Big data-based live broadcast preference content recommendation method and system

Abstract

The invention discloses a live broadcast preference content recommendation method and system based on big data, relates to the technical field of big data, and aims to solve the problem of low benefit caused by live broadcast recommendation of invalid users. Screening out the live broadcast of the corresponding label according to the live broadcast historical data, classifying the live broadcast into earning live broadcast and public welfare live broadcast, collecting effective user behavior data, classifying the users into promotion type users and transaction type users, carrying out matching recommendation on the promotion type users and the public welfare live broadcast, collecting live broadcast data, and carrying out user recommendation. Live broadcast promotion requirements are analyzed through fuzzy reasoning according to live broadcast data, the live broadcast promotion requirements are low, and transaction type users and revenue live broadcast are matched and recommended. And collecting matched recommendation feedback data, synthesizing the user data and the matched recommendation feedback data to construct a support vector machine to evaluate the user promotion, re-evaluating effective users according to the user promotion, and improving the problem of low live broadcast recommendation benefit.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and more specifically, to a live broadcast preference content recommendation method and system based on big data. Background Art

[0002] Big data technology is a digital technology that includes advantages such as improving business insights, optimizing decision-making, and innovating services. The application of big data technology in live broadcast recommendations can push appropriate live broadcast content to users with higher interests, thereby increasing the benefits of live broadcasts and meeting user needs at the same time.

[0003] The prior art has the following deficiencies:

[0004] In the past, live broadcast push recommended live broadcasts to users based on user preference tags without confirming the promotion of users to live broadcasts, resulting in problems such as large traffic but low benefits in the pushed live broadcast content, thus leading to low efficiency of live broadcast push. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a live broadcast preference content recommendation method and system based on big data, and reasonably plan the live broadcast push by analyzing user activity, live broadcast type, and requirements to solve the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A live broadcast preference content recommendation method based on big data, comprising the following steps:

[0008] Step S1: Collect user data and live broadcast historical data, screen valid users according to user data and obtain user preference tags, screen out the live broadcasts corresponding to the tags in the live broadcast historical data and classify the live broadcast types;

[0009] Step S2: Collect valid user behavior data, determine the user type, and match and recommend public welfare live broadcasts according to the user type;

[0010] Step S3: Collect live broadcast data, use fuzzy inference to analyze the live broadcast promotion requirements, and match and recommend revenue live broadcasts according to the live broadcast promotion requirements;

[0011] Step S4: Collect matching recommendation feedback data, construct a support vector machine to evaluate user promotion by integrating user data and matching recommendation feedback data, and re-evaluate valid users according to user promotion.

[0012] In a preferred embodiment, in step S1, the user data includes user login frequency and user preference tag percentage, and the live broadcast historical data includes the status of the reward function, the number of product link, and the number of online users in the live broadcast room.

[0013] In a preferred embodiment, in step S1, the active threshold is set by using the percentile method according to the user login frequency, and the users whose login frequency exceeds the active threshold are marked as valid users; the live broadcasts are classified into public welfare live broadcasts and revenue live broadcasts by using the binary classification method.

[0014] In a preferred embodiment, in step S2, the valid user behavior data includes user transaction volume and user interaction data, and the user interaction data includes the user live broadcast message volume, the user live broadcast forwarding volume, and the user live broadcast like volume.

[0015] In a preferred embodiment, in step S2, different interaction weights are set for the user live broadcast message volume, the user live broadcast forwarding volume, and the user live broadcast like volume, and the weighted sum is calculated to obtain the user interaction coefficient; when calculating the user interaction coefficient, the control variable method is used for simulation testing to set different interaction weights, and the user transaction volume and the user interaction coefficient are respectively compared with the preset transaction threshold and interaction threshold, and the user types are determined and classified and marked according to the comparison.

[0016] In a preferred embodiment, in step S2, when determining the user type, the users whose transaction volume exceeds the transaction threshold are marked as transactional users; the users whose interaction coefficient exceeds the interaction threshold are marked as promotional users; if the transaction volume of the user exceeds the transaction threshold and the interaction coefficient of the user exceeds the interaction threshold at the same time, then two classification marks are given to it; otherwise, the user is not classified and marked.

[0017] In a preferred embodiment, in step S3, the live broadcast data includes the proportion of the live broadcast room transaction amount and the click-through rate of the goods in the live broadcast room. The specific steps for analyzing the live broadcast promotion requirements by using fuzzy inference are as follows:

[0018] Define the proportion of the live broadcast room transaction amount and the click-through rate of the goods in the live broadcast room as input variables, define the live broadcast promotion requirements as output variables, and divide them into different fuzzy sets;

[0019] Formulate fuzzy rules and obtain the output result through the input variables according to the fuzzy rules;

[0020] Judge the live broadcast promotion requirements according to the output result, and push the live broadcast to users with different classification marks according to the live broadcast promotion requirements.

[0021] In a preferred embodiment, in step S4, the matching recommended feedback data is the change in the number of online users in the live broadcast room and the change in the user login frequency. The specific steps for constructing a support vector machine to evaluate the user promotion ability are as follows:

[0022] Data preparation: The change in the number of online users and the change in the user login frequency are used as input features; the user promotion ability is used as the output result;

[0023] Feature transformation: Select a kernel function to calculate the output result of the support vector machine;

[0024] Model training: Divide the input features into a training set and a test set, and use a loss function to optimize the kernel function through the output results of the support vector machine for the training set and the test set and make the output results converge;

[0025] Evaluation threshold setting: Set the finally converged output result of the support vector machine as the evaluation threshold;

[0026] Evaluate user promotion: Compare the change in the number of online users in the live broadcast room and the user login frequency with the evaluation threshold through the support vector machine to evaluate user promotion.

[0027] A live broadcast preference content recommendation system based on big data, used to implement the above-mentioned live broadcast preference content recommendation method based on big data, including a data collection module, a time update module, a threshold adjustment module, and a feedback adjustment module;

[0028] The data collection module is used to collect all the analysis data involved in the method and send them to different subsequent modules for processing respectively;

[0029] The type discrimination module is used to receive user data and live broadcast history data and screen and classify users and live broadcasts according to the received data;

[0030] The matching recommendation module is used to receive live broadcast data, analyze the live broadcast promotion requirements, and match and recommend the classified users and live broadcasts;

[0031] The feedback update module is used to receive the matching recommendation feedback data, analyze the user promotion, and update the user tags.

[0032] The technical effects and advantages of the live broadcast preference content recommendation method and system based on big data of the present invention:

[0033] The present invention collects user data and live broadcast history data, obtains user favorite tags according to user data and screens out valid users, screens out live broadcasts with corresponding tags according to live broadcast history data and classifies them into revenue live broadcasts and public welfare live broadcasts, collects valid user behavior data, classifies users into promotional users and transactional users, matches and recommends promotional users with public welfare live broadcasts, collects live broadcast data, analyzes the live broadcast promotion requirements according to the live broadcast data. If the live broadcast promotion requirements are low, only match and recommend transactional users with revenue live broadcasts. Collect the matching recommendation feedback data, comprehensively evaluate the user promotion based on the user data and the matching recommendation feedback data, and re-evaluate the valid users according to the user promotion. Description of the drawings

[0034] Figure 1 It is the main schematic diagram of the live broadcast preference content recommendation method based on big data of the present invention,

[0035] Figure 2 This is a schematic sub - figure of the live - preference content recommendation method based on big data according to the present invention.

[0036] Figure 3 This is a flowchart of the live - preference content recommendation system based on big data according to the present invention. Specific implementation manners

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts belong to the scope of protection of the present invention.

[0038] The present invention collects user data and live - broadcast historical data, obtains user - favorite tags according to the user data and screens out valid users, screens out the live broadcasts corresponding to the tags in the live - broadcast historical data and classifies them into revenue - generating live broadcasts and public - welfare live broadcasts, collects valid user behavior data to classify users into promotional users and transactional users, matches and recommends the promotional users with the public - welfare live broadcasts, collects live - broadcast data, analyzes the live - broadcast promotion requirements according to the live - broadcast data. If the live - broadcast promotion requirements are low, only match and recommend the transactional users with the revenue - generating live broadcasts. Collect matching - recommendation feedback data, comprehensively evaluate user promotion based on the user data and the matching - recommendation feedback data, and re - evaluate the valid users according to the user promotion, so as to plan and improve problems such as low live - broadcast efficiency.

[0039] Embodiment 1, a live - preference content recommendation method based on big data, as Figure 1 and Figure 2 shown, includes the following steps:

[0040] Step S1: Collect user data and live - broadcast historical data, screen out valid users according to the user data and obtain user - favorite tags, and screen out the live broadcasts corresponding to the tags in the live - broadcast historical data and classify the live - broadcast types.

[0041] Step S2: Collect valid user behavior data, determine the user type, and match and recommend the public - welfare live broadcasts according to the user type.

[0042] Step S3: Collect live - broadcast data, use fuzzy inference to analyze the live - broadcast promotion requirements, and match and recommend the revenue - generating live broadcasts according to the live - broadcast promotion requirements.

[0043] Step S4: Collect matching - recommendation feedback data, comprehensively construct a support vector machine based on the user data and the matching - recommendation feedback data to evaluate user promotion, and re - evaluate the valid users according to the user promotion.

[0044] The specific implementation is as follows:

[0045] In step S1, the user data includes the user login frequency and the percentage of user favorite tags, and the live broadcast historical data includes the status of the reward function, the number of product link counts, and the number of online users in the live broadcast room. The user data can be accessed from the software management system, and the live broadcast historical data can be obtained from the live broadcast historical database. When screening users, a sufficient number of users are selected as sample users, and the active threshold is set using the percentile method based on the user frequency.

[0046] Before setting the active threshold using the percentile method, the user login frequencies are arranged in ascending order, the threshold percentage is set, and the user login frequency at the threshold percentage sequence is used as the active threshold. For example, if the threshold percentage is set to 70%, the user login frequency exceeding 70% of the sample users is used as the active threshold.

[0047] The users with user login frequencies exceeding the active threshold are marked as valid users, and the percentage of favorite tags of the marked users is collected. The selection parameter is set, and the number of favorite tags within the selection parameter is used as the valid tags. The live broadcast historical data corresponding to the tags is collected. For example, if the selection parameter is set to 3, the top three favorite tags with the highest percentage of favorite tags are used as the valid tags.

[0048] Sufficient live broadcast historical data of valid tag types is collected, and the live broadcast classification threshold is set by integrating the status of the reward function and the number of product link counts in the live broadcast historical data. The formula FL = d × h can be used, where FL is the live broadcast revenue coefficient, d is the reward status variable, and h is the product link coefficient. It should be noted that the reward status variable can be set using binary classification. For example, if the current live broadcast enables the reward function, the reward status variable is set to 1; if the current live broadcast disables the reward function, the reward status variable is set to 0.

[0049] When setting the live broadcast classification threshold, the average value of the product link counts of sufficient valid tag types can be used as the product link coefficient, and the reward status variable is set to 1. The calculated live broadcast revenue coefficient is used as the live broadcast classification threshold; when calculating the live broadcast revenue coefficient for the current live broadcast, the product link coefficient is the number of product links in the current live broadcast room.

[0050] If the calculated live broadcast revenue coefficient of the current live broadcast exceeds the live broadcast classification threshold, the current live broadcast is marked as a revenue live broadcast; if the calculated live broadcast revenue coefficient of the current live broadcast is lower than the live broadcast classification threshold, the current live broadcast is marked as a public welfare live broadcast.

[0051] It should be noted that the number of product links is the number of product links listed in the live broadcast. In public welfare live broadcasts, there are mostly public welfare product links, and the number of such links is relatively small. The greater the number of product links listed in the live broadcast, the greater the corresponding live broadcast revenue. The software management system is a system used to manage and control the development process, which contains user data to be collected. The live broadcast history database stores various live broadcast history data in the software, including the status of the reward function in the historical live broadcast room, the number of product links, and the number of online users. The setting method of the activity threshold and the calculation formula of the live broadcast revenue coefficient are not unique and can be adjusted according to the actual situation.

[0052] In step S2, the user behavior data includes user transaction volume and user interaction data. The user interaction data includes the number of user live broadcast messages, the number of user live broadcast forwards, and the number of user live broadcast likes. After screening and marking users through step S1, sufficient marked user behavior data is collected. When determining the user type based on the marked user behavior data, it is necessary to preprocess the user interaction data. Set different interaction weights for the number of user live broadcast messages, the number of user live broadcast forwards, and the number of user live broadcast likes, and perform weighted summation to calculate the user interaction coefficient. The control variable method can be used for simulation tests when setting different interaction weights.

[0053] Select a period of time as the sample time, and use multiple users to conduct single-item tests on the same live broadcast. Compare the maximum value of the live broadcast online number within the sample time, and then use the calculation result of dividing the maximum value of the live broadcast online number corresponding to the single-item test by the total maximum value of the live broadcast online number of the test live broadcast as the corresponding single-item interaction weight. For example, use multiple users to only give a like to the same live broadcast once without forwarding or leaving a message, and collect the maximum value of the live broadcast online number within the sample time. Similarly, obtain the maximum values of the live broadcast online numbers of the three single-item tests and calculate the corresponding single-item interaction weights respectively.

[0054] Collect sufficient marked user behavior data, calculate the user interaction coefficients of different users respectively, use the percentile method to set the interaction threshold and the transaction threshold, and compare the current user behavior data with the set thresholds to determine the user type. The rules for determining the user type in this example are as follows:

[0055] If the user transaction volume of the current user exceeds the transaction threshold, classify and mark the current user as a transaction-type user; if the user interaction coefficient of the current user exceeds the interaction threshold, classify and mark the current user as a promotion-type user; if the user transaction volume of the current user exceeds the transaction threshold and the user interaction coefficient exceeds the interaction threshold at the same time, then mark the current user with both classification marks; otherwise, do not classify and mark the current user.

[0056] Match and recommend public welfare live broadcasts according to different user tags. If there is a promotional user tag in the current user's tags, push the public welfare live broadcasts to the user according to the percentage of the user's favorite tags; if there is only a transactional user tag in the current user's tags, collect the revenue live broadcast data that needs to be recommended and screen and recommend the revenue live broadcasts.

[0057] It should be noted that the user interaction data exists but is not limited to the three types given in this example. The methods of setting interaction weights, relevant thresholds, calculating user interaction coefficients, and determining user types are not unique, and the above is only for illustrative purposes.

[0058] In step S3, the live broadcast data of the revenue live broadcast includes the proportion of the transaction volume in the live broadcast room and the click-through rate of the goods in the live broadcast room. The proportion of the transaction volume in the live broadcast room can be calculated according to the type of goods.

[0059] Collect the total transaction amount and transaction volume of goods with the same type of label in sufficient quantity. Use the calculation result of dividing the total transaction amount of goods by the transaction volume of goods as the proportion of the transaction volume of the corresponding label goods. Sum and average the calculated proportions of the transaction volumes of all types of label goods hung in the live broadcast room to obtain the proportion of the transaction volume in the live broadcast room.

[0060] It should be noted that the click-through rate of the goods in the live broadcast room, the total transaction amount of goods with the same type of label, and the transaction volume of goods can be obtained by accessing the live broadcast background database. The live broadcast background database is used to store and manage various information and data in the live broadcast system, including the click-through rate of goods, various goods labels, the total transaction amount of goods, and the transaction volume of goods, etc.

[0061] Comprehensively utilize the proportion of the transaction volume in the live broadcast room and the click-through rate of the goods in the live broadcast room to analyze the live broadcast promotion requirements through fuzzy inference. The specific steps are as follows:

[0062] Define the proportion of the transaction volume in the live broadcast room and the click-through rate of the goods in the live broadcast room as input variables, and divide them into different fuzzy sets to analyze the revenue live broadcast.

[0063] For example, define the proportion of the transaction volume in the live broadcast room as "High", "Low", and define the click-through rate of the goods in the live broadcast room as "more", "less".

[0064] Define the live broadcast promotion requirements as the output variable and divide it into fuzzy sets.

[0065] For example, define the live broadcast promotion requirements as "High", "Low".

[0066] Formulate fuzzy rules. Obtain the output result through the input variables passing through the fuzzy rules. Judge the live broadcast promotion requirements of the current revenue live broadcast according to the output result. The formulation of the fuzzy rules can be set according to the actual situation.

[0067] For example, if the live broadcast room transaction amount ratio is marked as J, the click-through rate of goods in the live broadcast room is marked as D, and the live broadcast promotion requirement is marked as T, then it can be defined as

[0068] Rule 1: IF (J is High) AND (D is more) THEN (T is Low)

[0069] Rule 2: IF (J is Low) AND (D is less) THEN (T is High) ...

[0071] According to the formulated fuzzy rules, perform fuzzy reasoning to analyze the live broadcast promotion requirements of the current revenue live broadcast. When the output result is "Low", it is judged that the live broadcast promotion requirements of the current revenue live broadcast are low, and only the current revenue live broadcast is pushed to users with transactional user marks in the user marks; when the output result is "High", it is judged that the live broadcast promotion requirements of the current revenue live broadcast are high, and the current revenue live broadcast is pushed to users with classified marks.

[0072] It should be noted that the division of the fuzzy set can be adjusted according to the actual situation. For example, if the live broadcast room transaction amount ratio of the current revenue live broadcast exceeds 0.6, it is judged that the live broadcast room transaction amount ratio of the current revenue live broadcast is high and marked as "High"; if the click-through rate of goods in the live broadcast room of the current revenue live broadcast exceeds 0.7, it is judged that the click-through rate of goods in the live broadcast room of the current revenue live broadcast is more and marked as "more". When the output result is "Low", the live broadcast room transaction amount ratio and the click-through rate of goods in the live broadcast room of the current revenue live broadcast are relatively high, and the lower the live broadcast promotion demand of the current revenue live broadcast, the more obvious the benefit of pushing to transactional users. In addition, in this example, the fuzzy set is only divided into two types, and the types of fuzzy sets can be adjusted or further divided according to the actual situation to improve the accuracy of the output result.

[0073] In step S4, the matching recommended feedback data is the change in the number of online users in the live broadcast room and the change in the user login frequency. Randomly select the number of online users in the same live broadcast at the time points before and after the matching recommendation and the user login frequency of the users recommended to match the current live broadcast. Subtract the number of online users in the live broadcast room before the matching recommendation from the number of online users in the live broadcast room after the matching recommendation to obtain the change in the number of online users in the live broadcast room; subtract the user login frequency before the matching recommendation from the user login frequency after the matching recommendation to obtain the change in the user login frequency.

[0074] Randomly select multiple sets of time points before and after the matching recommendation to obtain the change in the number of online users in the live broadcast room and the change in the user login frequency, and combine them into an online change dataset and a login change dataset respectively. Use the change in the number of online users in the live broadcast room and the change in the user login frequency as input variables to construct a support vector machine model, and evaluate the user promotion through the output results of the support vector machine model. The specific steps are as follows:

[0075] Data preparation: Use the online change dataset and the login change dataset as input features; use user promotion as the output result.

[0076] Feature transformation: Select the radial basis function as the kernel function to map the input features to a high-dimensional feature space. The expression of the radial basis function is as follows: Where is the output result of the support vector machine, x and are two randomly selected input vectors in the online change dataset and the login change dataset. e is the natural base, and γ is a hyperparameter used to control the width of the kernel function.

[0077] Model training: Step A1, data preprocessing, remove outliers and normalize the input features. The online change dataset or the login change dataset can be sorted from small to large and the mean value and standard deviation are calculated. The calculation result obtained by summing the mean value and two standard deviations is used as the screening coefficient to compare with the data in the dataset, and the data in the dataset that exceeds the screening coefficient is removed. The data can be normalized using the Min-Max normalization method, that is, the data in the online change dataset or the login change dataset is passed through the Min-Max normalization formula: Where, x nore is the result after normalizing the data in the corresponding dataset, x max is the data with the largest value in the corresponding dataset, x min is the data with the smallest value in the corresponding dataset.

[0078] Step A2, divide the training set and the test set. Randomly divide the online change dataset and the login change dataset into the training set and the test set. For example, divide the data in the online change dataset and the login change dataset into the training set and the test set according to a ratio of 7:3. Randomly select a set of data in the training set and the dataset respectively to calculate through the radial basis function to obtain the output result of the support vector machine for the training set and the output result of the support vector machine for the test set.

[0079] Step A3, function optimization. Select the hinge loss function to optimize the support vector machine. The hinge loss formula is as follows: L(y, f(x)) = max(0, 1 - yf(x)), where L(y, f(x)) is the kernel function loss, y is the output result of the support vector machine calculated from the test set, and f(x) is the output result of the support vector machine calculated from the training set. Converge the output result of the support vector machine according to the kernel function loss, and use the converged result as the user promotion evaluation coefficient. Its convergence formula can be: where is the user promotion evaluation coefficient, is the output result of the support vector machine before convergence.

[0080] Setting of the evaluation threshold: Set different hyperparameters and calculate multiple user promotion evaluation coefficients, and take their average value as the evaluation threshold.

[0081] Evaluating user promotion: Take the change in the number of online users in the current live room and the change in the login frequency of the current user who is matched and recommended as inputs. Take the average value of the hyperparameters set when calculating the user promotion evaluation coefficient and obtain the support vector machine calculation result through the radial basis function, and compare it with the evaluation threshold to determine the user promotion.

[0082] When the support vector machine calculation result exceeds the evaluation threshold, it is determined that the user promotion of the current user is relatively high and the matching recommendation effect is good; when the support vector machine calculation result is lower than the evaluation threshold, it is determined that the user promotion of the current user is relatively low, remove all marks of the current user, set the mark update time, and after the mark update time, collect the current user data again and analyze whether the current user is an effective user and mark it according to the user data analysis.

[0083] It should be noted that the kernel function and loss function in the support vector machine are not unique and can be adjusted or switched according to the actual situation. The mark update time is set by professionals in this field and will not be elaborated here.

[0084] Embodiment 2, a live broadcast preference content recommendation system based on big data, as Figure 3 shown, is used to implement the above-mentioned live broadcast preference content recommendation method based on big data, and includes a data collection module, a type discrimination module, a matching recommendation module, and a feedback update module;

[0085] The data collection module is used to collect all the analysis data involved in the method and send them to subsequent different modules for processing respectively;

[0086] The type discrimination module is used to receive user data and live broadcast history data and screen and classify users and live broadcasts according to the received data;

[0087] The matching and recommending module is used to receive live broadcast data analysis, analyze the requirements for live broadcast promotion, and perform matching and recommendation on the classified users and live broadcasts.

[0088] The feedback and update module is used to receive the feedback of the matching and recommendation data analysis, analyze the user motivation, and update the user tags.

[0089] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product.

[0090] Those of ordinary skill in the art can realize that the modules and algorithm steps of each example described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application of the technical solution and the invention constraints. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0091] In addition, the functional modules in each embodiment of this application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0092] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all of them should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0093] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for recommending live broadcast preference content based on big data, characterized in that: The following steps are included: Step S1: Collect user data and live broadcast history data, filter valid users according to user data and obtain user favorite tags, filter live broadcasts with corresponding tags in the live broadcast history data and classify the live broadcast types; Step S2: Collect effective user behavior data, determine user types, and match and recommend charity live broadcasts based on user types; Step S3: Collect live broadcast data, use fuzzy reasoning to analyze live broadcast promotion requirements, and match and recommend profitable live broadcasts according to the live broadcast promotion requirements; Step S4: Collect matching recommendation feedback data, integrate user data and matching recommendation feedback data to construct a support vector machine to evaluate user motivation, and re-evaluate effective users based on user motivation.

2. The method for recommending live broadcast preference content based on big data according to claim 1, characterized in that: In step S1, user data includes user login frequency and percentage of user favorite tags, and live broadcast history data includes reward function status, number of product links and number of online users in the live broadcast room.

3. The method for recommending live broadcast preference content based on big data according to claim 1 or 2, characterized in that: In step S1, an active threshold is set using the percentile method according to the user login frequency, and users whose login frequency exceeds the active threshold are marked as valid users; the live broadcast is classified into public welfare live broadcast and revenue live broadcast using the binary classification method.

4. The method for recommending live broadcast preference content based on big data according to claim 1, characterized in that: In step S2, the effective user behavior data includes user transaction volume and user interaction data, and the user interaction data includes the user live broadcast message volume, user live broadcast forwarding volume and user live broadcast like volume.

5. The method for recommending live broadcast preference content based on big data according to claim 1 or 4, characterized in that: In step S2, different interaction weights are set for the number of user live broadcast messages, the number of user live broadcast forwarding, and the number of user live broadcast likes to perform weighted summation to calculate the user interaction coefficient; When calculating the user interaction coefficient, the control variable method is used to conduct simulation tests to set different interaction weights. The user transaction volume and user interaction coefficient are compared with the preset transaction threshold and interaction threshold respectively. Based on the comparison, the user type is determined and classified.

6. The method for recommending live broadcast preference content based on big data according to claim 1 or 5, characterized in that: In step S2, when determining the user type, users whose transaction volume exceeds the transaction threshold are marked as transaction users; users whose interaction coefficient exceeds the interaction threshold are marked as promotion users; if the user transaction volume exceeds the transaction threshold and the user interaction coefficient exceeds the interaction threshold, they will be marked with both classification marks; otherwise, the user will not be classified.

7. The method for recommending live broadcast preference content based on big data according to claim 1 or 6, characterized in that: In step S3, the live broadcast data includes the transaction volume ratio and the click-through rate of goods in the live broadcast room. The specific steps of using fuzzy reasoning to analyze the live broadcast promotion requirements are as follows: Define the transaction volume ratio and the click-through rate of goods in the live broadcast room as input variables, and define the live broadcast promotion requirements as output variables, and divide them into different fuzzy sets; Formulate fuzzy rules and obtain output results through input variables according to the fuzzy rules; Determine the live broadcast promotion requirements based on the output results, and push the live broadcast to users with different classification tags based on the live broadcast promotion requirements.

8. The method for recommending live broadcast preference content based on big data according to claim 1, characterized in that: In step S4, matching recommendation feedback data is the change in the number of online users in the live broadcast room and the change in the frequency of user logins, and constructing a support vector machine to assess user motivation. The specific steps are as follows: Data preparation: The change in online users and the change in user login frequency are used as input features; Use user motivation as output; Feature conversion: select the kernel function to calculate the output of the support vector machine; Model training: Divide the input features into training set and test set, and use the loss function to optimize the kernel function and converge the output results through the support vector machine output results of the training set and the test set; Evaluation threshold setting: The output result of the final convergence of the support vector machine is set as the evaluation threshold; Assessing user motivation: The change in online users in the live broadcast room and the user login frequency are compared with the assessment threshold through a support vector machine to assess user motivation.

9. A live broadcast preference content recommendation system based on big data, based on the live broadcast preference content recommendation method based on big data according to any one of claims 1 to 8, characterized in that: It includes a data acquisition module, a time update module, a threshold adjustment module and a feedback adjustment module; The data acquisition module is used to collect all analytical data involved in the method and send them to different subsequent modules for processing; The type identification module is used to receive user data and live broadcast history data and filter and classify users and live broadcasts according to the received data; The matching recommendation module is used to receive live broadcast data, analyze live broadcast promotion requirements, and make matching recommendations for classified users and live broadcasts; The feedback update module is used to receive matching recommendation feedback data and analyze user motivation to update user tags.

Images