Methods and devices for predicting social media feature data

Abstract

The embodiment of the present invention provides a method, device and electronic device for estimating social media feature data, belonging to the field of data processing technology, the method comprising: performing data cleaning and formatting operations on a plurality of collected social media data to obtain standard data; arranging the constructed prediction data D={D1, D2, D3} in chronological order, and dividing the arranged prediction data into a training set and a test set, using the training set to train an LSTM model, minimizing the prediction error by adjusting the weight and bias terms, and after the training is completed, inputting new target feature propagation data into the LSTM model, and outputting the target feature propagation prediction value D4 within a preset time period in the future based on the learned pattern; and using a chart library to generate a trend chart of the target feature propagation, a sentiment analysis chart and a user interaction chart. The present invention improves the accuracy of social media feature data estimation.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device and electronic device for estimating social media feature data. Background Art

[0002] The spread of social media feature data usually refers to statistics related to the spread of specific content or topics on social media platforms. These data can be used to measure the influence and coverage of a post, article, video, advertisement or topic on the Internet. Specific indicators of spread may include but are not limited to the following aspects:

[0003] 1. Views / Reads: refers to the number of times the content is visited or displayed.

[0004] 2. Likes: The number of times a user expresses their approval or like for the content.

[0005] 3. Comments: The number of comments or feedback from users on the content.

[0006] 4. Shares / Retweets: The number of times the content is forwarded to others or shared on other platforms by users.

[0007] 5. Saves / Favorites: The number of times a user saves content for later viewing.

[0008] 6. Engagement Rate: This is a ratio calculated by taking into account likes, comments, shares, and other behaviors. It is usually used to measure the degree of user interaction with content.

[0009] 7. Mentions: How often a specific keyword or topic is mentioned.

[0010] The existing technology has the following problems in the process of implementing social media data feature extraction:

[0011] 1. Incomplete data sources: Existing tools can usually only monitor data from a single platform and are unable to integrate target feature dissemination data from multiple social media and news platforms, resulting in limitations in prediction results.

[0012] 2. Lack of predictive capabilities: Most tools can only perform historical data analysis on the target feature propagation volume, lack the ability to predict future trends, and are unable to predict market changes in advance.

[0013] 3. Poor real-time performance: Most existing systems are in offline analysis mode, which cannot capture changes in the propagation volume of target features in real time and are difficult to reflect market dynamics in a timely manner.

[0014] 4. Lack of multi-dimensional analysis: Existing methods mainly rely on simple data aggregation and statistics, and lack comprehensive analysis of multi-dimensional factors (such as emotional tendencies and user interactions) that affect the spread of target features.

[0015] The above problems become technical problems that need to be solved. Summary of the invention

[0016] In view of this, an embodiment of the present invention provides a method, device and electronic device for estimating social media feature data, which at least partially solve the problems existing in the prior art.

[0017] In a first aspect, an embodiment of the present invention provides a method for estimating social media feature data, comprising:

[0018] Perform data cleaning and formatting operations on multiple social media data collected to obtain standard data;

[0019] Count the number of times the target feature is mentioned on each social media platform, use keyword matching and entity recognition technology to identify target feature related information, and summarize to obtain the target feature mention data set D1;

[0020] Based on the results of sentiment analysis, the proportion of positive, negative and neutral evaluations of the target feature among users is calculated, and a sentiment tendency score data set D2 is generated;

[0021] Count the user interaction behaviors of likes, comments, shares, and reposts related to the target feature, and evaluate the user engagement data set D3 of the target feature;

[0022] The core algorithm for the prediction of the target feature propagation trend is set based on the long short-term memory network LSTM. The four parts of the forget gate, input gate, cell state update and output gate are constructed. The constructed prediction data D={D1,D2,D3} are sorted in chronological order, and the sorted prediction data are divided into a training set and a test set. The training set is used to train the LSTM model. The prediction error is minimized by adjusting the weights and bias items. After the training is completed, the new target feature propagation data is input into the LSTM model, and the target feature propagation prediction value D4 in the future preset time period is output based on the learned pattern.

[0023] The chart library is used to generate trend charts, sentiment analysis charts, and user interaction charts of the target feature propagation volume. Customized analysis reports containing historical data, current status, and future trends of the target feature propagation volume are automatically generated. The front-end framework is combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume.

[0024] According to a specific implementation of an embodiment of the present invention, performing data cleaning and formatting operations on the collected multiple social media data to obtain standard data includes:

[0025] After removing duplicate data, invalid data, and noise data from social media data, the data formats of different platforms are unified.

[0026] According to a specific implementation of an embodiment of the present invention, the statistical number of mentions of the target feature on each social media platform, the use of keyword matching and entity recognition technology to identify target feature related information, and the summary to obtain the target feature mention amount data set D1, including:

[0027] Identify a list of keywords related to the target feature, find where the keywords appear in the standard data, and record the number of times each keyword is mentioned on each social media platform;

[0028] Use natural language processing technology to identify entities in the text, determine whether the identified entities are associated with the target feature, and combine the number of entity mentions related to the target feature into the mention count;

[0029] The number of times each keyword and its related entities are mentioned on different platforms is summarized to generate a mention volume data set D1.

[0030] According to a specific implementation of an embodiment of the present invention, the ratio of positive, negative and neutral evaluations of the target feature among users is calculated based on the result of sentiment analysis, and a sentiment tendency score data set D2 is generated, including:

[0031] Load the pre-trained sentiment analysis model, perform sentiment analysis on each content that mentions the target feature, and derive the sentiment polarity of each comment;

[0032] Count the number of positive, negative, and neutral sentiments in all mentions, and calculate the proportion of positive, negative, and neutral sentiments in all mentions;

[0033] According to the ratio of positive, negative and neutral emotions, a sentiment tendency score is assigned to each mentioned content, and the sentiment tendency scores of all mentioned contents are summarized to obtain the final sentiment tendency score data set D2.

[0034] According to a specific implementation of an embodiment of the present invention, the statistics of user interactive behaviors such as likes, comments, shares and forwarding related to the target feature and the user engagement data set D3 for evaluating the target feature include:

[0035] Calculate the total number of likes, comments, shares, and reposts for each mention;

[0036] The engagement score is calculated based on the number of interactions, and the interactive behavior data of each mentioned content is summarized to obtain the user engagement data set D3.

[0037] According to a specific implementation of an embodiment of the present invention, the core algorithm for setting the target feature propagation amount trend prediction based on the long short-term memory network LSTM constructs four parts: a forget gate, an input gate, a cell state update, and an output gate, and arranges the constructed prediction data D={D1, D2, D3} in chronological order, including:

[0038] Arrange D1, D2 and D3 in chronological order to form time series data, and convert the arranged data D = {D1, D2, D3} into the input data of the LSTM model;

[0039] Control the cell state of the previous time step through the forget gate The forgotten information in the forget gate is the current input and the hidden state of the previous time step , the output of the forget gate is :

[0040]

[0041] The input gate is used to determine the new information to be stored in the cell state. The input of the input gate is the current input and the hidden state of the previous time step , the output of the input gate is :

[0042]

[0043] New candidate value It is expressed as:

[0044]

[0045] Update cell status It is expressed as:

[0046]

[0047] The output gate determines the information to be used as the current output, and the input of the output gate is the current input and the hidden state of the previous time step , the output of the output gate It is expressed as:

[0048]

[0049] Hidden state at the current time step It is expressed as:

[0050]

[0051] W f , W i , W C and Wo is the weight matrix, b f 、b i 、b C and b o It represents the bias term and σ is the sigmoid activation function.

[0052] According to a specific implementation method of an embodiment of the present invention, the sorted prediction data is divided into a training set and a test set, the training set is used to train the LSTM model, and the prediction error is minimized by adjusting the weights and bias items. After the training is completed, the new target feature propagation amount data is input into the LSTM model, and the target feature propagation amount prediction value D4 within a preset time period in the future is output based on the learned pattern, including:

[0053] After setting the number of LSTM layers and neurons, create an LSTM model instance and define the dimensions of input and output;

[0054] Select MSE as the loss function, select Adam optimizer, iteratively train the LSTM model on the training set, and adjust the weights and bias terms;

[0055] Convert the new target feature propagation data into the same format as the training data, input the new data into the trained LSTM model, and output the predicted value D4 of the target feature propagation within the future preset time period.

[0056] According to a specific implementation method of an embodiment of the present invention, the trend graph, sentiment analysis graph and user interaction graph of the target feature propagation volume are generated by using a chart library, and a customized analysis report containing historical data, current status and future trends of the target feature propagation volume is automatically generated. The front-end framework is combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume, including:

[0057] Select a preset chart library to generate a trend chart based on the historical data of the target feature's spread to show changes in the time series; generate a pie chart or bar chart based on the sentiment analysis results to show the ratio of positive, negative, and neutral sentiments; generate a bar chart or stacked chart based on user interaction data to show the number of likes, comments, shares, and reposts;

[0058] Use the template engine to create a report template, fill the historical data, current status and future trend forecast values ​​of the target feature propagation volume into the template, and export the generated report to PDF, HTML or Excel format;

[0059] Use the front-end framework to build a Web application interface, including a chart display area and other interactive elements. Use WebSocket to receive real-time data and dynamically update the chart display area to display the real-time changes in the target feature propagation volume and the prediction results.

[0060] In a second aspect, an embodiment of the present invention provides a device for estimating social media feature data, including:

[0061] The standard module cleans and formats the collected data from multiple social media platforms to obtain standard data.

[0062] The aggregation module counts the number of times the target feature is mentioned on various social media platforms, uses keyword matching and entity recognition technology to identify target feature related information, and aggregates the target feature mention data set D1;

[0063] The calculation module calculates the proportion of positive, negative and neutral evaluations of the target feature among users based on the results of sentiment analysis, and generates a sentiment tendency score data set D2;

[0064] A statistical module that counts the likes, comments, shares, and forwarding user interactions related to the target feature, and evaluates the user engagement data set D3 of the target feature;

[0065] The prediction module sets the core algorithm for predicting the trend of target feature propagation based on the long short-term memory network LSTM, constructs four parts: forget gate, input gate, cell state update and output gate, sorts the constructed prediction data D={D1,D2,D3} in chronological order, and divides the sorted prediction data into training set and test set. The training set is used to train the LSTM model, and the prediction error is minimized by adjusting the weights and bias items. After the training is completed, the new target feature propagation data is input into the LSTM model, and the predicted value D4 of the target feature propagation in the future preset time period is output based on the learned pattern;

[0066] The generation module uses the chart library to generate trend charts, sentiment analysis charts and user interaction charts of the target feature propagation volume, automatically generates customized analysis reports containing the historical data, current status and future trends of the target feature propagation volume, and uses the front-end framework combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume.

[0067] In a third aspect, an embodiment of the present invention further provides an electronic device, the electronic device comprising:

[0068] at least one processor; and,

[0069] a memory communicatively connected to the at least one processor; wherein,

[0070] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for estimating social media feature data in any of the first aspects or any of the implementations of the first aspect.

[0071] In a fourth aspect, an embodiment of the present invention further provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the method for estimating social media feature data in the aforementioned first aspect or any implementation of the first aspect.

[0072] In a fifth aspect, an embodiment of the present invention further provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the method for estimating social media feature data in the aforementioned first aspect or any implementation of the first aspect.

[0073] The social media feature data estimation scheme in the embodiment of the present invention includes: performing data cleaning and formatting operations on multiple social media data collected to obtain standard data; counting the number of times a target feature is mentioned on each social media platform, using keyword matching and entity recognition technology to identify target feature related information, and summarizing to obtain a target feature mention quantity data set D1; based on the results of sentiment analysis, calculating the proportion of positive, negative and neutral evaluations of the target feature among users, and generating a sentiment tendency score data set D2; counting the likes, comments, shares and forwarding user interaction behaviors related to the target feature, and evaluating the user participation data set D3 of the target feature; setting the core algorithm for predicting the target feature propagation quantity trend based on the long short-term memory network LSTM, and constructing a forget gate and an input gate , cell state update and output gate, organize the constructed prediction data D={D1,D2,D3} in chronological order, and divide the organized prediction data into training set and test set, use the training set to train the LSTM model, minimize the prediction error by adjusting the weight and bias terms, and after the training is completed, input the new target feature propagation data into the LSTM model, and output the target feature propagation prediction value D4 in the future preset time period based on the learned pattern; use the chart library to generate the target feature propagation trend chart, sentiment analysis chart and user interaction chart, automatically generate a customized analysis report containing the historical data, current status and future trends of the target feature propagation, and use the front-end framework combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation. The present invention has the following beneficial effects:

[0074] aAccurate long-term and short-term trend forecasts:

[0075] The LSTM network can capture long-term dependencies in time series and is suitable for processing data with complex change patterns such as target feature propagation, achieving high-precision long- and short-term trend forecasts.

[0076] bHigh real-time performance:

[0077] The system has real-time data collection and processing capabilities. Combined with the dynamic prediction capabilities of LSTM, it can timely capture changes in the propagation volume of target features and quickly provide prediction results.

[0078] cMulti-dimensional comprehensive analysis:

[0079] By integrating multi-dimensional data such as the number of target feature mentions, sentiment tendencies, and user interactions, the system can more accurately reflect future changes in the target feature's dissemination volume.

[0080] dAutomation and visualization:

[0081] The system automates the entire process from data collection, processing, analysis to report generation, and helps users quickly understand the analysis results through intuitive visualization tools. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0083] Figure 1 A schematic diagram of a method for estimating social media feature data provided by an embodiment of the present invention;

[0084] Figure 2 A schematic diagram of another method for estimating social media feature data provided by an embodiment of the present invention;

[0085] Figure 3 A schematic diagram of the structure of a device for estimating social media feature data provided by an embodiment of the present invention;

[0086] Figure 4 A schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0087] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0088] The following describes the embodiments of the present disclosure through specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0089] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.

[0090] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0091] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0092] The present disclosure provides a method for estimating social media feature data. The method for estimating social media feature data provided in this embodiment can be performed by a computing device, which can be implemented as software, or as a combination of software and hardware, and can be integrated in a server, terminal device, etc.

[0093] See also Figure 1 and Figure 2 The present disclosure provides a method for estimating social media feature data, including:

[0094] S101, performing data cleaning and formatting operations on the collected multiple social media data to obtain standard data.

[0095] You can use API interfaces and web crawler technology to obtain data related to target features in real time from multiple social media platforms as well as news websites, blogs, and forums. The specific steps are as follows:

[0096] Set keywords based on target feature names, product names, competitor information, marketing activities, etc. to ensure comprehensiveness and pertinence of data collection.

[0097] Establish connections with various social media and news platforms through the API interface, set the frequency and scope of data collection, and ensure real-time data inflow.

[0098] The collected raw data is stored in a distributed database (such as Hadoop) to support efficient storage and fast access to large-scale data.

[0099] Clean, format and preprocess the collected data to extract key features related to the propagation of target features. The specific steps are as follows:

[0100] aData cleaning:

[0101] Remove duplicate data, invalid data, and noisy data, such as advertising spam or irrelevant content. Use natural language processing (NLP) technology to segment and denoise text.

[0102] bData formatting:

[0103] Unify the data formats of different platforms, such as standardizing timestamps and user identifiers, to ensure data consistency:

[0104] cSentiment analysis and topic identification:

[0105] Use NLP technology to perform sentiment analysis (positive, negative, neutral classification) and topic identification on text data to extract users' emotional tendencies towards target features and the main topics of discussion.

[0106] S102, counting the number of times the target feature is mentioned on each social media platform, using keyword matching and entity recognition technology to identify target feature related information, and summarizing to obtain a target feature mention quantity data set D1.

[0107] You can define a list of keywords that are relevant to the target feature. Find the locations where the keywords appear in the preprocessed text. Record the number of times each keyword is mentioned on each social media platform.

[0108] You can use natural language processing technology to identify entities in text, such as names of people, places, and organizations. Determine whether the identified entity is associated with the target feature. Combine the number of entity mentions related to the target feature into the mention count.

[0109] The number of times each keyword and its related entities are mentioned on different platforms is summarized to generate a mention volume data set D1.

[0110] S103, based on the result of sentiment analysis, calculate the ratio of positive, negative and neutral evaluations of the target feature among users, and generate a sentiment tendency score data set D2.

[0111] You can choose a suitable sentiment analysis tool or model, such as a rule-based approach, a machine learning model, or a deep learning model (such as BERT, TextBlob, etc.). Load a pre-trained sentiment analysis model. Perform sentiment analysis on each content that mentions the target feature to obtain the sentiment polarity of each comment (positive, negative, or neutral).

[0112] Count the number of positive, negative, and neutral sentiments in all mentions. Calculate the ratio of positive, negative, and neutral sentiments in all mentions.

[0113] The score of positive sentiment is set to be high, the score of negative sentiment is set to be low, and the score of neutral sentiment is set to be medium. According to the ratio of positive, negative and neutral sentiment, a sentiment tendency score is assigned to each mention.

[0114] S104, counting the likes, comments, shares and forwarding user interactive behaviors related to the target feature, and evaluating the user engagement data set D3 of the target feature.

[0115] For each mention, count the total number of likes, comments, shares, and reposts. Calculate an engagement score based on the number of interactions. You can use a simple weighted average or a more complex scoring mechanism. Summarize the interaction behavior data for each mention to get the user engagement data set 𝐷3.

[0116] S105, based on the long short-term memory network LSTM, the core algorithm for predicting the trend of target feature propagation volume is set, and four parts including forgetting gate, input gate, cell state update and output gate are constructed. The constructed prediction data D={D1, D2, D3} are sorted in chronological order, and the sorted prediction data are divided into training set and test set. The training set is used to train the LSTM model, and the prediction error is minimized by adjusting the weights and bias items. After the training is completed, the new target feature propagation volume data is input into the LSTM model, and the predicted value D4 of the target feature propagation volume in the future preset time period is output based on the learned pattern.

[0117] Arrange D1, D2 and D3 in chronological order to form time series data, and convert the arranged data D = {D1, D2, D3} into the input data of the LSTM model;

[0118] Control the cell state of the previous time step through the forget gate The forgotten information in the forget gate is the current input and the hidden state of the previous time step , the output of the forget gate is :

[0119]

[0120] The input gate is used to determine the new information to be stored in the cell state. The input of the input gate is the current input and the hidden state of the previous time step , the output of the input gate is :

[0121]

[0122] New candidate value It is expressed as:

[0123]

[0124] Update cell status It is expressed as:

[0125]

[0126] The output gate determines the information to be used as the current output, and the input of the output gate is the current input and the hidden state of the previous time step , the output of the output gate It is expressed as:

[0127]

[0128] Hidden state at the current time step It is expressed as:

[0129]

[0130] W f , W i , W C and Wo is the weight matrix, b f 、b i 、b C and b o It represents the bias term and σ is the sigmoid activation function.

[0131] After setting the number of LSTM layers and neurons, you can create an LSTM model instance and define the dimensions of the input and output;

[0132] Select MSE as the loss function, select Adam optimizer, iteratively train the LSTM model on the training set, and adjust the weights and bias terms;

[0133] Convert the new target feature propagation data into the same format as the training data, input the new data into the trained LSTM model, and output the predicted value D4 of the target feature propagation within the future preset time period.

[0134] In order to improve the prediction accuracy, the present invention also introduces the following optimization measures:

[0135] Hyperparameter tuning: Optimize the hyperparameters of the LSTM network, such as learning rate, number of hidden layer nodes, batch size, etc., through grid search or random search.

[0136] Data augmentation: Use data augmentation techniques to expand the training dataset to help the model better capture the complex variation patterns of target feature propagation.

[0137] Hybrid model: Combine LSTM with other time series models (such as ARIMA) and integrate the prediction results of different models through weighted averaging to improve the final prediction accuracy.

[0138] S106, using the chart library to generate trend charts, sentiment analysis charts and user interaction charts of the target feature propagation volume, automatically generates customized analysis reports containing historical data, current status and future trends of the target feature propagation volume, and uses the front-end framework combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume.

[0139] Use visualization tools to display the historical data, current status, and future trends of target feature propagation volume, and generate a detailed analysis report. The specific steps are as follows:

[0140] a Data Visualization:

[0141] Use chart libraries (such as D3.js, Plotly) to generate trend graphs of target feature propagation, sentiment analysis graphs, user interaction graphs, etc. to intuitively display the changing trend of target feature propagation.

[0142] b Report generation:

[0143] According to user needs, customized target feature propagation analysis reports are automatically generated, supporting output in multiple formats such as PDF, HTML, Excel, etc., and can be integrated into the company's reporting system.

[0144] c Real-time monitoring dashboard:

[0145] Build a real-time monitoring dashboard, using a front-end framework (such as React or Angular) combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume, and provide decision support for target feature management.

[0146] The present invention adopts LSTM network to predict the trend of target feature propagation volume, which has the following advantages:

[0147] a Accurate long-term and short-term trend forecasts:

[0148] The LSTM network can capture long-term dependencies in time series and is suitable for processing data with complex change patterns such as target feature propagation, achieving high-precision long- and short-term trend forecasts.

[0149] bHigh real-time performance:

[0150] The system has real-time data collection and processing capabilities. Combined with the dynamic prediction capabilities of LSTM, it can timely capture changes in the propagation volume of target features and quickly provide prediction results.

[0151] cMulti-dimensional comprehensive analysis:

[0152] By integrating multi-dimensional data such as the number of target feature mentions, sentiment tendencies, and user interactions, the system can more accurately reflect future changes in the target feature's dissemination volume.

[0153] dAutomation and visualization:

[0154] The system automates the entire process from data collection, processing, analysis to report generation, and helps users quickly understand the analysis results through intuitive visualization tools.

[0155] According to a specific implementation of an embodiment of the present invention, performing data cleaning and formatting operations on the collected multiple social media data to obtain standard data includes:

[0156] After removing duplicate data, invalid data, and noise data from social media data, the data formats of different platforms are unified.

[0157] See also Figure 2 According to a specific implementation of an embodiment of the present invention, the statistical number of mentions of the target feature on each social media platform, the use of keyword matching and entity recognition technology to identify target feature related information, and the summary to obtain the target feature mention amount data set D1, including:

[0158] S201, determining a list of keywords related to the target feature, finding the locations where the keywords appear in the standard data, and recording the number of times each keyword is mentioned on each social media platform;

[0159] S202, using natural language processing technology to identify entities in the text, determining whether the identified entities are associated with the target feature, and combining the number of entity mentions associated with the target feature into the number of mentions;

[0160] S203, summarizing the number of times each keyword and its related entities are mentioned on different platforms to generate a mention quantity data set D1.

[0161] According to a specific implementation of an embodiment of the present invention, the ratio of positive, negative and neutral evaluations of the target feature among users is calculated based on the result of sentiment analysis, and a sentiment tendency score data set D2 is generated, including:

[0162] Load the pre-trained sentiment analysis model, perform sentiment analysis on each content that mentions the target feature, and derive the sentiment polarity of each comment;

[0163] Count the number of positive, negative, and neutral sentiments in all mentions, and calculate the proportion of positive, negative, and neutral sentiments in all mentions;

[0164] According to the ratio of positive, negative and neutral emotions, a sentiment tendency score is assigned to each mentioned content, and the sentiment tendency scores of all mentioned contents are summarized to obtain the final sentiment tendency score data set D2.

[0165] According to a specific implementation of an embodiment of the present invention, the statistics of user interactive behaviors such as likes, comments, shares and forwarding related to the target feature and the user engagement data set D3 for evaluating the target feature include:

[0166] Calculate the total number of likes, comments, shares, and reposts for each mention;

[0167] The engagement score is calculated based on the number of interactions, and the interactive behavior data of each mentioned content is summarized to obtain the user engagement data set D3.

[0168] According to a specific implementation of an embodiment of the present invention, the core algorithm for setting the target feature propagation amount trend prediction based on the long short-term memory network LSTM constructs four parts: a forget gate, an input gate, a cell state update, and an output gate, and arranges the constructed prediction data D={D1, D2, D3} in chronological order, including:

[0169] Arrange D1, D2 and D3 in chronological order to form time series data, and convert the arranged data D = {D1, D2, D3} into the input data of the LSTM model;

[0170] Control the cell state of the previous time step through the forget gate The forgotten information in the forget gate is the current input and the hidden state of the previous time step , the output of the forget gate is :

[0171]

[0172] The input gate is used to determine the new information to be stored in the cell state. The input of the input gate is the current input and the hidden state of the previous time step , the output of the input gate is :

[0173]

[0174] New candidate value It is expressed as:

[0175]

[0176] Update cell status It is expressed as:

[0177]

[0178] The output gate determines the information to be used as the current output, and the input of the output gate is the current input and the hidden state of the previous time step , the output of the output gate It is expressed as:

[0179]

[0180] Hidden state at the current time step It is expressed as:

[0181]

[0182] W f , W i , W C and Wo is the weight matrix, b f , b i , b C and b o It represents the bias term and σ is the sigmoid activation function.

[0183] According to a specific implementation method of an embodiment of the present invention, the sorted prediction data is divided into a training set and a test set, the training set is used to train the LSTM model, and the prediction error is minimized by adjusting the weights and bias items. After the training is completed, the new target feature propagation amount data is input into the LSTM model, and the target feature propagation amount prediction value D4 within a preset time period in the future is output based on the learned pattern, including:

[0184] After setting the number of LSTM layers and neurons, create an LSTM model instance and define the dimensions of input and output;

[0185] Select MSE as the loss function, select Adam optimizer, iteratively train the LSTM model on the training set, and adjust the weights and bias terms;

[0186] Convert the new target feature propagation data into the same format as the training data, input the new data into the trained LSTM model, and output the predicted value D4 of the target feature propagation within the future preset time period.

[0187] According to a specific implementation method of an embodiment of the present invention, the trend graph, sentiment analysis graph and user interaction graph of the target feature propagation volume are generated by using a chart library, and a customized analysis report containing historical data, current status and future trends of the target feature propagation volume is automatically generated. The front-end framework is combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume, including:

[0188] Select a preset chart library to generate a trend chart based on the historical data of the target feature's spread to show changes in the time series; generate a pie chart or bar chart based on the sentiment analysis results to show the ratio of positive, negative, and neutral sentiments; generate a bar chart or stacked chart based on user interaction data to show the number of likes, comments, shares, and reposts;

[0189] Use the template engine to create a report template, fill the historical data, current status and future trend forecast values ​​of the target feature propagation volume into the template, and export the generated report to PDF, HTML or Excel format;

[0190] Use the front-end framework to build a Web application interface, including a chart display area and other interactive elements. Use WebSocket to receive real-time data and dynamically update the chart display area to display the real-time changes in the target feature propagation volume and the prediction results.

[0191] In addition to calculating the predicted value D4 of the target feature propagation amount within the preset time period, the present invention can also evaluate the influence of the target feature to obtain the evaluation value of the target feature, including the following steps S31-S33:

[0192] S31, calculate the features in the standard data through the feature calculation function Gf to obtain the feature data set Kt={Tn,Eq,Ud,Tf}, where Tn represents the number of occurrences of the target feature, Eq represents the emotional tendency, Ud represents the user interaction amount, and Tf represents the dissemination range.

[0193] The standard data is a data set, the target feature can be a keyword, phrase, or other pattern of interest, and Tn represents the total number of times the target feature appears in the standard data.

[0194]

[0195] Where N is the total number of samples in the dataset, di is the i-th sample, and Is an indicator function, which takes value 1 if the condition is true; otherwise it takes value 0.

[0196] Sentiment tendency is usually calculated through sentiment analysis, which can be based on dictionary methods or machine learning models. Suppose there is a sentiment score function , which returns a numeric value representing the sentiment score of a single sample (e.g., between -1 and +1, with negative numbers representing negative sentiment and positive numbers representing positive sentiment).

[0197]

[0198] User interaction can refer to the number of likes, comments, shares, and other indicators of user interaction with content. Represents the number of interactions of the i-th sample.

[0199]

[0200] Reach refers to how many different users have seen or engaged with the content. represents the number of unique users covered by the i-th sample.

[0201]

[0202] If we consider the total number of unique users instead of the number of users per interaction, then:

[0203]

[0204] where Ri is the set of all unique users associated with the i-th sample, and |.| represents the cardinality of the set, i.e., the number of elements in the set.

[0205] S32, calculate the historical data on n social media platforms M={M1,M2,…Mn} through the weight calculation function Gs to obtain the platform weight vector KM=[KM1,KM2,…KMn].

[0206] The weight calculation function Gs is used to determine the weight of each platform based on the historical data of the social media platform, and finally form a platform weight vector KM. These weights can be based on a variety of factors, such as user activity, content dissemination efficiency, influence, etc.

[0207] Each social media platform Mi can be evaluated based on multiple indicators, such as: user activity (such as the number of daily active users), content dissemination speed, average interaction rate (likes, comments, shares, etc.), user growth rate, brand or content exposure, and user satisfaction score.

[0208] In order to ensure the comparability between different indicators, each indicator needs to be standardized. Suppose for an indicator X, its value is , where i represents the i-th platform and j represents the j-th indicator.

[0209] The normalization formula can be:

[0210]

[0211] in are the minimum and maximum values ​​of the j-th indicator for all platforms respectively.

[0212] Assign a weight coefficient w to each indicator j , for each platform Mi, calculate its comprehensive score Si:

[0213]

[0214] Normalize the comprehensive score Si of each platform to get the platform weight KMi:

[0215] .

[0216] S33, using the drift calculation function Gp to calculate the drift feature of the historical data of the target feature on n social media platforms, and obtain the drift weight vector KP=[KP1, KP2, ...KPn].

[0217] Data at two time points t1 and t2 can be obtained, and for each feature F and each platform Mi, the drift value is defined for:

[0218]

[0219] represents the platform at time t The value of the upper feature F;

[0220] For each platform Mi, calculate its comprehensive drift score Pi:

[0221]

[0222] w F is the weight coefficient;

[0223] Quantify the comprehensive drift score Pi of each platform to obtain the drift weight KPi:

[0224] .

[0225] S34, calculating the evaluation value of the target feature based on the feature data set Kt, the platform weight vector KM and the drift weight vector KP.

[0226] For each platform Mi, define weighted feature data as follows:

[0227]

[0228] in is the value of the jth feature in the feature data set Kt on the platform Mi, is the number of days from the current time to the feature data, and α is the time decay factor (0<α<1).

[0229] Set the feature interaction coefficient matrix I to a 4×4 matrix, where represents the interaction coefficient between feature p and feature q. Then the weighted feature data for:

[0230]

[0231] For each platform Mi, calculate its comprehensive feature score Si:

[0232]

[0233] Add up the comprehensive feature scores of all platforms to get the evaluation value E of the target feature:

[0234]

[0235] Corresponding to the above method embodiment, see Figure 3 The embodiment of the present invention further discloses a device 30 for estimating social media feature data, comprising:

[0236] The standard module 301 performs data cleaning and formatting operations on the collected multiple social media data to obtain standard data;

[0237] Aggregation module 302 counts the number of times the target feature is mentioned on each social media platform, uses keyword matching and entity recognition technology to identify target feature related information, and aggregates to obtain a target feature mention quantity data set D1;

[0238] The calculation module 303 calculates the proportion of positive, negative and neutral evaluations of the target feature among users based on the results of the sentiment analysis, and generates a sentiment tendency score data set D2;

[0239] A statistics module 304 is used to count the likes, comments, shares and forwarding user interaction behaviors related to the target feature, and evaluate the user engagement data set D3 of the target feature;

[0240] Prediction module 305 sets the core algorithm for predicting the trend of target feature propagation volume based on long short-term memory network LSTM, constructs four parts including forget gate, input gate, cell state update and output gate, sorts the constructed prediction data D={D1, D2, D3} in chronological order, and divides the sorted prediction data into training set and test set, trains LSTM model with training set, minimizes prediction error by adjusting weight and bias term, inputs new target feature propagation volume data into LSTM model after training, and outputs the predicted value D4 of target feature propagation volume in the future preset time period based on the learned pattern;

[0241] Generation module 306 uses the chart library to generate trend charts, sentiment analysis charts and user interaction charts of the target feature propagation volume, automatically generates a customized analysis report containing the historical data, current status and future trends of the target feature propagation volume, and uses the front-end framework combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume.

[0242] See also Figure 4 The embodiment of the present invention further provides an electronic device 60, the electronic device comprising:

[0243] at least one processor; and,

[0244] a memory communicatively connected to the at least one processor; wherein,

[0245] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for estimating social media feature data in the aforementioned method embodiment.

[0246] An embodiment of the present invention further provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable the computer to execute the aforementioned method embodiment.

[0247] An embodiment of the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the method for estimating social media feature data in the aforementioned method embodiment.

[0248] Reference below Figure 4 , which shows a schematic diagram of the structure of an electronic device 60 suitable for implementing the embodiment of the present disclosure. The electronic device in the embodiment of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 4 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0249] like Figure 4 As shown, the electronic device 60 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage device 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 60 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0250] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, a gage, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device 60 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 4 The electronic device 60 is shown with various devices, but it should be understood that it is not required to implement or have all the devices shown. More or fewer devices may be implemented or have instead.

[0251] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device 609, or installed from a storage device 608, or installed from a ROM 602. When the computer program is executed by the processing device 601, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0252] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for estimating social media feature data, characterized in that: include: Perform data cleaning and formatting operations on multiple social media data collected to obtain standard data; Count the number of times the target feature is mentioned on each social media platform, use keyword matching and entity recognition technology to identify target feature related information, and summarize to obtain the target feature mention data set D1; Based on the results of sentiment analysis, the proportion of positive, negative and neutral evaluations of the target feature among users is calculated, and a sentiment tendency score data set D2 is generated; Count the user interaction behaviors of likes, comments, shares, and reposts related to the target feature, and evaluate the user engagement data set D3 of the target feature; The core algorithm for the prediction of the target feature propagation trend is set based on the long short-term memory network LSTM. The four parts of the forget gate, input gate, cell state update and output gate are constructed. The constructed prediction data D={D1,D2,D3} are sorted in chronological order, and the sorted prediction data are divided into a training set and a test set. The training set is used to train the LSTM model. The prediction error is minimized by adjusting the weights and bias items. After the training is completed, the new target feature propagation data is input into the LSTM model, and the target feature propagation prediction value D4 in the future preset time period is output based on the learned pattern. Use the chart library to generate trend charts, sentiment analysis charts, and user interaction charts of the target feature propagation volume, automatically generate customized analysis reports containing historical data, current status, and future trends of the target feature propagation volume, and use the front-end framework combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume; The method is also used to evaluate the influence of the target feature to obtain an evaluation value of the target feature, including: The features in the standard data are calculated by the feature calculation function Gf to obtain the feature data set Kt={Tn,Eq,Ud,Tf}, where Tn represents the number of occurrences of the target feature, Eq represents the emotional tendency, Ud represents the user interaction amount, and Tf represents the spread range. N is the total number of samples in the data set, di is the i-th sample, is an indicator function. If the condition is true, then The value is 1, otherwise it is 0; in, is the sentiment score function, which is used to represent the sentiment score of a single sample di; use represents the number of interactions of the i-th sample, then: use represents the number of unique users covered by the i-th sample, then: If we consider the total number of unique users instead of the number of users per interaction, then: Where Ri is the set of all unique users associated with the i-th sample, and |.| represents the cardinality of the set, that is, the number of elements in the set; The historical data on n social media platforms M={M1,M2,…Mn} are calculated through the weight calculation function Gs to obtain the platform weight vector KM=[KM1,KM2,…KMn]. Assume that for a certain indicator X, its value is , where i represents the i-th platform and j represents the j-th indicator, then: in are the minimum and maximum values ​​of all platforms on the jth indicator respectively; Assign a weight coefficient w to each indicator j , for each platform Mi, calculate its comprehensive score Si: Normalize the comprehensive score Si of each platform to get the platform weight KMi: The drift calculation function Gp is ​​used to calculate the drift characteristics of the historical data of the target feature on n social media platforms, and the drift weight vector KP=[KP1, KP2, …KPn] is obtained. The data at two time points t1 and t2 are obtained. For each feature F and each platform Mi, the drift value is defined for: Indicates that at time t, the platform The value of the upper feature F; For each platform Mi, calculate its comprehensive drift score Pi: w F is the weight coefficient; Quantify the comprehensive drift score Pi of each platform to obtain the drift weight KPi: Based on the feature data set Kt, the platform weight vector KM and the drift weight vector KP, the evaluation value of the target feature is calculated. For each platform Mi, the weighted feature data is defined as as follows: in is the value of the jth feature in the feature data set Kt on the platform Mi, is the number of days from the time of the feature data to the current time, and α is the time decay factor; Set the feature interaction coefficient matrix I to a 4×4 matrix, where represents the interaction coefficient between feature p and feature q, then the weighted feature data for: For each platform Mi, calculate its comprehensive feature score Si: Add up the comprehensive feature scores of all platforms to get the evaluation value E of the target feature: 。 2. The method according to claim 1, characterized in that The data cleaning and formatting operations are performed on the collected multiple social media data to obtain standard data, including: After removing duplicate data, invalid data, and noise data from social media data, the data formats of different platforms are unified.

3. The method according to claim 2, characterized in that The statistical target feature is mentioned on each social media platform, keyword matching and entity recognition technology are used to identify target feature related information, and the target feature mention quantity data set D1 is obtained by summarizing the data: Identify a list of keywords related to the target feature, find where the keywords appear in the standard data, and record the number of times each keyword is mentioned on each social media platform; Use natural language processing technology to identify entities in the text, determine whether the identified entities are associated with the target feature, and combine the number of entity mentions related to the target feature into the mention count; The number of times each keyword and its related entities are mentioned on different platforms is summarized to generate a mention volume data set D1.

4. The method according to claim 3, characterized in that Based on the result of sentiment analysis, the ratio of positive, negative and neutral evaluations of the target feature among users is calculated, and a sentiment tendency score data set D2 is generated, including: Load the pre-trained sentiment analysis model, perform sentiment analysis on each content that mentions the target feature, and derive the sentiment polarity of each comment; Count the number of positive, negative, and neutral sentiments in all mentions, and calculate the proportion of positive, negative, and neutral sentiments in all mentions; According to the ratio of positive, negative and neutral emotions, a sentiment tendency score is assigned to each mentioned content, and the sentiment tendency scores of all mentioned contents are summarized to obtain the final sentiment tendency score data set D2.

5. The method according to claim 4, characterized in that The data set D3 of user interaction behaviors such as likes, comments, shares and forwarding related to the target feature is used to evaluate the user engagement of the target feature, including: Calculate the total number of likes, comments, shares, and reposts for each mention; The engagement score is calculated based on the number of interactions, and the interactive behavior data of each mentioned content is summarized to obtain the user engagement data set D3.

6. The method according to claim 5, characterized in that The core algorithm for predicting the trend of target feature propagation volume based on the long short-term memory network LSTM is constructed to construct four parts: forget gate, input gate, cell state update and output gate. The constructed prediction data D={D1, D2, D3} is sorted in chronological order, including: Arrange D1, D2 and D3 in chronological order to form time series data, and convert the arranged data D = {D1, D2, D3} into the input data of the LSTM model; Control the cell state of the previous time step through the forget gate The forgotten information in the forget gate is the current input and the hidden state of the previous time step , the output of the forget gate is : The input gate is used to determine the new information to be stored in the cell state. The input of the input gate is the current input and the hidden state of the previous time step , the output of the input gate is : New candidate value It is expressed as: Update cell status It is expressed as: The output gate determines the information to be used as the current output, and the input of the output gate is the current input and the hidden state of the previous time step , the output of the output gate It is expressed as: Hidden state at the current time step It is expressed as: W f , W i , W C and Wo is the weight matrix, b f , b i , b C and b o It represents the bias term and σ is the sigmoid activation function.

7. The method according to claim 6, characterized in that The sorted prediction data is divided into a training set and a test set, and the training set is used to train the LSTM model. The prediction error is minimized by adjusting the weight and bias items. After the training is completed, the new target feature propagation amount data is input into the LSTM model, and the target feature propagation amount prediction value D4 within the future preset time period is output based on the learned pattern, including: After setting the number of LSTM layers and neurons, create an LSTM model instance and define the dimensions of input and output; Select MSE as the loss function, select Adam optimizer, iteratively train the LSTM model on the training set, and adjust the weights and bias terms; Convert the new target feature propagation data into the same format as the training data, input the new data into the trained LSTM model, and output the predicted value D4 of the target feature propagation within the future preset time period.

8. The method according to claim 7, characterized in that The chart library is used to generate a trend chart, sentiment analysis chart, and user interaction chart of the target feature propagation volume, automatically generate a customized analysis report containing the historical data, current status, and future trends of the target feature propagation volume, and use the front-end framework combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume, including: Select a preset chart library to generate a trend chart based on the historical data of the target feature's spread to show changes in the time series; generate a pie chart or bar chart based on the sentiment analysis results to show the ratio of positive, negative, and neutral sentiments; generate a bar chart or stacked chart based on user interaction data to show the number of likes, comments, shares, and reposts; Use the template engine to create a report template, fill the historical data, current status and future trend forecast values ​​of the target feature propagation volume into the template, and export the generated report to PDF, HTML or Excel format; Use the front-end framework to build a Web application interface, including a chart display area and other interactive elements. Use WebSocket to receive real-time data and dynamically update the chart display area to display the real-time changes in the target feature propagation volume and the prediction results.

9. A device for estimating social media feature data, characterized in that: include: The standard module cleans and formats the collected data from multiple social media platforms to obtain standard data. The aggregation module counts the number of times the target feature is mentioned on various social media platforms, uses keyword matching and entity recognition technology to identify target feature related information, and aggregates the target feature mention data set D1; The calculation module calculates the proportion of positive, negative and neutral evaluations of the target feature among users based on the results of sentiment analysis, and generates a sentiment tendency score data set D2; The statistics module counts the likes, comments, shares and forwarding user interaction behaviors related to the target feature, and evaluates the user engagement data set D3 of the target feature; The prediction module sets the core algorithm for predicting the trend of target feature propagation based on the long short-term memory network LSTM, constructs four parts: forget gate, input gate, cell state update and output gate, sorts the constructed prediction data D={D1,D2,D3} in chronological order, and divides the sorted prediction data into training set and test set. The training set is used to train the LSTM model, and the prediction error is minimized by adjusting the weights and bias items. After the training is completed, the new target feature propagation data is input into the LSTM model, and the predicted value D4 of the target feature propagation in the future preset time period is output based on the learned pattern; The generation module uses the chart library to generate trend charts, sentiment analysis charts, and user interaction charts of the target feature propagation volume, automatically generates customized analysis reports containing historical data, current status, and future trends of the target feature propagation volume, and uses the front-end framework combined with WebSocket technology to display the real-time changes and prediction results of the target feature propagation volume; The device is also used to evaluate the influence of the target feature to obtain an evaluation value of the target feature, including: The features in the standard data are calculated by the feature calculation function Gf to obtain the feature data set Kt={Tn,Eq,Ud,Tf}, where Tn represents the number of occurrences of the target feature, Eq represents the emotional tendency, Ud represents the user interaction amount, and Tf represents the spread range. N is the total number of samples in the data set, di is the i-th sample, is an indicator function. If the condition is true, then The value is 1, otherwise it is 0; in, is the sentiment score function, which is used to represent the sentiment score of a single sample di; use represents the number of interactions of the i-th sample, then: use represents the number of unique users covered by the i-th sample, then: If we consider the total number of unique users instead of the number of users per interaction, then: Where Ri is the set of all unique users associated with the i-th sample, and |.| represents the cardinality of the set, that is, the number of elements in the set; The historical data on n social media platforms M={M1,M2,…Mn} are calculated through the weight calculation function Gs to obtain the platform weight vector KM=[KM1,KM2,…KMn]. Assume that for a certain indicator X, its value is , where i represents the i-th platform and j represents the j-th indicator, then: in are the minimum and maximum values ​​of all platforms on the jth indicator respectively; Assign a weight coefficient w to each indicator j , for each platform Mi, calculate its comprehensive score Si: Normalize the comprehensive score Si of each platform to get the platform weight KMi: The drift calculation function Gp is ​​used to calculate the drift characteristics of the historical data of the target feature on n social media platforms, and the drift weight vector KP=[KP1, KP2, …KPn] is obtained. The data at two time points t1 and t2 are obtained. For each feature F and each platform Mi, the drift value is defined for: Indicates that at time t, the platform The value of the upper feature F; For each platform Mi, calculate its comprehensive drift score Pi: w F is the weight coefficient; Quantify the comprehensive drift score Pi of each platform to obtain the drift weight KPi: Based on the feature data set Kt, the platform weight vector KM and the drift weight vector KP, the evaluation value of the target feature is calculated. For each platform Mi, the weighted feature data is defined as as follows: in is the value of the jth feature in the feature data set Kt on the platform Mi, is the number of days from the time of the feature data to the current time, and α is the time decay factor; Set the feature interaction coefficient matrix I to a 4×4 matrix, where represents the interaction coefficient between feature p and feature q, then the weighted feature data for: For each platform Mi, calculate its comprehensive feature score Si: Add up the comprehensive feature scores of all platforms to get the evaluation value E of the target feature: 。 10. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for estimating social media feature data as described in any one of claims 1 to 8.

Images