[0134] A1=N1·T1,

[0135] A2=H2·ΔP2,

[0136]

[0137] Among them, A1 represents the anchor's attractiveness, A2 represents the product's attractiveness, D1 represents the audience dispersion, F1 represents the number of audience interactions per minute, S1 represents the number of new viewers entering per minute, S2 represents the number of product orders placed per minute, w1 and w2 represent the adjustment coefficients respectively; N1 represents the number of fans of the anchor, T1 represents the cumulative live broadcast time of the anchor, H2 represents the weighted comprehensive value of the product's views, collections, and purchases, ΔP2 represents the preferential price of the product, and K i represents the proportion of the audience in the ith region to the total audience, and n represents the number of regions;

[0138] When the live broadcast reaches the preset time T1, the first live broadcast traffic authenticity monitoring is started for the first preset time period ΔT2 thereafter, and a concentration G growth rate comparison model for the first preset time period ΔT2 is constructed, the first comparison average growth rate R1 is calculated, and the real average growth rate R2 of the concentration G for the first preset time period ΔT2 is calculated, and the real average growth rate R2 is compared with the first comparison average growth rate R1, wherein the concentration G growth rate comparison model for the first preset time period ΔT2 is:

[0139] Wherein, α1 and b1 represent the set growth rate curve parameters, and t1 represents the time of the first preset time period ΔT2;

[0140] After the first preset time period ΔT2, when it is detected that the concentration G continues to decrease for a preset time period ΔT3, the second live broadcast traffic authenticity monitoring is started for the second preset time period ΔT4 thereafter, and a concentration G reduction rate comparison model for the second preset time period ΔT4 is constructed, and the second comparison average reduction rate R3 is calculated. The real average reduction rate R4 of the concentration G in the second preset time period ΔT2 is calculated, and the real average reduction rate R4 is compared with the second comparison average reduction rate R3. The concentration G reduction rate comparison model for the second preset time period ΔT4 is:

[0141] Among them, α2 and b2 represent the set reduction rate curve parameters, t2 represents the time of the second preset time period ΔT4;

[0142] When R2>R1 and R4<R3, the monitored live traffic is deemed to be false traffic.

[0143] The live broadcast traffic authenticity monitoring model of the present invention has technical effects such as comprehensive data processing, comprehensive feature variable construction, dynamic monitoring model establishment, multi-parameter comprehensive judgment and real-time false traffic determination, etc., which provides an efficient and accurate traffic monitoring means for e-commerce live broadcast shopping platforms, ensuring the authenticity and security of live broadcast transactions.

[0144] In a preferred embodiment of the present invention, the present invention establishes a first judgment standard and a second judgment standard for live broadcast monitoring for potential risk label and dangerous label anchors, specifically including:

[0145] A first judgment standard is established for potential risk label anchors. The first judgment standard adopts a first preset time period ΔT2 aggregation degree G growth rate comparison model as follows: The comparison model of the reduction rate of the aggregation degree G in the second preset time period ΔT4 is:

[0146] A second judgment standard is established for the dangerous label anchor, and the second judgment standard adopts the adjusted first preset time period ΔT2 aggregation degree G growth rate comparison model as follows: The comparison model of the reduction rate of the aggregation degree G in the second preset time period ΔT4 is:

[0147] Set α1>α1 * , b1>b1 * , α2<α2 * , b2>b2 * .

[0148] The first judgment criterion and the second judgment criterion of the present invention have technical characteristics such as targeting, dynamic monitoring, parameter adjustment and real-time judgment, and provide an efficient and accurate solution for the live broadcast monitoring of anchors with potential risk tags and dangerous tags.

[0149] Based on the above, please refer to Figure 2 , an embodiment of the present invention provides a data analysis system for e-commerce live shopping, the system comprising:

[0150] The request management module is used to receive the verification request and initial bonus of the target anchor sent by the willing seller, and compare them with the pre-stored blacklist. If there is no match, the live broadcast record data of the target anchor shared by other sellers in response to the request is obtained;

[0151] The risk assessment module is used to conduct a potential risk assessment on the target anchor based on the live broadcast record data. If the assessment shows that there is no potential risk, the target anchor is monitored for public opinion in the recent preset time period. If negative public opinion related to the target anchor is monitored, the target anchor is assigned a potential risk label, otherwise the target anchor is assigned a potential safety label; if the assessment shows that there is a potential risk, the target anchor is assigned a potential risk label, and a bonus of a preset proportion of the initial bonus is paid to the seller who shared the target anchor's live broadcast record data according to the preset smart contract;

[0152] The anomaly monitoring module is used to establish a relationship map for the target anchors with potential risk labels, and use a convolutional neural network to conduct in-depth analysis of the anchors' relationship map to detect whether there are anomalies. If an anomaly is detected, the target anchor's potential risk label is replaced with a dangerous label, and the detection result is sent to the willing sellers;

[0153] The live broadcast monitoring module is used to establish a live broadcast traffic authenticity monitoring model, and establish the first judgment standard and the second judgment standard for potential risk label and dangerous label anchors for live broadcast monitoring. If false traffic is monitored, the target anchor will be added to the blacklist and a warning report will be sent to the willing sellers.

[0154] The present invention combines live broadcast record data and public opinion monitoring through risk assessment module, and can comprehensively and multi-angle evaluate the risk of target anchor through potential risk assessment. This comprehensiveness ensures the effective identification of various potential risks, thereby providing a strong foundation for subsequent monitoring; the abnormal monitoring module uses convolutional neural network to conduct in-depth analysis of the anchor's relationship map, and can efficiently detect whether there is abnormality in the anchor with potential risk label. This intelligent monitoring method ensures the timely discovery and processing of abnormal situations, and reduces the possibility of potential risks; the live broadcast monitoring module establishes a monitoring model for the authenticity of live broadcast traffic, and establishes corresponding judgment criteria according to anchors with different labels. This dynamic monitoring mechanism enables the system to adjust the monitoring strategy according to actual conditions, ensuring accurate monitoring of false traffic; once false traffic is monitored, the system can timely add the target anchor to the blacklist, and send a warning report to the willing seller, realizing rapid early warning and disposal, and this mechanism effectively curbs the spread of false traffic and ensures the fairness and authenticity of live shopping.

[0155] The data analysis system of the present invention has technical characteristics such as comprehensive risk assessment, intelligent anomaly monitoring, dynamic live broadcast monitoring and timely early warning, which provides reliable protection for e-commerce live broadcast shopping and improves the security and trust of the entire live broadcast transaction system.

[0156] Based on the above, a computer-readable storage medium is also provided, on which a computer program stored implements the above method when running.

[0157] It should be understood that the system and its modules shown above can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art will understand that the above methods and systems can be implemented using computer executable instructions and / or included in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The system and its modules of the present application can not only be implemented by hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but can also be implemented by software such as executed by various types of processors, and can also be implemented by a combination of the above hardware circuits and software (e.g., firmware).

[0158] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other beneficial effects that may be obtained.

[0159] The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is only for example and does not constitute a limitation of the present application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements and amendments to the present application. Such modifications, improvements and amendments are suggested in the present application, so such modifications, improvements and amendments still belong to the spirit and scope of the exemplary embodiments of the present application.

[0160] At the same time, the present application uses specific words to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0161] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any new and useful improvements thereto. Accordingly, various aspects of the present application may be performed entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "data blocks", "modules", "engines", "units", "components" or "systems". In addition, various aspects of the present application may be represented as a computer product located in one or more computer-readable media, which includes computer-readable program code.

[0162] A computer storage medium may include a propagated data signal containing computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may be in a variety of forms, including electromagnetic, optical, etc., or a suitable combination. A computer storage medium may be any computer-readable medium other than a computer-readable storage medium, which can be connected to an instruction execution system, device or apparatus to communicate, propagate or transmit the program for use. The program code on the computer storage medium may be transmitted via any suitable medium, including radio, cable, fiber optic cable, RF, or similar media, or any combination of the above media.

[0163] The computer program coding required for the operation of each part of the application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C language, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages, etc. The program coding can be run completely on the user's computer, or run on the user's computer as an independent software package, or run partly on the user's computer and partly on the remote computer, or run completely on the remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0164] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0165] Similarly, it should be noted that in order to simplify the description of the disclosure of this application and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this application, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the object of this application are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0166] In some embodiments, numbers describing the number of components and attributes are used. It should be understood that such numbers used for the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers allow adaptive changes. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which can be changed according to the required characteristics of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining the digits. Although the numerical domains and parameters used to confirm the breadth of the range in some embodiments of the present application are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0167] Each patent, patent application, patent application disclosure, and other materials, such as articles, books, instructions, publications, documents, etc., cited in this application are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the content of this application, documents that limit the broadest scope of the claims of this application (currently or later attached to this application) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the attached materials of this application are inconsistent or conflicting with the content described in this application, the descriptions, definitions, and / or use of terms in this application shall prevail.

[0168] Finally, it should be understood that the embodiments described in this application are only used to illustrate the principles of the embodiments of the present application. Other variations may also fall within the scope of the present application. Therefore, as an example and not a limitation, the alternative configurations of the embodiments of the present application may be considered to be consistent with the teachings of the present application. Accordingly, the embodiments of the present application are not limited to the embodiments explicitly introduced and described in the present application.

[0169] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A data analysis method for e-commerce live shopping. It is characterized in that The method comprises: Receive verification requests and initial bonuses from willing sellers, and compare them with the pre-stored blacklist. If there is no match, obtain live broadcast record data of the target anchor shared by other sellers in response to the request; The target anchor is evaluated for potential risks based on the live broadcast record data. If the evaluation shows that there is no potential risk, the target anchor is monitored for public opinion in the recent preset time period. If negative public opinion related to the target anchor is monitored, the target anchor is given a potential risk label, otherwise the target anchor is given a potential safety label; if the evaluation shows that there is a potential risk, the target anchor is given a potential risk label, and a bonus of a preset proportion of the initial bonus is paid to the seller who shared the target anchor's live broadcast record data according to the preset smart contract; Establish a relationship map for the target anchor with a potential risk label, determine whether there is an abnormality based on the relationship map, and if it is determined that there is an abnormality, replace the potential risk label of the target anchor with a dangerous label, and send the judgment result to the willing seller; specifically, it includes: collecting the fan attention data of the target anchor, obtaining the attention relationship between the target anchor and fans and between fans; according to the attention relationship, connect the target anchor as the root node and the fans as child nodes in a tree structure to obtain a relationship map represented by a tree structure; use an anomaly detection algorithm to detect whether there is an abnormality in the relationship map, and / or collect anchor relationship map data for training, mark and classify it as normal or abnormal, pre-process the relationship map, convert the relationship map into an adjacency matrix or a node feature matrix, build a convolutional neural network model for abnormality judgment, train the convolutional neural network model through label training data, and use the trained convolutional neural network to conduct in-depth analysis of the anchor's relationship map after evaluation; if it is determined that there is an abnormality, replace the potential risk label of the target anchor with a dangerous label, and send the judgment result to the willing seller; Establish a model for monitoring the authenticity of live broadcast traffic, and establish the first and second judgment criteria for potential risk label and dangerous label anchors to monitor live broadcasts. If false traffic is monitored, the target anchor will be added to the blacklist and a warning report will be sent to the intended sellers; The live broadcast record data includes audience interaction data, product sales data, audience quality data, and interaction difference data between buyers and non-buyers; Conduct potential risk assessment on the target anchor based on live broadcast record data, including: Calculate the average potential risk assessment value The average potential risk assessment value The calculation formula is as follows: Where N represents the number of live recorded data, R i represents the potential risk assessment value of the i-th live broadcast record data; Among them, R represents the potential risk assessment value, I A An indicator representing product sales data, I B Indicator representing audience interaction records, I C An indicator of audience quality, I D An indicator of the difference in interactions between buyers and non-buyers, w A 、w B 、w C and w D Respectively represent weights; I B =a B ·C+β B ·L+γ B ·S, Among them, M * Indicates the total number of products purchased by viewers after watching the live broadcast, M 0 represents the number of repeat purchases by viewers, Q represents product sales, Z represents the return rate, C represents the number of comments during the live broadcast, L represents the number of likes during the live broadcast, S represents the number of shares during the live broadcast, n represents the number of viewers, and x represents the number of i represents the longitude or latitude of the i-th viewer coordinates, represents the mean of the longitude or latitude of the audience coordinates, Md represents the median of audience interaction, and M 1 represents the number of buyer interactions, M 2 represents the number of non-buyer interactions, α A , β A , α B , β B , γ B , α C and β C represent the regulating factors respectively; Comparison of average potential risk assessment values and the preset risk threshold R 0 The size of the potential risk assessment value Greater than the preset risk threshold R 0 , the target anchor is deemed to have potential risks.

2. The data analysis method for e-commerce live shopping according to claim 1, It is characterized in that The method further comprises: Establishing a bonus pool that is bound to the seller's account and receives the initial bonus, wherein the bonus pool is set with read-only permissions, allowing only authenticated sellers to view the bonus pool balance and transaction records; Compare the target anchor's identity information with the pre-stored blacklist; If the target anchor's identity information does not match any record in the blacklist, the initial bonus will be pre-deposited into the bonus pool, and a response request for the target anchor's live broadcast record data will be initiated to other sellers; If the identity information of the target anchor matches the record in the blacklist, the initial bonus will be deducted by the first preset proportion of system service points, the remaining bonus will be returned to the willing seller, and the blacklist data of the target anchor will be sent.

3. The data analysis method for e-commerce live shopping according to claim 2, It is characterized in that The target anchor is monitored for public opinion in the recent preset time period. If negative public opinion related to the target anchor is monitored, the target anchor is assigned a potential risk label; otherwise, the target anchor is assigned a potential safety label. Specifically, it includes: Collect the target anchor’s name and keywords, and set a monitoring time window; Use the name and keywords of the target anchor to conduct real-time searches on news media and social media to obtain public opinions related to the target anchor within the monitoring time window, including news articles, social media posts, comments, and discussions; Conduct sentiment analysis on public opinion content and classify it into three types of sentiment polarity: positive, negative, or neutral. The sentiment analysis method is as follows: converting the acquired public opinions into text data and preprocessing the text, splitting the text into words or word sequences, using a predefined sentiment vocabulary, assigning a sentiment score to each word in the text, increasing the score for positive words and decreasing the score for negative words, and then calculating the summary score to classify the sentiment polarity of the text; If the public opinion is marked as negative sentiment and is directly related to the target anchor, it will be recorded as negative public opinion; Conduct content analysis on negative public opinions to determine whether they are related to the target anchor’s behavior, products, or speeches. If so, include the content of the public opinions in the risk assessment; Count the number of negative public opinions within a preset time period. If the number exceeds the preset threshold, it is judged that there is a potential risk. The target anchor is assigned a potential risk label, otherwise the target anchor is assigned a potential safety label.

4. The data analysis method for e-commerce live shopping according to claim 3, It is characterized in that The smart contract preset method is as follows: Deploy a smart contract to enable all sellers to participate; Generate a unique contract address, which is used for interaction; The execution conditions set in the smart contract are: if there is no potential risk in the assessment, the initial bonus will be deducted by a second preset proportion of system service points, and the remaining bonus will be returned to the willing seller; If the assessment shows potential risks, a bonus equal to a preset proportion of the initial bonus will be paid to the seller who shared the live broadcast record data of the target anchor, and a third preset proportion of system service points will be deducted from the initial bonus.

5. The data analysis method for e-commerce live shopping according to claim 4, It is characterized in that The third preset ratio is greater than the second preset ratio, and the second preset ratio is greater than the first preset ratio.

6. The data analysis method for e-commerce live shopping according to claim 1, It is characterized in that The establishment of a live broadcast traffic authenticity monitoring model includes: Collect live broadcast data, clean and preprocess it, and handle missing values ​​and outliers; Construct the characteristic variables of the aggregation model and calculate the aggregation degree G, where the calculation formula of the aggregation degree G is as follows: A 1 =N 1 ·T 1 , A 2 =H 2 ·ΔP 2 , Among them, A 1 Indicates the anchor's attractiveness, A 2 Indicates product attractiveness, D 1 represents the audience dispersion, F 1 Indicates the number of audience interactions per minute, S 1 represents the number of new viewers entering per minute, S 2 Indicates the number of product orders per minute, w 1 and w 2 Respectively represent the adjustment coefficient; N 1 Indicates the number of fans of the anchor, T 1 Indicates the cumulative live broadcast time of the anchor, H 2 Indicates the weighted comprehensive value of the product's views, favorites, and purchases, ΔP 2 Indicates the preferential price of the product, K i represents the proportion of the audience in the ith region to the total audience, and n represents the number of regions; When the live broadcast reaches the preset time T 1 When the first preset time period ΔT 2 Start the first live broadcast traffic authenticity monitoring and build the first preset time period ΔT 2 Aggregation G growth rate comparison model, calculate the first comparison average growth rate R 1 , calculate the first preset time period ΔT 2 The real average growth rate R of the aggregation degree G 2 , compare the real average growth rate R 2 Compared with the first average growth rate R 1 The size of the first preset time period ΔT 2 The comparison model of the aggregation degree G growth rate is: Among them, α 1 and b 1 Represents the set growth rate curve parameter, t 1 Indicates the first preset time period ΔT 2 time; In the first preset time period ΔT 2 After that, when it is detected that the aggregation degree G continues to decrease for a preset period of time ΔT 3 When the second preset time period ΔT 4 Start the second live broadcast traffic authenticity monitoring and build the second preset time period ΔT 4 Aggregation G reduction rate comparison model, calculate the second comparison average reduction rate R 3 , calculate the second preset time period ΔT 2 The real average reduction rate R of the aggregation G 4 , compare the actual average reduction rate R 4 Compared with the second average reduction rate R 3 The size of the second preset time period ΔT 4 The comparison model of aggregation degree G reduction rate is: Among them, α 2 and b 2 Indicates the set reduction rate curve parameters, t2 Indicates the second preset time period ΔT 4 time; When R 2 〉R 1 And R 4 〈R 3 When the monitored live broadcast traffic is detected, it is determined that the monitored live broadcast traffic is false traffic.

7. The data analysis method for e-commerce live shopping according to claim 6, It is characterized in that The first judgment standard and the second judgment standard are respectively established for the anchors with potential risk labels and dangerous labels to monitor the live broadcast, specifically including: A first judgment standard is established for a potential risk label anchor, and the first judgment standard adopts a first preset time period ΔT 2 The comparison model of the aggregation degree G growth rate is: The second preset time period ΔT 4 The comparison model of aggregation degree G reduction rate is: A second judgment standard is established for the dangerous label anchor, and the second judgment standard adopts the adjusted first preset time period ΔT 2 The comparison model of the aggregation degree G growth rate is: The second preset time period ΔT 4 The comparison model of aggregation degree G reduction rate is: set α 1 >α 1 * ,b 1 >b 1 * ,α 2 <α 2 * ,b 2 >b 2 * 。 8. A data analysis system for e-commerce live shopping using the data analysis method for e-commerce live shopping according to any one of claims 1 to 7, It is characterized in that The system comprises: The request management module is used to receive the verification request and initial bonus of the target anchor sent by the willing seller, and compare them with the pre-stored blacklist. If there is no match, the live broadcast record data of the target anchor shared by other sellers in response to the request is obtained; The risk assessment module is used to conduct a potential risk assessment on the target anchor based on the live broadcast record data. If the assessment shows that there is no potential risk, the target anchor is monitored for public opinion in the recent preset time period. If negative public opinion related to the target anchor is monitored, the target anchor is assigned a potential risk label, otherwise the target anchor is assigned a potential safety label; if the assessment shows that there is a potential risk, the target anchor is assigned a potential risk label, and a bonus of a preset proportion of the initial bonus is paid to the seller who shared the target anchor's live broadcast record data according to the preset smart contract; The live broadcast record data includes audience interaction data, product sales data, audience quality data, and interaction difference data between buyers and non-buyers; Conduct potential risk assessment on the target anchor based on live broadcast record data, including: Calculate the average potential risk assessment value The average potential risk assessment value The calculation formula is as follows: Where N represents the number of live recorded data, R i represents the potential risk assessment value of the i-th live broadcast record data; Among them, R represents the potential risk assessment value, I A An indicator representing product sales data, I B Indicator representing audience interaction records, I C An indicator of audience quality, I D An indicator of the difference in interactions between buyers and non-buyers, w A 、w B 、w C and w D Respectively represent weights; I B =a B ·C+β B ·L+γ B ·S, Among them, M * Indicates the total number of products purchased by viewers after watching the live broadcast, M 0 represents the number of repeat purchases by viewers, Q represents product sales, Z represents the return rate, C represents the number of comments during the live broadcast, L represents the number of likes during the live broadcast, S represents the number of shares during the live broadcast, n represents the number of viewers, and x represents the number of i represents the longitude or latitude of the i-th viewer coordinates, represents the mean of the longitude or latitude of the audience coordinates, Md represents the median of audience interaction, and M 1 represents the number of buyer interactions, M 2 represents the number of non-buyer interactions, α A , β A , α B , β B , γ B , α C and β C represent the regulating factors respectively; Comparison of average potential risk assessment values and the preset risk threshold R 0 When the average potential risk assessment value R is greater than the preset risk threshold R 0 , the target anchor is considered to have potential risks; The abnormality monitoring module is used to establish a relationship map for the target anchor with potential risk labels, and use a convolutional neural network to perform in-depth analysis on the anchor's relationship map to detect whether there is an abnormality. If an abnormality is detected, the target anchor's potential risk label is replaced with a dangerous label, and the detection result is sent to the willing seller; specifically, it includes: collecting the fan attention data of the target anchor, obtaining the attention relationship between the target anchor and fans and between fans; according to the attention relationship, the target anchor is used as the root node and the fans are used as child nodes to connect in a tree structure to obtain a relationship map represented by a tree structure; using an abnormality detection algorithm to detect whether there is an abnormality in the relationship map, and / or, collecting anchor relationship map data for training, marking and classifying it as normal or abnormal, pre-processing the relationship map, converting the relationship map into an adjacency matrix or a node feature matrix, constructing a convolutional neural network model for abnormality judgment, training the convolutional neural network model through label training data, and using the trained convolutional neural network to perform in-depth analysis on the anchor's relationship map after evaluation; if it is judged that there is an abnormality, the target anchor's potential risk label is replaced with a dangerous label, and the judgment result is sent to the willing seller; The live broadcast monitoring module is used to establish a live broadcast traffic authenticity monitoring model, and establish the first judgment standard and the second judgment standard for potential risk label and dangerous label anchors respectively for live broadcast monitoring. If false traffic is monitored, the target anchor will be added to the blacklist and a warning report will be sent to the willing sellers.

Images