Big data-based live broadcast operation system and method

Abstract

The invention belongs to the technical field of operation systems, and particularly relates to a live broadcast operation system and method based on big data, and the method comprises the following steps: obtaining commodity historical marketing data and user behavior data in a live broadcast process, and respectively summarizing the data into a commodity database and a user database; the method comprises the steps of obtaining a hot search event in a live broadcast process in real time, extracting a specific information element in the hot search event, finding out a corresponding marketing commodity which is not broadcasted according to the specific information element, and marking the marketing commodity as a hot search associated commodity. According to the method, two dimensions of the hot search trend value and the trend adjustment value are introduced on the basis of the sales volume trend value, the sales trend of the commodities can respond to emergencies in time, the lagging of commodity sales trend judgment is reduced, meanwhile, the display sequence of the to-be-displayed commodities is dynamically adjusted, and the display efficiency of the commodities is improved. The displayed commodities can be matched with more online users, and the user retention rate and the commodity conversion rate are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of operation systems, and specifically relates to a live broadcast operation system and method based on big data. Background Art

[0002] With the development of Internet technology, the live broadcast industry has rapidly become an important part of modern media and entertainment. Live broadcast platforms provide users with opportunities for real-time interaction, allowing viewers to communicate and exchange with anchors instantly. However, traditional live broadcast operation methods often rely on manual management and experience judgment, which has obvious limitations in processing large amounts of data, optimizing user experience, and improving operational efficiency.

[0003] At present, the operation of live broadcast platforms mainly faces the following challenges:

[0004] 1. During the live broadcast, the order of product display is fixed, and the number and composition of users watching the live broadcast change in real time. Each user has their own preferences. When the displayed products cannot deeply match more users, the user retention rate decreases, which in turn reduces the conversion rate of the products, forming a vicious cycle and resulting in poor live broadcast operation results;

[0005] 2. In the existing technology, the judgment of product sales trends mainly relies on data such as changes in sales volume and product browsing volume. The judgment factors are too single and there is a lag. In some emergencies, the operation response is delayed, which leads to the inability to adjust the product sales strategy in time according to the sudden situation;

[0006] Based on the above problems, this application document proposes a live broadcast operation system and method based on big data to improve the problems existing in the existing live broadcast operation system. Summary of the invention

[0007] The purpose of the present invention is to provide a live broadcast operation system and method based on big data. By introducing two dimensions, hot search trend value and trend adjustment value, on the basis of sales trend value, the sales trend of goods can respond to emergencies in a timely manner and reduce the lag in judging the sales trend of goods. At the same time, by dynamically adjusting the exhibition order of the exhibited goods, the exhibited goods can be matched with more online users, thereby improving user retention rate and product conversion rate.

[0008] The technical solution adopted by the present invention is as follows:

[0009] A live broadcast operation method based on big data, comprising the following steps:

[0010] The following steps are involved:

[0011] Obtain the historical product marketing data and user behavior data during the live broadcast process, and summarize them into a product database and a user database respectively. The historical product marketing data includes product attribute tags, category tags, historical product sales, and product conversion rates. The user behavior data includes user basic attribute tags, behavior tags, and preference tags.

[0012] Acquire hot search events during the live broadcast in real time, extract specific information elements from the hot search events, find corresponding unbroadcasted marketing products based on the specific information elements, and mark them as hot search related products;

[0013] Output the popularity trend value of each undisplayed marketing product based on the acquired product historical sales volume, hot search related products, and product conversion rate;

[0014] Obtaining in real time the preference tags of online users during the live broadcast and the marketing coefficients of the marketing products that have not been broadcasted, and dynamically adjusting the broadcasting order of the marketing products that have not been broadcasted according to the marketing coefficients and the popularity trend values ​​of the products;

[0015] The operation data after the live broadcast and the trend adjustment value of each marketing commodity are obtained, the operation data and the trend adjustment value are respectively merged into the commodity database and the user database, and the commodity database and the user database are updated.

[0016] In a preferred solution, historical product marketing data and user behavior data during the live broadcast are obtained and summarized into a product database and a user database, respectively. The historical product marketing data includes product attribute tags, category tags, historical product sales, and product conversion rates. The user behavior data includes user basic attribute tags, behavior tags, and preference tags. The following steps are included:

[0017] Extract all product historical marketing data from historical live broadcast records, where the product historical marketing data includes product brand, product name, product attributes, product category, product historical sales volume, and product conversion rate;

[0018] Extracting user data from historical live broadcast records, wherein the user data includes viewing behavior data, purchase behavior data, and search behavior data of the user;

[0019] Preprocessing the acquired commodity historical marketing data and user data, and generating corresponding commodity feature tags and user feature tags, and aggregating the commodity historical marketing data, commodity feature tags, user data and user feature tags into a commodity database and a user database respectively;

[0020] Among them, the commodity feature tags at least include commodity attribute tags and category tags, and the user feature tags at least include user basic attribute tags, behavior tags, and preference tags.

[0021] In a preferred solution, hot search events during a live broadcast are obtained in real time, specific information elements in the hot search events are extracted, and corresponding unbroadcasted marketing products are found according to the specific information elements and marked as hot search related products, including the following steps:

[0022] Constructing an associated database, constructing matching words, classification labels and mapping rules in the associated database, and associating matching words and classification labels according to the mapping rules;

[0023] Monitor the real-time hot search lists of major social media platforms, news websites, and live broadcast platforms in real time, obtain all hot search events in the hot search lists, extract specific information elements of the hot search events, compare the specific information elements with the matching words in the associated database, and when the specific information elements are consistent with the matching words, extract the corresponding classification labels from the associated database according to the mapping rules;

[0024] Obtain the category labels of the marketing products that are not displayed, and compare the extracted classification labels with the category labels of the marketing products that are not displayed;

[0025] Find the undisplayed marketing products whose category labels and classification labels are consistent, and mark them as hot search related products.

[0026] In a preferred solution, the step of outputting the popularity trend value of each undisplayed marketing product according to the acquired historical sales volume of the product, hot search related products, and product conversion rate includes the following steps:

[0027] Obtain the historical sales volume and conversion rate of all marketing products that have not been displayed;

[0028] Outputting a sales trend value and a trend adjustment value of the marketing product according to the historical sales volume and product conversion rate of the product;

[0029] Obtain hot search related products from all marketing products that have not been displayed, and output the hot search trend values ​​of the hot search related products;

[0030] Substitute the sales trend value, hot search trend value and trend adjustment value into the first expression to obtain the popularity trend value of the product. The first expression is: In the formula, f0 represents the popularity trend value of the product, f1 represents the sales trend value of the marketing product, f2 represents the hot search trend value of the marketing product, and f3 represents the trend adjustment value obtained after the last live broadcast of the marketing product.

[0031] Among them, for marketing products that have not been broadcast except for hot search related products, the hot search trend value f2 is 0. The marketing products have not been broadcast during the historical live broadcast, and the trend adjustment value f3 is 0.

[0032] In a preferred embodiment, outputting the sales volume trend value of a commodity according to the historical sales volume of the commodity comprises the following steps:

[0033] Establishing a weight evaluation interval, wherein there are multiple weight evaluation intervals, and each weight evaluation interval corresponds to a weight factor;

[0034] Taking the current live broadcast start time as the starting node, a backtracking period is established in reverse chronological order, and multiple backtracking cycles with equal intervals are set within the backtracking period;

[0035] Get the period sales of the product in each retrospective period, as well as the total sales of the product in the retrospective period;

[0036] Compare the total sales volume of the product with multiple weight evaluation intervals in sequence to obtain the weight factor corresponding to the total sales volume of the product in the retrospective period;

[0037] Substitute the product cycle sales volume and weight factor into the second expression to obtain the sales volume trend value of the product. The second expression is: Among them, C i represents the sales volume of the product in the i-th retrospective period, C i-1 It represents the sales volume of the product in the i-1th retrospective period, T represents the length of the retrospective period, n represents the number of retrospective periods, δ represents the weight factor corresponding to the product, and i represents the number of the retrospective period.

[0038] In a preferred solution, the step of obtaining hot search related products from all marketing products that have not been displayed, and outputting the hot search trend values ​​of the hot search related products, includes the following steps:

[0039] The time when the hot search event is obtained from the hot search list is used as the starting node, a hot search period is established, and multiple hot search cycles with equal intervals are set within the hot search period until the hot search event is removed from the hot search list;

[0040] Get the hot search sales of hot search related products in each hot search cycle;

[0041] Substitute the hot search sales volume into the third expression to obtain the hot search trend value of the product. The third expression is: Among them, K j represents the hot search sales volume of the product in the jth hot search cycle, K j-1 It represents the sales volume of the hot search of the product in the j-1th hot search cycle, L represents the duration of the hot search period, m represents the number of hot search cycles, and i represents the number of the hot search cycle.

[0042] In a preferred solution, the preference tags of online users during the live broadcast and the marketing coefficients of the marketing products that have not been broadcasted are obtained in real time, and the broadcasting order of the marketing products that have not been broadcasted is dynamically adjusted according to the marketing coefficients and the popularity trend values, including the following steps:

[0043] Constructing a matching evaluation interval, wherein there are multiple matching evaluation intervals, and each matching evaluation interval corresponds to a marketing coefficient;

[0044] Extract the preference tags of online users during the live broadcast in real time through the user database;

[0045] Extract category labels of marketing products that are not on display from the product database in real time;

[0046] Compare the category labels of the marketing products that have not been displayed with the preference labels of online users one by one, count the comparison results, and obtain the number of matches between the marketing products that have not been displayed and the online users;

[0047] Compare the matching quantity with the matching evaluation interval to obtain the marketing coefficient corresponding to the marketing product that has not been displayed;

[0048] Substituting the marketing coefficient and the popularity trend value into the fourth expression to obtain the ranking index of each marketing product that has not been displayed, the fourth expression is: Among them, P represents the ranking index of marketing products that have not been displayed. Indicates the marketing coefficient of the marketing products that are not displayed;

[0049] Arrange the ranking indexes of all marketing products that have not been displayed in descending order to obtain a product display sequence;

[0050] The marketing products that have not been displayed are displayed according to the arrangement order of the display sequence, among which the products that have been displayed are removed from the display sequence and will not be displayed again.

[0051] In a preferred solution, the acquisition of the operational data after the live broadcast and the trend adjustment value of each marketing commodity, merging the operational data and the trend adjustment value into the commodity database and the user database respectively, and updating the commodity database and the user database includes the following steps:

[0052] Constructing a conversion rate evaluation interval, wherein there are multiple conversion rate evaluation intervals, and each conversion rate evaluation interval corresponds to a trend adjustment value;

[0053] After the current live broadcast ends, the operation data during the current live broadcast is obtained, and the operation data includes the number of clicks, transaction volume, browsing history, and purchase history of online users of each marketing product;

[0054] Preprocessing the operational data and generating corresponding product marketing data and user data;

[0055] Output the current conversion rate of marketing products according to the number of clicks and transaction volume of each marketing product during the current live broadcast;

[0056] Compare the current conversion rate with the conversion rate evaluation interval to obtain a trend adjustment value corresponding to each marketing commodity;

[0057] The product marketing data, trend adjustment values ​​and user data obtained during the current live broadcast are respectively merged into the product database and the user database, and the data in the product database and the user database are updated.

[0058] A live broadcast operation system based on big data, applicable to any of the above-mentioned live broadcast operation methods based on big data, comprising:

[0059] A storage module, wherein the storage module is used to store user data and product data;

[0060] Collection module: The collection module is used to collect hot search data and product sales data;

[0061] A processing module, which is used to process relevant data and calculate the display index and conversion rate of the product;

[0062] The feedback module is used to obtain conversion data and user data of the product.

[0063] The technical effects achieved by the present invention are:

[0064] When judging the sales trend of commodities, the present invention introduces two dimensions, namely, the hot search trend value and the trend adjustment value, on the basis of the original sales trend value, and refines the influencing factors of the sales trend of commodities. In emergencies, the sales trend of commodities can respond to emergencies in a timely manner, thereby effectively reducing the lag of judging the sales trend of commodities and improving operational efficiency.

[0065] The present invention realizes dynamic optimization of the order of commodity display during the live broadcast by a series of measures such as real-time acquisition of online user data during the live broadcast, accurate matching of commodities with user data, dynamic adjustment of commodity display order, and updating of commodity popularity trend value and user data according to live broadcast feedback, so that the displayed commodities can be matched with more online users, thereby improving user retention rate and commodity conversion rate, which not only improves the viewing experience of the live broadcast, but also helps it to better meet user needs and market trends, and improves the live broadcast operation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a flowchart of the live broadcast operation method of the present invention;

[0067] Figure 2 It is a schematic diagram of the framework of the live broadcast operation system of the present invention. DETAILED DESCRIPTION

[0068] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0069] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0070] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure or characteristic that may be included in at least one implementation of the present invention. The phrase "in a preferred embodiment" that appears in different places in this specification does not refer to the same embodiment, nor is it a separate or selective embodiment that is mutually exclusive with other embodiments.

[0071] Please see attached Figure 1 As shown, it is the first embodiment of the present invention, which provides a live broadcast operation method based on big data, including the following steps:

[0072] S100: Obtaining historical product marketing data and user behavior data during the live broadcast process, and summarizing them into a product database and a user database respectively, wherein the historical product marketing data includes product attribute tags, category tags, historical product sales, and product conversion rates, and the user behavior data includes user basic attribute tags, behavior tags, and preference tags;

[0073] S200: Acquire hot search events during the live broadcast in real time, extract specific information elements from the hot search events, find corresponding marketing products that have not been broadcasted according to the specific information elements, and mark them as hot search related products;

[0074] S300: Outputting the popularity trend value of each marketing product that has not been displayed according to the acquired historical sales volume of the product, hot search related products, and product conversion rate;

[0075] S400: obtaining in real time the preference tags of online users during the live broadcast and the marketing coefficients of the marketing products that have not been broadcasted, and dynamically adjusting the broadcasting order of the marketing products that have not been broadcasted according to the marketing coefficients and the popularity trend values ​​of the products;

[0076] S500: Acquire the operation data after the live broadcast and the trend adjustment value of each marketing product, merge the operation data and the trend adjustment value into the product database and the user database respectively, and update the product database and the user database.

[0077] Here, the product conversion rate refers to the proportion of users who browse or visit a product within a certain time frame who are converted into users who actually purchase the product during the product sales process. Specifically, if the number of views of product A in April is 4,000 and the sales volume is 1,000, then the conversion rate of product A in April is 25%. The product conversion rate is an important indicator for evaluating sales results and user purchasing behavior, which helps optimize marketing and improve sales results. The product conversion rate can be used to analyze user behavior, evaluate sales results, and optimize marketing strategies. A higher conversion rate usually indicates better sales results and user willingness to buy, while a lower conversion rate may require adjustments to products, pricing, or marketing activities.

[0078] As described in the above steps S100 to S500, by establishing a user database and a product database, a data foundation is provided for the live broadcast operation work, which facilitates the detailed sorting and analysis of user data and product data, and measures the degree of recognition of the product among the audience by evaluating the popularity trend value of the product. In addition, by collecting the popularity trend value of the product and the dynamic operation data of the user during the live broadcast in real time, the display order of the product is further optimized and adjusted. For example, the current live broadcast products include A, B, C, D, E, F, and G. First, the hot search events in the current hot search list are obtained through several Internet platforms, and the specific information elements (such as keywords) in the hot search events are obtained by performing relevant text processing on the hot search events. The specific information elements and products are matched, wherein the products that can match the specific information elements are marked as hot search related products, and the relevant data of the hot search related products are obtained, and then the historical marketing data of A, B, C, D, E, F, and G are respectively obtained through the product database, (such as) historical sales data and conversion rate, and each of the above products is evaluated according to the historical sales data, conversion rate and related data of each of the above products. The heat trend value of a commodity indirectly reflects the audience degree of the commodity, which is convenient for adjusting the display order of multiple commodities of the same type in the future. Then, the behavior data of each online user during the live broadcast (such as user basic attribute labels, behavior labels, preference labels, etc.) can be obtained in real time through the user database. Through the behavior data of online users, their consumption needs and consumption preferences can be analyzed. According to the preferences of online users and the heat trend values ​​of commodities, the above commodities are sorted in real time, so that the next displayed commodity can meet the preferences of more online users as much as possible. This dynamic adjustment strategy based on real-time data ensures that the live broadcast content can continue to be consistent with the interests and needs of online users, thereby significantly improving the viewing experience of online users and the user retention rate of the live broadcast platform. The exposure rate and sales conversion rate of commodities have also been significantly improved, thereby enhancing the commercial value of the live broadcast. After the live broadcast activity, by collecting live broadcast operation data, user data and commodity data are updated and optimized, so that the heat trend values ​​of user data and commodities are more accurate and reliable, so as to improve user satisfaction and promote the continuous optimization and development of the live broadcast platform.

[0079] In a preferred embodiment, historical product marketing data and user behavior data during the live broadcast are obtained and summarized into a product database and a user database, respectively. The historical product marketing data includes product attribute tags, category tags, historical product sales, and product conversion rates. The user behavior data includes user basic attribute tags, behavior tags, and preference tags. The following steps are included:

[0080] S101: extracting all product historical marketing data from historical live broadcast records, where the product historical marketing data includes product brand, product name, product attribute, product category, product historical sales volume, and product conversion rate;

[0081] S102: extracting user data from historical live broadcast records, where the user data includes viewing behavior data, purchase behavior data, and search behavior data of the user;

[0082] S103: pre-processing the acquired commodity historical marketing data and user data, and generating corresponding commodity feature tags and user feature tags, and aggregating the commodity historical marketing data, commodity feature tags, user data and user feature tags into a commodity database and a user database, respectively, wherein the pre-processing includes cleaning, deduplication and standardization processing;

[0083] Among them, the commodity feature tags at least include commodity attribute tags and category tags, and the user feature tags at least include user basic attribute tags, behavior tags, and preference tags.

[0084] As described in the above steps S101 to S103, the accuracy and reliability of the data are ensured by cleaning, deduplicating and standardizing the data, and the pre-processed historical marketing data of the goods and user behavior data are respectively summarized into a product database and a user database, providing a data basis for the subsequent dynamic adjustment of the exhibition order of the goods to be exhibited.

[0085] In a preferred embodiment, the real-time acquisition of hot search events during the live broadcast process, extraction of specific information elements in the hot search events, finding corresponding unbroadcasted marketing products according to the specific information elements, and marking them as hot search related products, includes the following steps:

[0086] S201: constructing an associated database, constructing matching words, classification labels and mapping rules in the associated database, and associating matching words and classification labels according to the mapping rules;

[0087] S202: Real-time monitoring of hot search lists on major social media platforms, news websites, and live broadcast platforms, obtaining all hot search events in the hot search lists, extracting specific information elements of the hot search events, comparing the specific information elements with matching words in the associated database, and when the specific information elements are consistent with the matching words, extracting corresponding classification labels from the associated database according to mapping rules;

[0088] S203: Obtaining category labels of marketing products that are not displayed, and comparing the extracted classification labels with the category labels of marketing products that are not displayed;

[0089] S204: Find the undisplayed marketing products whose category labels and classification labels are consistent, and mark them as hot search related products.

[0090] Furthermore, in this embodiment, specific information elements can be extracted from the text content through data mining technology, machine learning algorithm, natural language processing and other technologies. Specifically, data mining technology, machine learning algorithm, natural language processing are all existing mature technologies and will not be further elaborated here.

[0091] As described in the above steps S201 to S204, by building an associated database, mapping and associating specific information elements (such as keywords, matching words, brands, etc.) with classification labels, specific text content can be quickly matched with commodities. For example, the keyword "back to school season" and the classification labels "stationery and sports supplies" and "teaching materials", the keyword "Mother's Day" and the classification labels "middle-aged and elderly clothing", "travel" and "cosmetics", the keyword "Valentine's Day" and the classification labels "film and television entertainment", "cosmetics", "accessories" and "gifts" can all be associated through mapping rules, and then the real-time hot search trends of major social media, news websites and live broadcast platforms are monitored based on the big data platform. status to collect the latest hot search events, extract search keywords, topics and other specific information from the hot search events based on natural language processing technology, extract corresponding classification labels through the associated database, and compare the extracted classification labels with the category labels of the marketing products on display one by one. Finally, the undisplayed marketing products whose category labels can correspond to the classification labels are marked as hot search related products, so that the display order of marketing products can be associated with hot content. By establishing an associated database and mining related products in hot search events, the sales trend of products can be mined in advance, and the lag in sales trend judgment caused by relying on product sales volume and product views in the existing technology can be reduced.

[0092] In a preferred embodiment, outputting the popularity trend value of each undisplayed marketing product according to the acquired historical sales volume of the product, hot search related products and product conversion rate includes the following steps:

[0093] S301: Obtain the historical sales volume and conversion rate of all marketing products that have not been displayed;

[0094] S302: Outputting a sales trend value and a trend adjustment value of the marketing product according to the historical sales volume and the product conversion rate of the product;

[0095] S303: Acquire hot search related products among all marketing products that have not been displayed, and output the hot search trend values ​​of the hot search related products;

[0096] S304: Substitute the sales trend value, hot search trend value and trend adjustment value into a first expression to obtain the popularity trend value of the product. The first expression is: In the formula, f0 represents the popularity trend value of the product, f1 represents the sales trend value of the marketing product, f2 represents the hot search trend value of the marketing product, and f3 represents the trend adjustment value obtained after the last live broadcast of the marketing product.

[0097] Among them, for marketing products that have not been broadcast except for hot search related products, the hot search trend value f2 is 0. The marketing products have not been broadcast during the historical live broadcast, and the trend adjustment value f3 is 0.

[0098] As described in the above steps S301 to S304, in the prior art, the prediction of commodity sales trends mainly relies on the analysis of historical sales and page views over a long period of time. The judgment factors are relatively simple, and sudden events have little impact on historical sales and page views over a long period of time. Therefore, there is an obvious lag in the judgment of sales trends, which in turn affects the operation effect. In this embodiment, on the basis of the original historical sales, the hot search trend value and the trend adjustment value are introduced to comprehensively judge the audience degree of the marketing commodity, and the popularity of the marketing commodity during the previous live broadcast period is judged by the historical sales. The popularity of the hot search event is determined by the heat trend value. The degree of user recognition of the marketing product during the last live broadcast is determined by the trend adjustment value. Multiple factors are combined to make a comprehensive judgment on the sales trend of the marketing product, making the sales trend judgment of the marketing product more accurate. For example, as a well-known sporting goods brand, Hongxing Erke has a relatively stable historical sales volume. However, in the event of Hongxing Erke donating to Henan in 2021, its donation behavior became a hot search event. If only relying on historical sales data (i.e., sales trend value f1), it is impossible to predict that the brand is Hongxing Erke's product. The sales trend will increase significantly in the short term. By introducing the hot search trend value f2, the product is associated with the hot search event. The introduction of the hot search trend value will increase the heat trend value f0, and then it can be quickly determined that its sales trend will increase in the short term. After the live broadcast, the trend adjustment value f3 is obtained according to its conversion rate. The higher conversion rate makes the heat trend value f0 further increase. At the same time, when the donation incident of Hongxing Erke gradually subsides, the sales volume of the marketing products of the brand Hongxing Erke during the live broadcast will gradually decrease until it returns to stability again, and the conversion rate of Hongxing Erke's products will gradually Decreases, the trend adjustment value f3 will also gradually decrease. In this process, by respectively calculating the sales trend value f1, hot search trend value f2 and trend adjustment value f3 of the marketing product, the popularity trend value f0 of the product is obtained, which is convenient for operators to timely obtain the sales trend changes of the product. This improvement, based on the original sales trend value, innovatively introduces the two dimensions of hot search trend value and trend adjustment value, which not only refines the influencing factors of the product sales trend, but also realizes the early prediction of sales trends by associating products with hot search events, thereby effectively reducing the lag of trend judgment and significantly improving operational efficiency.

[0099] In a preferred embodiment, the step of outputting the sales volume trend value of a commodity according to the historical sales volume of the commodity comprises the following steps:

[0100] S310: Setting a weight evaluation interval, wherein there are multiple weight evaluation intervals, and each weight evaluation interval corresponds to a weight factor;

[0101] S311: Taking the current live broadcast start time as the starting node, establishing a backtracking period in reverse chronological order, and setting a plurality of equally spaced backtracking cycles within the backtracking period;

[0102] S312: Obtain the periodic sales volume of the product in each retrospective period, and the total sales volume of the product in the retrospective period;

[0103] S313: Compare the total sales volume of the product with the multiple weight evaluation intervals in sequence to obtain the weight factor corresponding to the total sales volume of the product in the retrospective period;

[0104] S314: Substitute the sales volume of the product in a period and the weight factor into a second expression to obtain the sales volume trend value of the product. The second expression is: Among them, C i represents the sales volume of the product in the i-th retrospective period, C i-1 It represents the sales volume of the product in the i-1th retrospective period, T represents the length of the retrospective period, n represents the number of retrospective periods, δ represents the weight factor corresponding to the product, and i represents the number of the retrospective period.

[0105] As described in the above steps S310 to S314, by setting multiple lookback periods, and counting the sales volume of the product in each lookback period, the sales trend value of the product is obtained through the second expression, wherein when the sales trend value is a positive value, it indicates that the sales trend of the product is on the rise, and when the sales trend value is a negative value, it indicates that the sales trend of the product is on the decline. At the same time, by setting a weight evaluation interval, and obtaining the total sales volume of the product in all lookback periods, and obtaining the corresponding weight factor according to the weight evaluation interval, substituting it into the second expression, the sales trend value of the product is obtained, and the sales trend of the product is comprehensively determined. By obtaining the sales trend value of the marketing product, it can provide a basis for adjusting the exhibition order of the same type of marketing products in the same live broadcast session during the subsequent live broadcast.

[0106] In a preferred embodiment, the step of obtaining hot search related products from all marketing products that have not been displayed, and outputting the hot search trend values ​​of the hot search related products, includes the following steps:

[0107] S320: Taking the time of the hot search event obtained from the hot search list as the starting node, establishing a hot search period, and setting a plurality of hot search cycles at equal intervals within the hot search period until the hot search event is removed from the hot search list;

[0108] S321: Obtain the hot search sales volume of hot search related products in each hot search cycle;

[0109] S322: Substitute the hot search sales volume into the third expression to obtain the hot search trend value of the product. The third expression is: Among them, Kj represents the hot search sales volume of the product in the jth hot search cycle, K j-1 It represents the sales volume of the hot search of the product in the j-1th hot search cycle, L represents the duration of the hot search period, m represents the number of hot search cycles, and i represents the number of the hot search cycle.

[0110] As described in the above steps S320 to S322, hot search events are captured in real time through a big data platform or a media platform. After matching the hot search events and products through an associated database, the time of the hot search event is obtained from the hot search list as the starting node, and a hot search period including multiple hot search cycles is established. The sales volume of the products in each hot search cycle is tracked and obtained. By analyzing the sales volume of the products in multiple hot search cycles, the hot search trend value of the product is obtained, and then it is determined whether the impact of the hot search event on the sales volume of the product is positive or negative. Specifically, if the hot search trend value during the monitoring period is positive, it indicates that the sales volume of the product is on an upward trend, and this hot search event has a positive impact; conversely, if the trend value is negative, it reflects a decline in sales volume of the product, and this hot search event has a negative impact.

[0111] It should be noted that, in this embodiment, a positive event refers to an event containing positive energy or positive meaning, and a negative event refers to an event containing negative energy or negative meaning.

[0112] Furthermore, if the product has no correlation with the hot search event, the hot search trend value of the product is 0.

[0113] In a preferred embodiment, the preference tags of online users during the live broadcast and the marketing coefficients of the marketing products that have not been broadcasted are obtained in real time, and the broadcasting order of the marketing products that have not been broadcasted is dynamically adjusted according to the marketing coefficients and the popularity trend values, including the following steps:

[0114] S401: constructing a matching evaluation interval, wherein there are multiple matching evaluation intervals, and each matching evaluation interval corresponds to a marketing coefficient;

[0115] S402: extracting preference tags of online users during the live broadcast in real time from the user database;

[0116] S403: extracting category labels of marketing products that are not displayed in real time from the product database;

[0117] S404: Compare the category labels of the marketing products that have not been displayed with the preference labels of the online users one by one, count the comparison results, and obtain the number of matches between the marketing products that have not been displayed and the online users;

[0118] S405: Compare the matching quantity with the matching evaluation interval to obtain the marketing coefficient corresponding to the marketing product that has not been displayed;

[0119] S406: Substitute the marketing coefficient and the popularity trend value into a fourth expression to obtain a ranking index of each marketing product that has not been displayed. The fourth expression is: Among them, P represents the ranking index of marketing products that have not been displayed. Indicates the marketing coefficient of the marketing products that are not displayed;

[0120] S407: Arrange the ranking indexes of all marketing products that have not been displayed in descending order to obtain a product display sequence;

[0121] S408: Display the marketing products that have not been displayed according to the arrangement order of the display sequence, wherein the products that have been displayed are removed from the display sequence and are no longer displayed.

[0122] As described in the above steps S401 to S408, during the live broadcast process, the number of users watching the live broadcast and their composition change in real time. Each user has his or her own preferences. In order to ensure that the displayed marketing products can attract online users more widely, the preference tags of all online users need to be captured in real time through the user database, and the preference tags of online users are matched with the category tags of marketing products to obtain the number of matches between marketing products that have not been displayed and online users, and the marketing coefficient corresponding to the marketing product is obtained according to the matching evaluation interval, and the ranking index of the marketing product is calculated according to the marketing coefficient and the heat trend value. During the live broadcast, the display order of the marketing products that have not been displayed is dynamically sorted based on the marketing coefficient and the heat trend value of the marketing products, so that the marketing products that have not been displayed can be matched with more online users, thereby improving the user retention rate and the conversion rate of the marketing products.

[0123] In a specific embodiment, the marketing products that are not displayed during the live broadcast include lipstick and razors. Category labels such as "female" and "cosmetics" are added to lipstick through the product database, and category labels such as "male" and "facial cleansing products" are added to razors. During the live broadcast, the preference labels of online users are obtained in real time, and the category labels of lipstick and razors are compared with the preference labels of online users respectively. For example, if the category label of lipstick can be compatible with the preference labels of 1,000 online users, the number of matches for lipstick is recorded as 1,000, and the category label of razors is recorded as 1,000. If the class label can be adapted to the preference labels of 500 online users, the matching number of razors is recorded as 500. By comparing the matching number and matching evaluation interval of the marketing products to be displayed, the corresponding marketing coefficients of lipstick and razor are obtained respectively. Then, the popularity trend value of lipstick and the popularity trend value of razor are combined respectively to calculate the ranking index of lipstick and the ranking index of razor. If the ranking index of lipstick is greater than the ranking index of razor, the display order of lipstick is adjusted to before razor. Otherwise, the display order of lipstick is adjusted to after razor.

[0124] In another specific embodiment, when the marketing products that are not displayed during the live broadcast include lipstick and razor, when the popularity trends are the same, the product display order is optimized according to the user group. In the early stage of the live broadcast, if the number of female viewers exceeds that of male viewers, the lipstick will have a higher matching number and ranking index, and its ranking index will also surpass the razor. Therefore, the lipstick will be displayed before the razor. In the middle of the live broadcast, if the number of male viewers is dominant and both products are not displayed, the razor will have a higher matching number and ranking index, and its ranking index will surpass the lipstick. The order is dynamically adjusted according to the ranking index of the products, so that the razor is displayed before the lipstick. This dynamic adjustment strategy based on real-time user data changes aims to improve the matching degree between the marketing products that are not displayed and the online users, and optimize the live broadcast effect and operation effect.

[0125] In a preferred embodiment, the acquisition of the operational data after the live broadcast and the trend adjustment value of each marketing commodity, merging the operational data and the trend adjustment value into the commodity database and the user database respectively, and updating the commodity database and the user database includes the following steps:

[0126] S501: Constructing a conversion rate evaluation interval, wherein there are multiple conversion rate evaluation intervals, and each conversion rate evaluation interval corresponds to a trend adjustment value;

[0127] S502: After the current live broadcast ends, the operation data during the current live broadcast is obtained, where the operation data includes the number of clicks, transaction volume, browsing history, and purchase history of each marketing product;

[0128] S503: Preprocessing the operation data and generating corresponding product marketing data and user data;

[0129] S504: Outputting the current conversion rate of the marketing products according to the click volume and transaction volume of each marketing product during the current live broadcast;

[0130] S505: Compare the current conversion rate with the conversion rate evaluation interval to obtain a trend adjustment value corresponding to each marketing commodity;

[0131] S506: Merge the product marketing data, trend adjustment value and user data acquired during the current live broadcast into the product database and the user database respectively, and update the data in the product database and the user database.

[0132] As described in the above steps S501 to S504, after the live broadcast ends, in order to accurately evaluate the live broadcast effect and provide data support for the next live broadcast, it is necessary to evaluate the conversion rate of the goods during the live broadcast and update the heat trend value of the goods according to the evaluation results. The conversion rate evaluation is achieved by setting up multiple conversion rate evaluation intervals, each interval corresponding to a trend adjustment value. By collecting the product sales data after the live broadcast in real time, the conversion rate of each product is calculated, and compared with the conversion rate evaluation interval to obtain the trend adjustment value corresponding to each product. This trend adjustment value will serve as an important basis for updating the heat trend value of the product, ensuring the dynamic and accuracy of the heat trend value of the product.

[0133] In a specific embodiment, if the conversion rate of lipstick during the live broadcast is higher than the upper limit of the preset conversion rate evaluation interval, it means that the lipstick has received extremely high attention during the live broadcast and has been successfully converted into a purchase behavior. Therefore, its corresponding trend adjustment value will be set to a positive value, and the positive value is large, which will cause the heat trend value of the lipstick to be significantly improved. On the contrary, if the conversion rate of the razor is lower than the lower limit of the preset conversion rate evaluation interval, it means that the razor has not attracted enough attention or converted into a purchase behavior during the live broadcast, and its corresponding trend adjustment value will be set to a negative value, which will cause the heat trend value of the razor to decrease. Through such adjustments, the heat trend value of the product will more accurately reflect the actual performance of the product in the live broadcast. At the same time, in order to more comprehensively understand user needs, it is also necessary to update user data based on live broadcast operation data, which includes collecting multi-dimensional data such as user purchase records, browsing records, and evaluation information. Through data analysis technology, extract information such as user purchase preferences and consumption habits, and update this information to the user database, which will help improve the accuracy of the next live broadcast and provide strong support for subsequent product recommendations and personalized services.

[0134] In another specific embodiment, if a product is displayed for the first time, the trend adjustment value of the product is 0.

[0135] In another specific embodiment, if a product is being displayed for the first time, the trend adjustment value of the product may refer to the trend adjustment values ​​of other products of the same type.

[0136] A live broadcast operation system based on big data, applicable to any of the above-mentioned live broadcast operation methods based on big data, comprising:

[0137] A storage module, wherein the storage module is used to store user data and product data;

[0138] Collection module: The collection module is used to collect hot search data and product sales data;

[0139] A processing module, which is used to process relevant data and calculate the display index and conversion rate of the product;

[0140] A feedback module is used to obtain conversion data and user data of the product.

[0141] In this embodiment, the storage module serves as the core storage unit of the system and is responsible for storing a large amount of user data and product data. These data include basic information of users, purchase records, browsing records, evaluation information, etc., as well as basic information, price, sales volume, popularity trend value, etc. of the products. The storage module provides real-time and accurate data support for the system through an efficient data storage and retrieval mechanism; the collection module is responsible for real-time collection of various data, including hot search data and sales data of products, etc. These data reflect the real-time needs of users and the sales situation of products during the live broadcast process, and are an important basis for the system to conduct data analysis and decision-making. The acquisition module realizes fast and accurate data acquisition through crawler technology, API interface and other methods; the processing module is the data processing center of the whole system. It processes and calculates data based on the data acquired by the acquisition module and the user data and product data in the storage module. The processing module first matches the correlation quantity of the product and the real-time user data through the correlation quantity evaluation interval to obtain the matching value of the product; then, it calculates the exhibition index of the product in combination with the popularity trend value of the product; finally, it calculates the conversion rate of the product based on the live broadcast operation data, and updates the popularity trend value of the product based on the conversion rate. The processing module provides accurate and effective data support for the system through a series of complex calculations and analyses; the feedback module is responsible for obtaining feedback information such as product conversion data and user data. These feedback information reflects the effect of live broadcast operation and changes in user needs, and is an important basis for continuous improvement and optimization of the system. The feedback module obtains product sales data, user evaluation and other information after the live broadcast through data interfaces, user feedback and other methods, and feeds this information back to the processing module for subsequent data analysis and decision-making.

[0142] The above is only a preferred embodiment of the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be considered as the protection scope of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention shall be implemented according to the conventional means in the art unless otherwise specified and limited.

Claims

1. A live broadcast operation method based on big data, characterized by: The following steps are involved: Obtain the historical product marketing data and user behavior data during the live broadcast process, and summarize them into a product database and a user database respectively. The historical product marketing data includes product attribute tags, category tags, historical product sales, and product conversion rates. The user behavior data includes user basic attribute tags, behavior tags, and preference tags. Acquire hot search events during the live broadcast in real time, extract specific information elements from the hot search events, find corresponding unbroadcasted marketing products based on the specific information elements, and mark them as hot search related products; Output the popularity trend value of each marketing product that has not been displayed based on the obtained historical sales volume of the product, hot search related products and product conversion rate; Obtaining in real time the preference tags of online users during the live broadcast and the marketing coefficients of the marketing products that have not been broadcasted, and dynamically adjusting the broadcasting order of the marketing products that have not been broadcasted according to the marketing coefficients and the popularity trend values ​​of the products; The operation data after the live broadcast and the trend adjustment value of each marketing commodity are obtained, the operation data and the trend adjustment value are respectively merged into the commodity database and the user database, and the commodity database and the user database are updated.

2. A live broadcast operation method based on big data according to claim 1, characterized in that: Obtaining the historical marketing data of commodities and user behavior data during the live broadcast process, and summarizing them into a commodity database and a user database respectively, wherein the historical marketing data of commodities include commodity attribute tags, category tags, historical sales volume of commodities, commodity conversion rate, and the user behavior data include user basic attribute tags, behavior tags, and preference tags, including the following steps: Extract all product historical marketing data from historical live broadcast records, where the product historical marketing data includes product brand, product name, product attributes, product category, product historical sales volume, and product conversion rate; Extracting user data from historical live broadcast records, wherein the user data includes viewing behavior data, purchase behavior data, and search behavior data of the user; Preprocessing the acquired commodity historical marketing data and user data, and generating corresponding commodity feature tags and user feature tags, and aggregating the commodity historical marketing data, commodity feature tags, user data and user feature tags into a commodity database and a user database respectively; Among them, the commodity feature tags at least include commodity attribute tags and category tags, and the user feature tags at least include user basic attribute tags, behavior tags, and preference tags.

3. A live broadcast operation method based on big data according to claim 1, characterized in that: Acquiring hot search events during a live broadcast in real time, extracting specific information elements from the hot search events, finding corresponding unbroadcasted marketing products according to the specific information elements, and marking them as hot search related products, including the following steps: Constructing an associated database, constructing matching words, classification labels and mapping rules in the associated database, and associating matching words and classification labels according to the mapping rules; Monitor the real-time hot search lists of major social media platforms, news websites, and live broadcast platforms in real time, obtain all hot search events in the hot search lists, extract specific information elements of the hot search events, compare the specific information elements with the matching words in the associated database, and when the specific information elements are consistent with the matching words, extract the corresponding classification labels from the associated database according to the mapping rules; Obtain the category labels of the marketing products that are not displayed, and compare the extracted classification labels with the category labels of the marketing products that are not displayed; Find the undisplayed marketing products whose category labels and classification labels are consistent, and mark them as hot search related products.

4. A live broadcast operation method based on big data according to claim 1, characterized in that: The method of outputting the popularity trend value of each undisplayed marketing product according to the acquired historical sales volume of the product, hot search related products and product conversion rate comprises the following steps: Obtain the historical sales volume and conversion rate of all marketing products that have not been displayed; Outputting a sales trend value and a trend adjustment value of the marketing product according to the historical sales volume and product conversion rate of the product; Obtain hot search related products from all marketing products that have not been displayed, and output the hot search trend values ​​of the hot search related products; Substitute the sales trend value, hot search trend value and trend adjustment value into the first expression to obtain the popularity trend value of the product. The first expression is: In the formula, f0 represents the popularity trend value of the product, f1 represents the sales trend value of the marketing product, f2 represents the hot search trend value of the marketing product, and f3 represents the trend adjustment value obtained after the last live broadcast of the marketing product. Among them, for marketing products that have not been broadcast except for hot search related products, the hot search trend value f2 is 0. The marketing products have not been broadcast during the historical live broadcast, and the trend adjustment value f3 is 0.

5. A live broadcast operation method based on big data according to claim 4, characterized in that: The step of outputting the sales volume trend value of a commodity according to the historical sales volume of the commodity comprises the following steps: Establishing a weight evaluation interval, wherein there are multiple weight evaluation intervals, and each weight evaluation interval corresponds to a weight factor; Taking the current live broadcast start time as the starting node, a backtracking period is established in reverse chronological order, and multiple backtracking cycles with equal intervals are set within the backtracking period; Get the period sales of the product in each retrospective period, as well as the total sales of the product in the retrospective period; Compare the total sales volume of the product with multiple weight evaluation intervals in sequence to obtain the weight factor corresponding to the total sales volume of the product in the retrospective period; Substitute the product cycle sales volume and weight factor into the second expression to obtain the sales volume trend value of the product. The second expression is: Among them, C i represents the sales volume of the product in the i-th retrospective period, C i-1 It represents the sales volume of the product in the i-1th retrospective period, T represents the length of the retrospective period, n represents the number of retrospective periods, δ represents the weight factor corresponding to the product, and i represents the number of the retrospective period.

6. A live broadcast operation method based on big data according to claim 4, characterized in that: The step of obtaining hot search related products from all marketing products that have not been displayed, and outputting the hot search trend values ​​of the hot search related products, includes the following steps: The time when the hot search event is obtained from the hot search list is used as the starting node, a hot search period is established, and multiple hot search cycles with equal intervals are set within the hot search period until the hot search event is removed from the hot search list; Get the hot search sales of hot search related products in each hot search cycle; Substitute the hot search sales volume into the third expression to obtain the hot search trend value of the product. The third expression is: Among them, K j represents the hot search sales volume of the product in the jth hot search cycle, K j-1 It represents the sales volume of the hot search of the product in the j-1th hot search cycle, L represents the duration of the hot search period, m represents the number of hot search cycles, and i represents the number of the hot search cycle.

7. The live broadcast operation method based on big data according to claim 1, characterized in that: Obtaining in real time the preference tags of online users during the live broadcast and the marketing coefficients of the marketing products that have not been broadcasted, and dynamically adjusting the broadcasting order of the marketing products that have not been broadcasted according to the marketing coefficients and the popularity trend values, including the following steps: Constructing a matching evaluation interval, wherein there are multiple matching evaluation intervals, and each matching evaluation interval corresponds to a marketing coefficient; Extract the preference tags of online users during the live broadcast in real time through the user database; Extract category labels of marketing products that are not on display from the product database in real time; Compare the category labels of the marketing products that have not been displayed with the preference labels of online users one by one, count the comparison results, and obtain the number of matches between the marketing products that have not been displayed and the online users; Compare the matching quantity with the matching evaluation interval to obtain the marketing coefficient corresponding to the marketing product that has not been displayed; Substituting the marketing coefficient and the popularity trend value into the fourth expression to obtain the ranking index of each marketing product that has not been displayed, the fourth expression is: Among them, P represents the ranking index of marketing products that have not been displayed. Indicates the marketing coefficient of the marketing products that are not displayed; Arrange the ranking indexes of all marketing products that have not been displayed in descending order to obtain a product display sequence; The marketing products that have not been displayed are displayed according to the arrangement order of the display sequence, among which the products that have been displayed are removed from the display sequence and will not be displayed again.

8. The live broadcast operation method based on big data according to claim 1, characterized in that: The method of obtaining the operation data after the live broadcast and the trend adjustment value of each marketing commodity, merging the operation data and the trend adjustment value into the commodity database and the user database respectively, and updating the commodity database and the user database includes the following steps: Constructing a conversion rate evaluation interval, wherein there are multiple conversion rate evaluation intervals, and each conversion rate evaluation interval corresponds to a trend adjustment value; After the current live broadcast ends, the operation data during the current live broadcast is obtained, and the operation data includes the number of clicks, transaction volume, browsing history, and purchase history of online users of each marketing product; Preprocessing the operational data and generating corresponding product marketing data and user data; Output the current conversion rate of marketing products according to the number of clicks and transaction volume of each marketing product during the current live broadcast; Compare the current conversion rate with the conversion rate evaluation interval to obtain a trend adjustment value corresponding to each marketing commodity; The product marketing data, trend adjustment values ​​and user data obtained during the current live broadcast are respectively merged into the product database and the user database, and the data in the product database and the user database are updated.

9. A live broadcast operation system based on big data, applicable to a live broadcast operation method based on big data as claimed in any one of claims 1 to 8, characterized in that: include: A storage module, wherein the storage module is used to store user data and product data; Collection module: The collection module is used to collect hot search data and product sales data; A processing module, which is used to process relevant data and calculate the display index and conversion rate of the product; A feedback module is used to obtain conversion data and user data of the product.

Images