Design method and device for public opinion monitoring and early warning platform

Abstract

The invention relates to the field of machine learning, and particularly provides a public opinion monitoring and early warning platform design method and device, and the method comprises the following steps: S1, public opinion information classification and denoising; s2, enterprise negative public opinion judgment; s3, enterprise public opinion abnormity judgment and early warning; s4, an early warning notification module; and S5, hotspot discovery and analysis. Compared with the prior art, the public opinion monitoring and early warning platform can find the attitude and opinion of the public to a certain event, object or service through public opinion monitoring and early warning, thereby helping enterprises to better understand the demands and expectations of the public and formulate more effective strategies and measures.

Description

Technical Field

[0001] The present invention relates to the field of machine learning, and specifically provides a design method and device for a public opinion monitoring and early warning platform. Background Art

[0002] Machine learning is a multi-disciplinary subject involving probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, etc. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance.

[0003] How to filter and denoise the public opinion information from news, forums, blogs, portals, forums, WeChat public accounts, Weibo, APP and other channels through the multi-channel public opinion data collected on the Internet, combine the content and characteristics of negative public opinion information dissemination from news, forums, Weibo and other channels, establish a public opinion early warning model based on multi-channel public opinion information, and discover hidden public opinion anomalies are problems that need to be urgently solved by technical personnel in this field. Summary of the invention

[0004] The present invention aims to overcome the deficiencies of the above-mentioned prior art and provides a design method for a public opinion monitoring and early warning platform with strong practicality.

[0005] A further technical task of the present invention is to provide a design device for a public opinion monitoring and early warning platform that is reasonably designed, safe and applicable.

[0006] The technical solution adopted by the present invention to solve its technical problem is:

[0007] A design method for a public opinion monitoring and early warning platform comprises the following steps:

[0008] S1, public opinion information classification and denoising;

[0009] S2. Determination of negative public opinion about enterprises;

[0010] S3. Determination and early warning of abnormal corporate public opinion;

[0011] S4, early warning notification module;

[0012] S5. Hotspot discovery and analysis.

[0013] Furthermore, in step S1, relevant public opinion data is identified from the massive public opinion data, one is an automatic classification method based on machine learning, and the other is an automatic classification method based on rules;

[0014] Automatic classification methods based on machine learning include KNN, SVM, naive Bayes and neural networks. Rule-based automatic classification is that users directly determine the classification rules for each category to form a category model. The rule classifier counts the number of rules satisfied in the test sample and the number of times the rules appear based on the category model, and uses the position information of the rules in the test text structure to measure the category to which the test sample belongs.

[0015] Furthermore, a combination of machine learning-based and rule-based automatic classification is adopted. When the training is expected to be sufficient and the overlap of category features is not large, the machine learning-based method can be used; when the training samples are small or the quality is not high, the rule-based automatic classification method can be used.

[0016] Furthermore, in step S2, the negative public opinion of the enterprise is determined by calculating and analyzing the enterprise public opinion data in real time, discovering the negative aspects related to the enterprise, and calculating the degree of negativity;

[0017] There are two commonly used methods for judging negative public opinion: one is the sentiment dictionary method, and the other is the machine learning method.

[0018] Furthermore, the method based on sentiment dictionary extracts sentiment units from the text, then scores the sentiment units in terms of polarity, and finally gives the sentiment score of the text by combining the scores of all sentiment units, which reflects the sentiment tendency and intensity of the text.

[0019] Machine learning-based methods make up for the lack of knowledge in sentiment dictionary methods. Common algorithms include support vector machines, convolutional neural networks, and deep belief networks. A negative judgment model is constructed by combining sentiment dictionary and machine learning.

[0020] Furthermore, in step S3, it further includes:

[0021] (1) Define enterprise exception types according to business needs, that is, build an exception labeling system and define exception characteristics for each exception type. In addition, classify exceptions according to their severity.

[0022] (2) After clarifying the anomaly type and corresponding features, an enterprise anomaly classification model is constructed. Usually, rule-based and machine learning-based methods are used. Through the enterprise anomaly classification model, the enterprise public opinion data is screened and classified to obtain enterprise-related anomaly graded public opinion data;

[0023] (3) Build an enterprise public opinion early warning model and early warning knowledge base based on channel type, abnormal level and abnormal number. After the enterprise abnormal public opinion data passes through the early warning model, it is determined whether to issue an early warning, and the early warning level is given according to the early warning index.

[0024] Furthermore, in step S4, a warning notification is made in the warning notification module by sending an email and a web page pop-up window.

[0025] Furthermore, in step S5, based on the real-time streaming clustering technology, relevant hot events in the public opinion information are discovered, specifically including:

[0026] It can automatically discover events or topics within a specified time, aggregate long and short texts to generate public opinion hotspots, process the latest collected public opinion data in a streaming manner and aggregate them to specific hotspots in real time, and perform multi-dimensional analysis of hotspots;

[0027] Opinion analysis is to study various typical opinions in hot events, use automated opinion analysis technology, combine domain words and attribute words in various fields to conduct positive and negative analysis on netizens' comments, analyze and summarize the main opinions of netizens, and count the distribution of netizens' positive and negative tendencies.

[0028] A design device for a public opinion monitoring and early warning platform, comprising: at least one memory and at least one processor;

[0029] The at least one memory is used to store a machine-readable program;

[0030] The at least one processor is used to call the machine-readable program to execute a design method for a public opinion monitoring and early warning platform.

[0031] Compared with the prior art, the design method and device of a public opinion monitoring and early warning platform of the present invention have the following outstanding beneficial effects:

[0032] Through the public opinion monitoring and early warning of this platform, the present invention can find out the public's attitude and views on a certain event, thing or service, thereby helping enterprises to better understand the public's needs and expectations and formulate more effective strategies and measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are 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.

[0034] Attached Figure 1 It is a flow chart of a design method of a public opinion monitoring and early warning platform. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the solution of the present invention, the present invention is further described in detail below in conjunction with specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] A best embodiment is given below:

[0037] like Figure 1 As shown, a design method of a public opinion monitoring and early warning platform in this embodiment has the following steps:

[0038] S1, public opinion information classification and denoising;

[0039] Public opinion information classification mainly realizes the identification and discovery of relevant public opinion data from massive public opinion data. There are usually two types of information classification technologies: one is the automatic classification method based on machine learning; the other is the automatic classification method based on rules.

[0040] The main automatic classification methods based on machine learning include KNN, SVM, naive Bayes, neural network, etc. The basic idea of ​​rule-based automatic classification is that the user directly determines the classification rules for each category to form a category model. The rule classifier counts the number of rules satisfied in the test sample and the number of times the rules appear based on the category model, and uses the position information of the rules in the test text structure to measure the category to which the test sample belongs.

[0041] This platform adopts a combination of machine learning and rule-based automatic classification. When the training is expected to be sufficient and the category feature overlap is not large, the machine learning method can be used; when the training samples are small or the quality is not high, the rule-based automatic classification method can be used.

[0042] Public opinion information denoising cleans up the relevant noise data identified in the public opinion classification module. On the one hand, we can optimize the feature extraction algorithm of the information classification model, such as using an intelligent word segmentation method that can eliminate ambiguity to extract company names, which can eliminate some noise. Compared with ordinary word segmentation methods, intelligent word segmentation can analyze the most reasonable word segmentation results. On the other hand, we can build an enterprise knowledge base to filter enterprise noise data in a targeted manner.

[0043] S2. Determination of negative public opinion about enterprises;

[0044] The judgment of negative public opinion about an enterprise is mainly through real-time calculation and analysis of enterprise public opinion data, discovering the negative aspects related to the enterprise and calculating its degree of negativity. There are two commonly used methods for judging negative public opinion: one is the method of sentiment dictionary; the other is the method of machine learning.

[0045] The method based on sentiment dictionary is to extract sentiment units from the text, then score the polarity of the sentiment units, and finally give the sentiment score of the text by combining the scores of all sentiment units, reflecting the sentiment tendency and intensity of the text. This method has a high accuracy rate. The reason is that the core of this method is the sentiment dictionary, and the quality and coverage of the sentiment dictionary determine the effectiveness of this method.

[0046] Methods based on machine learning can make up for the lack of knowledge of sentiment dictionary methods. Common algorithms include support vector machines, convolutional neural networks, deep belief networks, etc.

[0047] The negative judgment model is constructed by combining sentiment dictionary and machine learning. The structure diagram is as follows: Figure 1 .

[0048] S3. Determination and early warning of abnormal corporate public opinion;

[0049] The enterprise public opinion anomaly judgment and early warning module discovers abnormal enterprise data through analysis of enterprise public opinion data, and builds an early warning model to warn the enterprise.

[0050] 1) Define the enterprise abnormality types according to business needs, that is, build an abnormal label system and define abnormal characteristics for each abnormality type. In addition, according to the severity of the abnormality, the abnormality is classified into levels. For example, serious abnormalities such as running away and corporate legal persons turning themselves in can be defined as level one abnormalities, while difficulties in cash withdrawal and false propaganda can be defined as level two abnormalities.

[0051] 2) After clarifying the anomaly type and corresponding features, the enterprise anomaly classification model is constructed, usually using two methods: rule-based and machine learning. Through the enterprise anomaly classification model, the enterprise public opinion data is screened and classified to obtain the enterprise-related anomaly graded public opinion data.

[0052] 3) Build an enterprise public opinion warning model and warning knowledge base based on channel type, abnormal level, abnormal number, etc. After the enterprise abnormal public opinion data passes through the warning model, it is determined whether to issue a warning, and the warning level is given according to the warning index. In addition, this module can rank the current enterprise warning index and query historical warning data by regularly calculating and storing enterprise abnormal data.

[0053] S4, early warning notification module;

[0054] Early warning notifications are provided by sending emails, pop-up windows on web pages, etc.

[0055] S5, hotspot discovery and analysis;

[0056] Based on real-time streaming clustering technology, relevant hot events in public opinion information are discovered. Specifically, it can automatically discover events or topics within a specified time, aggregate long and short texts (long texts such as news, short texts such as Weibo, etc.) to generate public opinion hotspots, stream the latest collected public opinion data and aggregate it to specific hotspots in real time, and perform multi-dimensional analysis of hotspots, including opinion analysis, regional analysis, and scope analysis.

[0057] Opinion analysis is achieved by studying various typical opinions in hot events, including netizens' opinions and media opinions, using automated opinion analysis technology, combining domain words and attribute words in various fields to conduct positive and negative analysis on netizens' comments, analyzing and summarizing the main opinions of netizens, and counting the distribution of netizens' positive and negative tendencies, to help users understand social conditions and public opinion more accurately.

[0058] Based on the above method, a design device of a public opinion monitoring and early warning platform in this embodiment includes: at least one memory and at least one processor;

[0059] The at least one memory is used to store a machine-readable program;

[0060] The at least one processor is used to call the machine-readable program to execute a design method for a public opinion monitoring and early warning platform.

[0061] The above-mentioned specific implementations are only specific cases of the present invention. The patent protection scope of the present invention includes but is not limited to the above-mentioned specific implementations. Any technical solutions that comply with the claims of the present invention and any appropriate changes or substitutions made by ordinary technicians in the relevant technical field shall fall within the patent protection scope of the present invention.

[0062] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A design method for a public opinion monitoring and early warning platform, characterized in that: The steps are as follows: S1, public opinion information classification and denoising; S2. Determination of negative public opinion about enterprises; S3. Determination and early warning of abnormal corporate public opinion; S4, early warning notification module; S5. Hotspot discovery and analysis.

2. The design method of a public opinion monitoring and early warning platform according to claim 1 is characterized in that: In step S1, relevant public opinion data is identified from the massive public opinion data, one is an automatic classification method based on machine learning, and the other is an automatic classification method based on rules; Automatic classification methods based on machine learning include KNN, SVM, naive Bayes and neural networks. Rule-based automatic classification is that users directly determine the classification rules for each category to form a category model. The rule classifier counts the number of rules satisfied in the test sample and the number of times the rules appear based on the category model, and uses the position information of the rules in the test text structure to measure the category to which the test sample belongs.

3. The design method of a public opinion monitoring and early warning platform according to claim 2 is characterized in that: A combination of machine learning and rule-based automatic classification is adopted. When the training is expected to be sufficient and there is not much overlap in category features, the machine learning method can be used; when the training samples are small or the quality is not high, the rule-based automatic classification method can be used.

4. The design method of a public opinion monitoring and early warning platform according to claim 3 is characterized in that: In step S2, the negative public opinion of the enterprise is determined by calculating and analyzing the enterprise public opinion data in real time, discovering the negative aspects related to the enterprise and calculating the degree of negativity; There are two commonly used methods for judging negative public opinion: one is the sentiment dictionary method, and the other is the machine learning method.

5. The design method of a public opinion monitoring and early warning platform according to claim 4 is characterized in that: The method based on sentiment dictionary is to extract sentiment units from the text, then score the sentiment units in terms of polarity, and finally give the sentiment score of the text by combining the scores of all sentiment units, which reflects the sentiment tendency and intensity of the text. Machine learning-based methods make up for the lack of knowledge in sentiment dictionary methods. Common algorithms include support vector machines, convolutional neural networks, and deep belief networks. A negative judgment model is constructed by combining sentiment dictionary and machine learning.

6. The design method of a public opinion monitoring and early warning platform according to claim 5 is characterized in that: In step S3, further comprising: (1) Define enterprise exception types according to business needs, that is, build an exception labeling system and define exception characteristics for each exception type. In addition, classify exceptions according to their severity. (2) After clarifying the anomaly type and corresponding features, an enterprise anomaly classification model is constructed. Usually, rule-based and machine learning-based methods are used. Through the enterprise anomaly classification model, the enterprise public opinion data is screened and classified to obtain enterprise-related anomaly graded public opinion data; (3) Build an enterprise public opinion early warning model and early warning knowledge base based on channel type, abnormal level and abnormal number. After the enterprise abnormal public opinion data passes through the early warning model, it is determined whether to issue an early warning, and the early warning level is given according to the early warning index.

7. The design method of a public opinion monitoring and early warning platform according to claim 6 is characterized in that: In step S4, a warning notification is made in the warning notification module by sending an email and a pop-up window on a web page.

8. The design method of a public opinion monitoring and early warning platform according to claim 7 is characterized in that: In step S5, based on the real-time streaming clustering technology, relevant hot events in the public opinion information are discovered, specifically including: It can automatically discover events or topics within a specified time, aggregate long and short texts to generate public opinion hotspots, process the latest collected public opinion data in a streaming manner and aggregate them to specific hotspots in real time, and perform multi-dimensional analysis of hotspots; Opinion analysis is to study various typical opinions in hot events, use automated opinion analysis technology, combine domain words and attribute words in various fields to conduct positive and negative analysis on netizens' comments, analyze and summarize the main opinions of netizens, and count the distribution of netizens' positive and negative tendencies.

9. A design device for a public opinion monitoring and early warning platform, characterized in that: include: at least one memory and at least one processor; The at least one memory is used to store a machine-readable program; The at least one processor is configured to call the machine-readable program to execute the method according to any one of claims 1 to 8.

Images