An intelligent operation platform and method based on global data insight

Abstract

The present invention discloses an intelligent operation platform and method based on global data insight, and relates to the technical field of public opinion monitoring. The platform includes a global data mining module, a data analysis and processing module, a correlation analysis training module, and a demand prediction feedback module that are sequentially connected in communication; the global data mining module is used to collect unstructured public opinion data related to the global public opinion; the data analysis and processing module is used to parse and preprocess the unstructured public opinion data to generate a parsed data set; the correlation analysis training module is used to perform correlation analysis and model training based on the parsed data set to generate a public opinion analysis model; the demand prediction feedback module is used to build a public opinion prediction model and perform public opinion prediction based on demand. The present invention obtains hot content and public opinion data in real time, performs correlation analysis on the processed data, and establishes a public opinion analysis model and a public opinion prediction model, thereby achieving insight and real-time monitoring of global public opinion data.

Description

Technical Field

[0001] The present invention belongs to the technical field of public opinion monitoring, and in particular relates to an intelligent operation platform and method based on global data insight. Background Art

[0002] The development of big data has led to a massive increase in data, which has also led to an increasing amount of public opinion data, resulting in resistance to public opinion monitoring and market insights. Traditional manual monitoring methods cannot adapt to the data changes brought about by big data, so how to achieve comprehensive, efficient and timely real-time monitoring and screening analysis of global data has become an urgent problem to be solved. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent operation platform and method based on global data insight, which obtains hot content and public opinion data in real time, performs correlation analysis on the processed data, and establishes public opinion analysis model and public opinion prediction model, thereby realizing insight and real-time monitoring of global public opinion data.

[0004] The purpose of the present invention can be achieved by the following technical solutions:

[0005] In a first aspect, an embodiment of the present application provides an intelligent operation platform based on global data insight, including a global data mining module, a data analysis and processing module, a correlation analysis training module, and a demand prediction feedback module that are sequentially communicatively connected;

[0006] The global data mining module is used to collect unstructured public opinion data related to global public opinion;

[0007] The data analysis and processing module is used to analyze and preprocess the unstructured public opinion data to generate an analysis data set;

[0008] The association analysis training module is used to perform association analysis and model training according to the parsed data set to generate a public opinion analysis model;

[0009] The demand prediction feedback module is used to build a public opinion prediction model and perform public opinion prediction according to demand;

[0010] dividing the parsed data set into a first data set and a second data set;

[0011] The first data set is used to generate the public opinion analysis model;

[0012] The second data set is used to construct the public opinion prediction model;

[0013] The unstructured public opinion data includes text data, image data and audio and video data; the preprocessing includes text processing, image processing and audio and video processing.

[0014] Preferably, the unstructured public opinion data is parsed and preprocessed to generate a parsed data set; the steps include:

[0015] Perform data cleaning on text data;

[0016] Wherein, the data cleaning includes removing noise, filtering irrelevant information, removing special characters and unifying the format;

[0017] Segment the text data into several vocabulary units and label the corresponding part of speech of the words in each vocabulary unit;

[0018] Perform sentiment analysis on text data through machine learning and sentiment dictionary to obtain emotional attitudes;

[0019] Extract keywords / phrases from text data;

[0020] Perform entity recognition and event tracking on text data;

[0021] Analyze the association relationship between entities in text data;

[0022] The processing information, emotional attitudes and associations in the text data are associated together to generate a text parsing dataset.

[0023] Preferably, a convolutional neural network and a recurrent neural network are combined to construct a CNN-RNN comprehensive model, and image processing is performed on the image data, comprising the following steps:

[0024] Preprocessing the image data, including denoising, scaling, cropping, normalization and edge detection, and using the preprocessed image data as an input image;

[0025] Build the CNN network structure, including convolutional layer, pooling layer, fully connected layer and activation function layer, and set the corresponding parameters;

[0026] Use CNN to extract features from the input image and obtain local feature representation of the input image;

[0027] The feature sequence extracted by CNN is input into RNN for processing to capture the dependencies between sequences and obtain the overall feature representation of the entire input image;

[0028] Build the RNN network structure, including the RNN layer and the Dropout layer;

[0029] Use labeled data to train the CNN-RNN integrated model;

[0030] Among them, the cross entropy loss function and back propagation algorithm are used to update the model parameters, and data enhancement is used to expand the training set during the training process;

[0031] Evaluate the model using the test data;

[0032] Specifically: use accuracy, precision and recall as indicators to evaluate the performance of the CNN-RNN comprehensive model;

[0033] Use the trained CNN-RNN comprehensive model to classify or predict unknown image data.

[0034] Preferably, a Dropout layer is added between the convolutional layer and the fully connected layer of the CNN network structure; and a Dropout layer is added at the connection between the CNN network structure and the RNN network structure.

[0035] Preferably, processing the audio and video data comprises the following steps:

[0036] Extracting audio data and video data from the audio and video data;

[0037] removing noise from the audio data;

[0038] Removing redundant information from the video data; the redundant information includes repeated frames and still images;

[0039] Compressing and encoding the audio and video data;

[0040] Encapsulate the encoded audio and video data;

[0041] Verify the processed audio and video data.

[0042] Preferably, generating the parsed data set comprises the following steps:

[0043] Extracting image text data from the processed image data;

[0044] Extracting audio and video text data from the processed audio and video data;

[0045] The image text data and the audio and video text data are fused into the text parsing data set using multimodal fusion technology to construct the parsing data set.

[0046] Preferably, the multimodal fusion technology is used to construct the analytical data set, comprising the following steps:

[0047] Parse and annotate the dataset

[0048] Extracting image features from the image text data;

[0049] Extracting audio features from the audio and video text data;

[0050] Extracting text features from the text parsing dataset;

[0051] Fuse image features, audio features, and text features to obtain a comprehensive feature vector;

[0052] The parsed data set is generated according to the comprehensive feature vector and parsed and annotated.

[0053] Preferably, performing association analysis based on the parsed data set comprises the following steps:

[0054] Obtaining a comprehensive feature vector in the parsed data set;

[0055] Calculating correlation coefficients between different features in the comprehensive feature vector;

[0056] Test and visualize the correlation coefficients;

[0057] According to the closeness of the correlation with public opinion, the correlation coefficient is divided into first-level correlation coefficient, second-level correlation coefficient, third-level correlation coefficient and fourth-level correlation coefficient;

[0058] Select the features of the first-level correlation coefficient and the features of the second-level correlation coefficient as the training feature set for model training, and select the features of the third-level correlation coefficient and the features of the fourth-level correlation coefficient as the test feature set for model training;

[0059] Among them, the closeness of the correlation coefficient is: first-level correlation coefficient > second-level correlation coefficient > third-level correlation coefficient > fourth-level correlation coefficient.

[0060] Preferably, constructing the public opinion prediction model includes the following steps:

[0061] Acquire the second data set and perform data screening on the second data set to generate a prediction data set;

[0062] Obtaining the output result of the public opinion analysis model;

[0063] Using the prediction data set and the output result as input of the model to generate a public opinion prediction model;

[0064] Perform public opinion prediction according to the public opinion prediction model and generate prediction results;

[0065] Based on the prediction results, a personalized recommendation algorithm is used to perform personalized recommendations.

[0066] In a second aspect, an embodiment of the present application provides an intelligent operation method based on global data insight, which is applied to an intelligent operation platform based on global data insight as described above, and includes the following steps:

[0067] Collect unstructured public opinion data related to global public opinion;

[0068] Parsing and preprocessing the unstructured public opinion data to generate a parsed data set;

[0069] Perform association analysis and model training based on the parsed data set to generate a public opinion analysis model;

[0070] Build a public opinion prediction model and make public opinion predictions based on demand;

[0071] dividing the parsed data set into a first data set and a second data set;

[0072] The first data set is used to generate the public opinion analysis model;

[0073] The second data set is used to construct the public opinion prediction model;

[0074] The unstructured public opinion data includes text data, image data and audio and video data; the preprocessing includes text processing, image processing and audio and video processing.

[0075] The beneficial effects of the present invention are:

[0076] (1) By acquiring hot content and public opinion data in real time, the processed data is subjected to correlation analysis and a public opinion analysis model and a public opinion prediction model are established, thereby achieving insight into and real-time monitoring of global public opinion data. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0078] Figure 1 A schematic diagram of the structure of an intelligent operation platform based on global data insights provided in an embodiment of the present application. DETAILED DESCRIPTION

[0079] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are only examples of methods and systems consistent with some aspects of the present application as detailed in the attached claims.

[0080] The terms used in this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in this article refers to any or all possible combinations of one or more associated listed items.

[0081] The specific implementation methods, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0082] Example 1

[0083] See also Figure 1 , the embodiment of the present application provides an intelligent operation platform based on global data insight, including a global data mining module, a data analysis and processing module, a correlation analysis training module and a demand prediction feedback module which are sequentially communicatively connected;

[0084] The global data mining module is used to collect unstructured public opinion data related to global public opinion;

[0085] The above-mentioned whole domain refers to the entire field related to the brand / enterprise, including but not limited to comments on various e-commerce platforms, forum topics, soft articles on self-media public accounts, and live broadcasts with KOLs.

[0086] The data analysis and processing module is used to analyze and preprocess the public opinion data to generate an analysis data set;

[0087] The association analysis training module is used to perform association analysis and model training based on the preprocessed data set to generate a public opinion analysis model;

[0088] The demand prediction feedback module is used to build a public opinion prediction model and perform public opinion prediction according to demand;

[0089] dividing the parsed data set into a first data set and a second data set;

[0090] The first data set is used to generate the public opinion analysis model;

[0091] The second data set is used to construct the public opinion prediction model;

[0092] The unstructured public opinion data includes text data, image data and audio and video data; the preprocessing includes text processing, image processing and audio and video processing.

[0093] Specifically, in order to gain insights into consumers' evaluations and opinions on brands / enterprises through monitoring public opinion, help enterprises / brands quickly understand online information, and thus comprehensively monitor the company's own product reputation, monitor the products of industry competitors, and monitor the company's own online public opinion, this application first collects global public opinion data in areas related to brands / enterprises, which are mainly unstructured data; secondly, pre-processes the above public opinion data, cleans up missing, duplicated or other data that is irrelevant to subsequent modeling, and then completes the missing data and retrieves the missed data to improve subsequent The efficiency, accuracy and completeness of data analysis and modeling; the pre-processed data is analyzed to extract useful information about market conditions and trends, customer needs, etc., and the associations and hidden connections between the data are mined to provide comprehensive information support for decision-making and model building; the above-mentioned analyzed data are then integrated into a data set, and a public opinion analysis model is established based on the data set; a public opinion prediction model is also constructed through the above-mentioned data, and the model is used to predict public opinion according to specific needs; finally, the above-mentioned public opinion analysis model and public opinion prediction model are used to better assist decision makers in adjusting product operations.

[0094] The above modules will be described in detail below.

[0095] Regarding the global data mining module, the public opinion data based on enterprises / brands contains consumers' evaluation and expectations of brands / enterprises, so the collection and mining of public opinion data is particularly important. This application mainly mines unstructured public opinion data, and unstructured data may include different types of data such as text, images, audio and video. Therefore, it is necessary to classify unstructured data and use different mining and analysis methods for different types of data.

[0096] In an embodiment provided in the present application, the unstructured public opinion data is parsed and preprocessed to generate a parsed data set; the steps include:

[0097] Performing data cleaning on the text data;

[0098] Wherein, the data cleaning includes removing noise, filtering irrelevant information, removing special characters and unifying the format;

[0099] Segmenting the text data into a plurality of vocabulary units and marking corresponding parts of speech for the words in each vocabulary unit;

[0100] Perform sentiment analysis on text data through machine learning and sentiment dictionary to obtain emotional attitudes;

[0101] Extract keywords / phrases from text data;

[0102] Perform entity recognition and event tracking on text data;

[0103] Analyze the association relationship between entities in text data;

[0104] The processing information, emotional attitudes and associations in the text data are associated together to generate a text parsing dataset.

[0105] Furthermore, a CNN-RNN comprehensive model is constructed by combining a convolutional neural network and a recurrent neural network, and image processing is performed on the image data, including the following steps:

[0106] Preprocessing the image data, including denoising, scaling, cropping, normalization and edge detection, and using the preprocessed image data as an input image;

[0107] Build the CNN network structure, including convolutional layer, pooling layer, fully connected layer and activation function layer, and set the corresponding parameters;

[0108] Use CNN to extract features from the input image and obtain local feature representation of the input image;

[0109] The feature sequence extracted by CNN is input into RNN for processing to capture the dependencies between sequences and obtain the overall feature representation of the entire input image;

[0110] Build the RNN network structure, including the RNN layer and the Dropout layer;

[0111] Use labeled data to train the CNN-RNN integrated model;

[0112] Among them, the cross entropy loss function and back propagation algorithm are used to update the model parameters, and data enhancement is used to expand the training set during the training process;

[0113] Evaluate the model using the test data;

[0114] Specifically: use accuracy, precision and recall as indicators to evaluate the performance of the CNN-RNN comprehensive model;

[0115] Use the trained CNN-RNN comprehensive model to classify or predict unknown image data.

[0116] Furthermore, a Dropout layer is added between the convolutional layer and the fully connected layer of the CNN network structure; and a Dropout layer is added at the connection between the CNN network structure and the RNN network structure.

[0117] In an embodiment provided in the present application, processing audio and video data includes the following steps:

[0118] Extracting audio data and video data from the audio and video data;

[0119] removing noise from the audio data;

[0120] Removing redundant information from the video data; the redundant information includes repeated frames and still images;

[0121] Compressing and encoding the audio and video data;

[0122] Encapsulate the encoded audio and video data;

[0123] Verify the processed audio and video data.

[0124] Furthermore, generating the parsed data set comprises the following steps:

[0125] Extracting image text data from the processed image data;

[0126] Extracting audio and video text data from the processed audio and video data;

[0127] The image text data and the audio and video text data are fused into the text parsing data set using multimodal fusion technology to construct the parsing data set.

[0128] Furthermore, the multimodal fusion technology is used to construct the analytical data set, including the following steps:

[0129] Parse and annotate the dataset

[0130] Extracting image features from the image text data;

[0131] Extracting audio features from the audio and video text data;

[0132] Extracting text features from the text parsing dataset;

[0133] Fuse image features, audio features, and text features to obtain a comprehensive feature vector;

[0134] The parsed data set is generated according to the comprehensive feature vector and parsed and annotated.

[0135] Furthermore, association analysis is performed based on the parsed data set, including the following steps:

[0136] Obtaining a comprehensive feature vector in the parsed data set;

[0137] Calculating correlation coefficients between different features in the comprehensive feature vector;

[0138] Test and visualize the correlation coefficients;

[0139] According to the closeness of the correlation with public opinion, the correlation coefficient is divided into first-level correlation coefficient, second-level correlation coefficient, third-level correlation coefficient and fourth-level correlation coefficient;

[0140] Select the features of the first-level correlation coefficient and the features of the second-level correlation coefficient as the training feature set for model training, and select the features of the third-level correlation coefficient and the features of the fourth-level correlation coefficient as the test feature set for model training;

[0141] Among them, the closeness of the correlation coefficient is: first-level correlation coefficient > second-level correlation coefficient > third-level correlation coefficient > fourth-level correlation coefficient;

[0142] That is: the first-level correlation coefficient is a strong correlation coefficient, indicating that the correlation of the degree of closeness is the strongest; the second-level correlation coefficient is a relatively strong correlation coefficient, indicating that the correlation of the degree of closeness is relatively strong; the third-level correlation coefficient is a relatively weak correlation coefficient, indicating that the correlation of the degree of closeness is relatively weak, weaker than the first and second levels, but stronger than the fourth level; the fourth-level correlation coefficient is a weak correlation coefficient, indicating that the correlation of the degree of closeness is the weakest.

[0143] By removing redundant or irrelevant features, the efficiency and accuracy of the model can be improved.

[0144] In an embodiment provided in the present application, constructing the public opinion prediction model includes the following steps:

[0145] Acquire the second data set and perform data screening on the second data set to generate a prediction data set;

[0146] Through the above method, suitable data are screened out as input data of the prediction model.

[0147] Obtaining the output result of the public opinion analysis model;

[0148] Using the prediction data set and the output result as input of the model to generate a public opinion prediction model;

[0149] Perform public opinion prediction according to the public opinion prediction model and generate prediction results;

[0150] Based on the prediction results, a personalized recommendation algorithm is used to perform personalized recommendations.

[0151] To summarize, this application obtains hot content and public opinion data in real time, performs correlation analysis on the processed data, and establishes public opinion analysis models and public opinion prediction models, thereby achieving insight and real-time monitoring of global public opinion data.

[0152] Example 2

[0153] The embodiment of the present application provides an intelligent operation method based on global data insight, which is applied to the intelligent operation platform based on global data insight as described above, and includes the following steps:

[0154] Collect unstructured public opinion data related to global public opinion;

[0155] Parsing and preprocessing the unstructured public opinion data to generate a parsed data set;

[0156] Perform association analysis and model training based on the parsed data set to generate a public opinion analysis model;

[0157] Build a public opinion prediction model and make public opinion predictions based on demand;

[0158] dividing the parsed data set into a first data set and a second data set;

[0159] The first data set is used to generate the public opinion analysis model;

[0160] The second data set is used to construct the public opinion prediction model;

[0161] The unstructured public opinion data includes text data, image data and audio and video data; the preprocessing includes text processing, image processing and audio and video processing.

[0162] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0163] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0164] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0165] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An intelligent operation platform based on global data insights, characterized by: It includes a global data mining module, a data analysis and processing module, a correlation analysis training module and a demand prediction feedback module which are sequentially connected in communication; The global data mining module is used to collect unstructured public opinion data related to global public opinion; The data analysis and processing module is used to analyze and preprocess the unstructured public opinion data to generate an analysis data set; The association analysis training module is used to perform association analysis and model training according to the parsed data set to generate a public opinion analysis model; The demand prediction feedback module is used to build a public opinion prediction model and perform public opinion prediction according to demand; dividing the parsed data set into a first data set and a second data set; The first data set is used to generate the public opinion analysis model; The second data set is used to construct the public opinion prediction model; The unstructured public opinion data includes text data, image data and audio and video data; the preprocessing includes text processing, image processing and audio and video processing; in, Performing association analysis based on the parsed data set includes the following steps: Obtaining a comprehensive feature vector in the parsed data set; Calculating correlation coefficients between different features in the comprehensive feature vector; Test and visualize the correlation coefficients; According to the closeness of the correlation with public opinion, the correlation coefficient is divided into first-level correlation coefficient, second-level correlation coefficient, third-level correlation coefficient and fourth-level correlation coefficient; Select the features of the first-level correlation coefficient and the features of the second-level correlation coefficient as the training feature set for model training, and select the features of the third-level correlation coefficient and the features of the fourth-level correlation coefficient as the test feature set for model training; Among them, the closeness of the correlation coefficient is: first-level correlation coefficient > second-level correlation coefficient > third-level correlation coefficient > fourth-level correlation coefficient.

2. The intelligent operation platform based on global data insight according to claim 1, characterized in that: Parsing and preprocessing the unstructured public opinion data to generate a parsed data set includes the following steps: Perform data cleaning on text data; Wherein, the data cleaning includes removing noise, filtering irrelevant information, removing special characters and unifying the format; Segment the text data into several vocabulary units and label the corresponding part of speech of the words in each vocabulary unit; Perform sentiment analysis on text data through machine learning and sentiment dictionary to obtain emotional attitudes; Extract keywords / phrases from text data; Perform entity recognition and event tracking on text data; Analyze the association relationship between entities in text data; The processing information, emotional attitudes and associations in the text data are associated together to generate a text parsing dataset.

3. The intelligent operation platform based on global data insight according to claim 1, characterized in that: Combining the convolutional neural network and the recurrent neural network to construct a CNN-RNN comprehensive model, and performing image processing on the image data, including the following steps: Preprocessing the image data, including denoising, scaling, cropping, normalization and edge detection, and using the preprocessed image data as an input image; Build the CNN network structure, including convolutional layer, pooling layer, fully connected layer and activation function layer, and set the corresponding parameters; Use CNN to extract features from the input image and obtain local feature representation of the input image; The feature sequence extracted by CNN is input into RNN for processing to capture the dependencies between sequences and obtain the overall feature representation of the entire input image; Build the RNN network structure, including the RNN layer and the Dropout layer; Use labeled data to train the CNN-RNN integrated model; Among them, the cross entropy loss function and back propagation algorithm are used to update the model parameters, and data enhancement is used to expand the training set during the training process; Evaluate the model using the test data; Specifically: use accuracy, precision and recall as indicators to evaluate the performance of the CNN-RNN comprehensive model; Use the trained CNN-RNN comprehensive model to classify or predict unknown image data.

4. The intelligent operation platform based on global data insight according to claim 3 is characterized by: Add a Dropout layer between the convolutional layer and the fully connected layer of the CNN network structure; add a Dropout layer at the connection between the CNN network structure and the RNN network structure.

5. The intelligent operation platform based on global data insight according to claim 3 is characterized by: Processing audio and video data includes the following steps: Extracting audio data and video data from the audio and video data; removing noise from the audio data; Removing redundant information from the video data; the redundant information includes repeated frames and still images; Compressing and encoding the audio and video data; Encapsulate the encoded audio and video data; Verify the processed audio and video data.

6. The intelligent operation platform based on global data insight according to claim 5, characterized in that: Generating the parsed data set includes the following steps: Extracting image text data from the processed image data; Extracting audio and video text data from the processed audio and video data; The image text data and the audio and video text data are fused into a text parsing data set using multimodal fusion technology to construct the parsing data set.

7. The intelligent operation platform based on global data insight according to claim 6 is characterized by: The analysis data set is constructed using multimodal fusion technology, including the following steps: Parse and annotate the dataset Extracting image features from the image text data; Extracting audio features from the audio and video text data; Extracting text features from the text parsing dataset; Fuse image features, audio features, and text features to obtain a comprehensive feature vector; The parsed data set is generated according to the comprehensive feature vector and parsed and annotated.

8. The intelligent operation platform based on global data insight according to claim 1, characterized in that: Constructing the public opinion prediction model includes the following steps: Acquire the second data set and perform data screening on the second data set to generate a prediction data set; Obtaining the output result of the public opinion analysis model; Using the prediction data set and the output result as input of the model to generate a public opinion prediction model; Perform public opinion prediction according to the public opinion prediction model and generate prediction results; Based on the prediction results, a personalized recommendation algorithm is used to perform personalized recommendations.

9. An intelligent operation method based on global data insight, applied to an intelligent operation platform based on global data insight as claimed in any one of claims 1 to 7, characterized in that: The steps include: Collect unstructured public opinion data related to global public opinion; Parsing and preprocessing the unstructured public opinion data to generate a parsed data set; Perform association analysis and model training based on the parsed data set to generate a public opinion analysis model; Build a public opinion prediction model and make public opinion predictions based on demand; dividing the parsed data set into a first data set and a second data set; The first data set is used to generate the public opinion analysis model; The second data set is used to construct the public opinion prediction model; The unstructured public opinion data includes text data, image data and audio and video data; the preprocessing includes text processing, image processing and audio and video processing.

Images