An opinion hot spot prediction method and system based on big data technology

Abstract

The application discloses a public opinion hotspot prediction method and system based on big data technology, which comprises the following steps: collecting historical public opinion hotspot data and real-time public opinion data of a specified webpage through a crawler technology, marking the historical public opinion hotspot data according to data attributes, and storing the data in a corresponding database; extracting features of all data in the database through a TF-IDF algorithm and a SIFT algorithm to generate a corresponding feature matrix; performing dimension reduction processing on the feature matrix; constructing a public opinion hotspot prediction network, inputting the dimension-reduced feature matrix and corresponding hotspot weight into the public opinion hotspot prediction network, training the public opinion hotspot prediction network by taking a mean square error as a loss function, and stopping the training until the loss function value is the minimum; and predicting a hotspot of real-time public opinion data by using the trained public opinion hotspot prediction network. The application can effectively improve the public opinion hotspot prediction accuracy.

Description

Technical Field

[0001] This invention relates to the technical field of predicting public opinion hotspots, and in particular to a method and system for predicting public opinion hotspots based on big data technology. Background Technology

[0002] With the development of the internet and mobile internet, the way people obtain information has changed dramatically, and the amount of information has increased dramatically, which has brought great challenges to brand marketing. In order to gain an advantage in fierce market competition, companies need to leverage technologies such as big data and public opinion monitoring to improve brand marketing efficiency and better meet consumer needs and expectations.

[0003] Currently, traditional data modeling methods are commonly used to predict online public opinion hotspots, but these methods typically do not consider the correlation analysis of multi-source data, making it difficult to meet the required prediction results. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: using web crawler technology to extract historical and real-time public opinion hotspot data from specified web pages, marking the historical data according to its attributes, and storing it in a corresponding database; wherein the data attributes are three-dimensional attributes, including one-dimensional, two-dimensional, and three-dimensional attributes, the one-dimensional attributes including text, images, and videos, the two-dimensional attributes including events involving people and events involving non-people, and the three-dimensional attributes including the duration of the public opinion hotspot; extracting features from all data in the database using TF-IDF and SIFT algorithms to generate corresponding feature matrices; performing dimensionality reduction on the feature matrices; constructing a public opinion hotspot prediction network, inputting the dimensionality-reduced feature matrices and corresponding hotspot weights into the network, and training the network using mean squared error as a loss function until the loss function value is minimized; the public opinion hotspot prediction network includes a first sub-network, a second sub-network, a third sub-network, and a feedback network; and using the trained network to predict hotspots in real-time public opinion data.

[0006] As a preferred embodiment of the public opinion hotspot prediction method based on big data technology described in this invention, the feature extraction includes: extracting text features using the TF-IDF algorithm; and extracting image / video features using the SIFT algorithm.

[0007] As a preferred embodiment of the public opinion hotspot prediction method based on big data technology described in this invention, the dimensionality reduction process includes: centering the feature matrix and calculating the covariance matrix of the centered feature matrix; performing eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and their corresponding eigenvectors; sorting the eigenvalues ​​according to their magnitude and taking the eigenvectors corresponding to the first n eigenvalues ​​as basis vectors; and multiplying the centered feature matrix by the basis vectors to obtain the dimensionality-reduced feature matrix.

[0008] As a preferred embodiment of the public opinion hotspot prediction method based on big data technology described in this invention, the public opinion hotspot prediction network includes: a first sub-network comprising a first input gate, a first forget gate, a first output gate, and a first attention mechanism layer; a second sub-network comprising a second input gate, a second forget gate, a second output gate, and a second attention mechanism layer; a third sub-network comprising a third input gate, a third forget gate, a third output gate, and a third attention mechanism layer; and a feedback network comprising an input layer, three hidden layers, and an output layer, used to fuse the hotspot weights output by the first, second, and third sub-networks, and to predict public opinion hotspots based on the fusion result, using the ELU activation function to output the prediction result.

[0009] As a preferred embodiment of the public opinion hotspot prediction method based on big data technology described in this invention, the hotspot weights include: defining the hotspot weight of the first attribute as x, the hotspot weight of the second attribute as y, and the hotspot weight of the third attribute as z.

[0010] S = algx + blgy 2 +clgz 3

[0011] Where a, b, and c are weighting coefficients, and S is the fusion result.

[0012] As a preferred embodiment of the public opinion hotspot prediction method based on big data technology described in this invention, the loss function includes:

[0013]

[0014] Where L is the loss function value, n is the number of feature matrices after dimensionality reduction, and y i Let y be the actual output corresponding to the i-th dimension-reduced feature matrix. ` i This is the predicted output corresponding to the i-th dimension-reduced feature matrix.

[0015] As a preferred embodiment of the public opinion hotspot prediction system based on big data technology described in this invention, the system includes: a data processing module configured to crawl historical and real-time public opinion hotspot data from specified web pages using web crawling technology, mark the historical data according to data attributes, and store it in a corresponding database; wherein the data attributes are three-dimensional attributes, including one-dimensional, two-dimensional, and three-dimensional attributes, the one-dimensional attributes including text, images, and videos, the two-dimensional attributes including events involving people and events involving non-people, and the three-dimensional attributes including the duration of the public opinion hotspot; a feature extraction module configured to extract features from all data in the database using TF-IDF and SIFT algorithms to generate a corresponding feature matrix; a data dimensionality reduction module configured to perform dimensionality reduction processing on the feature matrix; a network training module configured to construct a public opinion hotspot prediction network, input the dimensionality-reduced feature matrix and corresponding hotspot weights into the network, train the network using mean squared error as a loss function until the loss function value is minimized; and a public opinion hotspot prediction module configured to use the trained network to predict hotspots in real-time public opinion data.

[0016] As a preferred embodiment of the public opinion hotspot prediction system based on big data technology described in this invention, the feature extraction module is specifically configured to extract text features using the TF-IDF algorithm and extract image / video features using the SIFT algorithm.

[0017] As a preferred embodiment of the public opinion hotspot prediction system based on big data technology described in this invention, the data dimensionality reduction module is specifically configured to perform centering processing on the feature matrix and calculate the covariance matrix of the centered feature matrix; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and their corresponding eigenvectors; sort the eigenvalues ​​according to their magnitude and take the eigenvectors corresponding to the first n eigenvalues ​​as basis vectors; multiply the centered feature matrix by the basis vectors to obtain the dimensionality-reduced feature matrix.

[0018] As a preferred embodiment of the public opinion hotspot prediction system based on big data technology described in this invention, the network training module is specifically configured to construct a public opinion hotspot prediction network, which includes a first sub-network, a second sub-network, a third sub-network, and a feedback network. The first sub-network includes a first input gate, a first forget gate, a first output gate, and a first attention mechanism layer. The second sub-network includes a second input gate, a second forget gate, a second output gate, and a second attention mechanism layer. The third sub-network includes a third input gate, a third forget gate, a third output gate, and a third attention mechanism layer. The feedback network includes an input layer, three hidden layers, and an output layer, used to fuse the hotspot weights output by the first, second, and third sub-networks, and to predict public opinion hotspots based on the fusion result, using the ELU activation function to output the prediction result.

[0019] The beneficial effects of this invention are as follows: By introducing an attention mechanism, the LSTM network can quickly filter target information from massive amounts of data, improving the efficiency and accuracy of task processing. At the same time, by fusing the Sigmoid and ReLU activation functions, the feedback network has good robustness, effectively improving the accuracy of predicting public opinion hotspots. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0021] Figure 1 This is a flowchart illustrating the public opinion hotspot prediction method based on big data technology described in the first embodiment of the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] This invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of this invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not adhering to the usual scale. Furthermore, the schematic diagrams are merely examples and should not be construed as limiting the scope of protection of this invention. In actual fabrication, the three-dimensional spatial dimensions of length, width, and depth should be included.

[0026] Furthermore, in the description of this invention, it should be noted that the terms "upper," "lower," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are used solely for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. In addition, the terms "first," "second," or "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0027] Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" in this invention should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; similarly, they can refer to mechanical connections, electrical connections, or direct connections, or indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0028] Example 1

[0029] Reference Figure 1 This is the first embodiment of the present invention, which provides a method for predicting public opinion hotspots based on big data technology, including:

[0030] S1: Use web crawling technology to capture historical and real-time public opinion data from specified web pages, and mark the historical public opinion data according to data attributes and store it in the corresponding database.

[0031] The data attributes are three-dimensional, including one-dimensional, two-dimensional and three-dimensional attributes. One-dimensional attributes include text, images and videos; two-dimensional attributes include events involving people and events involving non-people; and three-dimensional attributes include the duration of public opinion hotspots.

[0032] Specifically, there are six types of tags: text + person / event + duration of public opinion hotspot, text + non-person / event + duration of public opinion hotspot, video + person / event + duration of public opinion hotspot, video + non-person / event + duration of public opinion hotspot, image + person / event + duration of public opinion hotspot, and image + non-person / event + duration of public opinion hotspot.

[0033] At the same time, six databases were established to store six types of historical public opinion hotspot data.

[0034] S2: Use the TF-IDF and SIFT algorithms to extract features from all data in the database and generate the corresponding feature matrix.

[0035] (1) Extract text features using the TF-IDF algorithm;

[0036] TF-IDF (term frequency–inverse document frequency) is a commonly used weighting technique for information retrieval and text mining. The main idea behind TF-IDF is that if a word or phrase has a high TF frequency in one document but rarely appears in other documents, it is considered to have good class-discriminating ability and is suitable for classification.

[0037] (2) Extract image / video features using the SIFT algorithm.

[0038] It should be noted that SIFT (Scale-invariant feature transform) is an algorithm for detecting local features. This algorithm obtains features by finding the feature points (interest points, or corner points) in an image and their related scale and orientation descriptors, and then performs image feature point matching. It has good uniqueness and rich information content, and is suitable for fast and accurate matching in massive feature databases.

[0039] S3: Perform dimensionality reduction on the feature matrix.

[0040] The feature matrix is ​​centered by taking the mean of each column to obtain the row vectors, and then subtracting the row vectors from the feature matrix (due to Python's broadcasting mechanism, the row vectors are expanded into a matrix with the same dimensions as the feature matrix) to complete the centering process.

[0041] Calculate the covariance matrix of the eigenvalue matrix after centering.

[0042] The covariance matrix is ​​decomposed into eigenvalues ​​and their corresponding eigenvectors.

[0043] Sort the eigenvalues ​​according to their magnitudes, and take the eigenvectors corresponding to the first k eigenvalues ​​as the basis vectors;

[0044] Multiply the centered feature matrix by the basis vectors to obtain the dimensionality-reduced feature matrix.

[0045] This invention uses principal component analysis for feature dimensionality reduction, which can effectively improve the model calculation speed.

[0046] S4: Construct a public opinion hotspot prediction network. Input the dimensionality-reduced feature matrix and the corresponding hotspot weights into the public opinion hotspot prediction network. Use the mean squared error as the loss function to train the public opinion hotspot prediction network until the loss function value is minimized and then stop training.

[0047] The public opinion hotspot prediction network includes a first sub-network, a second sub-network, a third sub-network, and a feedback network.

[0048] Specifically, the first sub-network includes a first input gate, a first forget gate, a first output gate, and a first attention mechanism layer; the second sub-network includes a second input gate, a second forget gate, a second output gate, and a second attention mechanism layer; the third sub-network includes a third input gate, a third forget gate, a third output gate, and a third attention mechanism layer; the hotspot weights x, y, and z of the first, second, and third attributes are predicted respectively through the first, second, and third sub-networks; preferably, the introduction of the attention mechanism enables the LSTM network to quickly filter target information from massive amounts of data, thereby improving the prediction accuracy of the hotspot weights.

[0049] The feedback network consists of an input layer, three hidden layers, and an output layer. It is used to fuse the hotspot weights output from the first, second, and third sub-networks, and to predict public opinion hotspots based on the fusion results. The prediction results are output using the ELU activation function.

[0050] It should be noted that the ELU activation function combines the Sigmoid and ReLU activation functions. The output mean of ELU is close to zero. By reducing the influence of bias offset, the normal gradient is made closer to the unit natural gradient, thereby accelerating the learning process with the mean approaching zero and increasing the computational intensity, which effectively improves the computational efficiency of the feedback network.

[0051] Define the hotspot weight of the first attribute as x, the hotspot weight of the second attribute as y, and the hotspot weight of the third attribute as z;

[0052] S = algx + blgy 2 +clgz 3

[0053] Where a, b, and c are weighting coefficients, and the values ​​of a, b, and c are in the range of (0, 1). S is the fusion result.

[0054] Furthermore, the mean squared error is used as the loss function to train the public opinion hotspot prediction network until the loss function value is minimized, at which point training stops. The loss function is as follows:

[0055]

[0056] Where L is the loss function value, n is the number of feature matrices after dimensionality reduction, and y i Let y be the actual output corresponding to the i-th dimension-reduced feature matrix. ` i This is the predicted output corresponding to the i-th dimension-reduced feature matrix.

[0057] S5: Use the trained public opinion hotspot prediction network to predict hotspots in real-time public opinion data.

[0058] Real-time public opinion hotspot data is input into the trained public opinion hotspot prediction network. If the fusion result corresponding to the real-time public opinion hotspot data is greater than or equal to the fusion result corresponding to the historical public opinion hotspot data, then the data is determined to be a public opinion hotspot.

[0059] Example 2

[0060] This embodiment differs from the first embodiment in that it provides a public opinion hotspot prediction system based on big data technology, including:

[0061] The data processing module is configured to use web crawling technology to retrieve historical and real-time public opinion data from specified web pages, label the data according to its attributes, and store it in the corresponding database. The data attributes are three-dimensional, including one-dimensional, two-dimensional, and three-dimensional attributes. One-dimensional attributes include text, images, and videos; two-dimensional attributes include events involving people and events involving non-people; and three-dimensional attributes include the duration of public opinion hotspots.

[0062] The feature extraction module is configured to extract features from all data in the database using the TF-IDF and SIFT algorithms, and generate the corresponding feature matrix.

[0063] The data dimensionality reduction module is configured to perform dimensionality reduction processing on the feature matrix.

[0064] The network training module is configured to construct a public opinion hotspot prediction network. The dimensionality-reduced feature matrix and the corresponding hotspot weights are input into the public opinion hotspot prediction network. The mean squared error is used as the loss function to train the public opinion hotspot prediction network until the loss function value is minimized and then training stops.

[0065] The public opinion hotspot prediction module is configured to use the trained public opinion hotspot prediction network to predict new public opinion hotspot data.

[0066] Furthermore, the feature extraction module is specifically configured to extract text features using the TF-IDF algorithm and extract image / video features using the SIFT algorithm.

[0067] Furthermore, the data dimensionality reduction module is specifically configured to perform centering processing on the feature matrix and calculate the covariance matrix of the centered feature matrix; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and their corresponding eigenvectors; sort the eigenvalues ​​according to their magnitude and take the eigenvectors corresponding to the first n eigenvalues ​​as basis vectors; multiply the centered feature matrix by the basis vectors to obtain the dimensionality-reduced feature matrix.

[0068] Furthermore, the network training module is specifically configured to construct a public opinion hotspot prediction network, which includes a first subnetwork, a second subnetwork, a third subnetwork, and a feedback network. The first subnetwork includes a first input gate, a first forget gate, a first output gate, and a first attention mechanism layer. The second subnetwork includes a second input gate, a second forget gate, a second output gate, and a second attention mechanism layer. The third subnetwork includes a third input gate, a third forget gate, a third output gate, and a third attention mechanism layer. The feedback network includes an input layer, three hidden layers, and an output layer, used to fuse the hotspot weights output from the first, second, and third subnetworks, and to predict public opinion hotspots based on the fusion result, using the ELU activation function to output the prediction result.

[0069] It should be recognized that embodiments of the present invention can be implemented or carried out by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable storage medium. The method can be implemented using standard programming techniques—including implementation in a computer program on a non-transitory computer-readable storage medium configured with a computer program, wherein such a storage medium causes the computer to operate in a specific and predefined manner—according to the methods and drawings described in the specific embodiments. Each program can be implemented in a high-level procedural or object-oriented programming language to communicate with the computer system. However, if desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. Furthermore, for this purpose, the program can run on a programmed application-specific integrated circuit (ASIC).

[0070] Furthermore, the procedures described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by the context. The procedures described herein (or variations and / or combinations thereof) may be executed under the control of one or more computer systems configured with executable instructions, and may be implemented by hardware or a combination thereof as code (e.g., executable instructions, one or more computer programs, or one or more applications) that commonly executes on one or more processors. The computer program comprises a plurality of instructions executable by one or more processors.

[0071] Furthermore, the method can be implemented in any suitable type of computing platform, including but not limited to personal computers, minicomputers, mainframes, workstations, networked or distributed computing environments, standalone or integrated computer platforms, or in communication with charged particle tools or other imaging devices, etc. Aspects of the invention can be implemented as machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, optical read and / or write storage medium, RAM, ROM, etc., such that it is readable by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the processes described herein. Furthermore, the machine-readable code, or portions thereof, can be transmitted via wired or wireless networks. The invention described herein includes these and other different types of non-transitory computer-readable storage media when such media comprises instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor. When programmed according to the methods and techniques described herein, the invention also includes the computer itself. A computer program can be applied to input data to perform the functions described herein, thereby transforming the input data to generate output data stored in non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the invention, the converted data represents physical and tangible objects, including specific visual depictions of physical and tangible objects generated on a display.

[0072] As used herein, the terms “component,” “module,” “system,” etc., are intended to refer to a computer-related entity, which may be hardware, firmware, a combination of hardware and software, software, or running software. For example, a component may be, but is not limited to, a process running on a processor, a processor, an object, an executable file, a running thread, a program, and / or a computer. As an example, an application running on a computing device and the computing device itself can both be components. One or more components may reside in a running process and / or thread, and components may be located in a single computer and / or distributed among two or more computers. Furthermore, these components are capable of execution from various computer-readable media having various data structures thereon. These components may communicate locally and / or remotely via signals, such as based on one or more data packets (e.g., data from a component that interacts with a local system, another component in a distributed system, and / or signals that interact with other systems via a network such as the Internet).

[0073] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting public opinion hotspots based on big data technology, characterized in that, include: The web crawler technology is used to extract historical and real-time public opinion data from specified web pages, and the historical public opinion data is tagged according to data attributes and stored in the corresponding database. The data attributes are three-dimensional attributes, including one-dimensional, two-dimensional and three-dimensional attributes. The one-dimensional attributes include text, images and videos. The two-dimensional attributes include events involving people and events involving non-people. The three-dimensional attributes include the duration of the public opinion hotspot. The TF-IDF and SIFT algorithms are used to extract features from all data in the database and generate the corresponding feature matrix. The feature matrix is ​​then subjected to dimensionality reduction processing; A public opinion hotspot prediction network is constructed. The dimensionality-reduced feature matrix and the corresponding hotspot weights are input into the public opinion hotspot prediction network. The mean squared error is used as the loss function to train the public opinion hotspot prediction network until the loss function value is minimized and then training stops. The public opinion hotspot prediction network includes a first sub-network, a second sub-network, a third sub-network, and a feedback network. The trained public opinion hotspot prediction network is used to predict hotspots in real-time public opinion data. The aforementioned public opinion hotspot prediction network includes: The first sub-network includes a first input gate, a first forget gate, a first output gate, and a first attention mechanism layer; The second sub-network includes a second input gate, a second forget gate, a second output gate, and a second attention mechanism layer; The third sub-network includes a third input gate, a third forget gate, a third output gate, and a third attention mechanism layer; The feedback network includes an input layer, three hidden layers, and an output layer. It is used to fuse the hotspot weights output by the first sub-network, the second sub-network, and the third sub-network, and to predict public opinion hotspots based on the fusion result. The prediction result is output using the ELU activation function. The hotspot weights include: Define the hotspot weight of the first attribute as x, the hotspot weight of the second attribute as y, and the hotspot weight of the third attribute as z; S=algx+blgy 2 +clgz 3； Where a, b, and c are weighting coefficients, and S is the fusion result.

2. The method for predicting public opinion hotspots based on big data technology as described in claim 1, characterized in that, The feature extraction includes: Text features are extracted using the TF-IDF algorithm; image / video features are extracted using the SIFT algorithm.

3. The method for predicting public opinion hotspots based on big data technology as described in claim 2, characterized in that, The dimensionality reduction process includes: The feature matrix is ​​centered, and the covariance matrix of the centered feature matrix is ​​calculated. The covariance matrix is ​​subjected to eigenvalue decomposition to obtain eigenvalues ​​and their corresponding eigenvectors; Sort the eigenvalues ​​according to their magnitudes, and take the eigenvectors corresponding to the first n eigenvalues ​​as the basis vectors; Multiply the centered feature matrix by the basis vector to obtain the dimensionality-reduced feature matrix.

4. The method for predicting public opinion hotspots based on big data technology as described in claim 1 or 3, characterized in that, The loss function includes: ； Where L is the loss function value, and n is the number of feature matrices after dimensionality reduction. This represents the actual output corresponding to the i-th dimension-reduced feature matrix. This is the predicted output corresponding to the i-th dimension-reduced feature matrix.

5. A public opinion hotspot prediction system based on big data technology, characterized in that, include: The data processing module is configured to use web crawler technology to retrieve historical and real-time public opinion hotspot data from specified web pages, and to mark the historical public opinion hotspot data according to data attributes and store it in the corresponding database; wherein, the data attributes are three-dimensional attributes, including one-dimensional attributes, two-dimensional attributes and three-dimensional attributes, the one-dimensional attributes include text, images and videos, the two-dimensional attributes include events involving people and events involving non-people, and the three-dimensional attributes include the duration of the public opinion hotspot; The feature extraction module is configured to extract features from all data in the database using the TF-IDF and SIFT algorithms, and generate the corresponding feature matrix. The data dimensionality reduction module is configured to perform dimensionality reduction processing on the feature matrix; The network training module is configured to construct a public opinion hotspot prediction network. The feature matrix after dimensionality reduction and the corresponding hotspot weights are input into the public opinion hotspot prediction network. The mean squared error is used as the loss function to train the public opinion hotspot prediction network until the loss function value is minimized and then training stops. The public opinion hotspot prediction network includes a first sub-network, a second sub-network, a third sub-network, and a feedback network. The public opinion hotspot prediction module is configured to use the trained public opinion hotspot prediction network to predict hotspots from real-time public opinion data. The aforementioned public opinion hotspot prediction network includes: The first sub-network includes a first input gate, a first forget gate, a first output gate, and a first attention mechanism layer; The second sub-network includes a second input gate, a second forget gate, a second output gate, and a second attention mechanism layer; The third sub-network includes a third input gate, a third forget gate, a third output gate, and a third attention mechanism layer; The feedback network includes an input layer, three hidden layers, and an output layer. It is used to fuse the hotspot weights output by the first sub-network, the second sub-network, and the third sub-network, and to predict public opinion hotspots based on the fusion result. The prediction result is output using the ELU activation function. The hotspot weights include: Define the hotspot weight of the first attribute as x, the hotspot weight of the second attribute as y, and the hotspot weight of the third attribute as z; S=algx+blgy 2 +clgz 3； Where a, b, and c are weighting coefficients, and S is the fusion result.

6. The public opinion hotspot prediction system based on big data technology as described in claim 5, characterized in that, The feature extraction module is specifically configured to extract text features using the TF-IDF algorithm and extract image / video features using the SIFT algorithm.

7. The public opinion hotspot prediction system based on big data technology as described in claim 6, characterized in that, The data dimensionality reduction module is specifically configured to perform centering processing on the feature matrix and calculate the covariance matrix of the centered feature matrix; perform eigenvalue decomposition on the covariance matrix to obtain eigenvalues ​​and their corresponding eigenvectors; sort the eigenvalues ​​according to their magnitude and take the eigenvectors corresponding to the first n eigenvalues ​​as basis vectors; and multiply the centered feature matrix by the basis vectors to obtain the dimensionality-reduced feature matrix.

Images