CN109376237B - Client stability prediction method, device, computer equipment and storage medium - Google Patents
Client stability prediction method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN109376237B CN109376237B CN201811025974.XA CN201811025974A CN109376237B CN 109376237 B CN109376237 B CN 109376237B CN 201811025974 A CN201811025974 A CN 201811025974A CN 109376237 B CN109376237 B CN 109376237B
- Authority
- CN
- China
- Prior art keywords
- information
- client
- target
- hidden layer
- emotion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000008451 emotion Effects 0.000 claims abstract description 91
- 238000012544 monitoring process Methods 0.000 claims abstract description 49
- 238000012545 processing Methods 0.000 claims description 23
- 238000004590 computer program Methods 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 12
- 239000011159 matrix material Substances 0.000 claims description 10
- 230000011218 segmentation Effects 0.000 claims description 9
- 230000002996 emotional effect Effects 0.000 claims description 5
- 238000013507 mapping Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000013473 artificial intelligence Methods 0.000 abstract 1
- 238000012549 training Methods 0.000 description 11
- 230000006399 behavior Effects 0.000 description 8
- 238000013145 classification model Methods 0.000 description 7
- 239000000284 extract Substances 0.000 description 7
- 230000004927 fusion Effects 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000007477 logistic regression Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000010801 machine learning Methods 0.000 description 5
- 238000003062 neural network model Methods 0.000 description 5
- 239000013598 vector Substances 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000007637 random forest analysis Methods 0.000 description 3
- 230000002829 reductive effect Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000007635 classification algorithm Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000013075 data extraction Methods 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 230000000670 limiting effect Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 230000009193 crawling Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003862 health status Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 230000003211 malignant effect Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to a client stability prediction method and device based on artificial intelligence, computer equipment and storage medium. The method comprises the following steps: monitoring a plurality of target information issued by a network platform; acquiring client data corresponding to a target client; the customer data includes a product identifier; calculating a public opinion index corresponding to the product identifier based on the target information; counting access data of a target client to target information in the monitoring period; determining emotion characteristics of the target client according to the client data; and inputting the public opinion index, the access data and the emotion characteristics into a preset information influence prediction model, and outputting the stability parameters of the target client. By adopting the method, the stability of the client can be predicted in time, and the prediction accuracy is improved.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to a method and apparatus for predicting customer stability, a computer device, and a storage medium.
Background
As the impact of internet media becomes greater, more and more network platforms exist that distribute information to users via the internet. Although the information issued by the network platform may not contain any emotion words, accessing the information can lead people to generate certain emotion tendencies, and the stability of enterprise clients is directly affected by some emotion tendencies. Although most enterprises respond to the information issued by the network platform, for example, refute a rumour information is issued, when responding to the related information, the adverse phenomena such as customer loss occur to the corresponding enterprises at the moment after the information event is issued and fermented by public opinion. As can be seen, the prior art lacks an early warning solution to provide an enterprise with respect to customer stability.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a client stability prediction method, apparatus, computer device, and storage medium that can predict client stability in time and improve prediction accuracy.
A method of predicting customer stability, the method comprising: monitoring a plurality of target information issued by a network platform; acquiring client data corresponding to a target client; the customer data includes a product identifier; calculating a public opinion index corresponding to the product identifier based on the target information; counting access data of a target client to target information in the monitoring period; determining emotion characteristics of the target client according to the client data; and inputting the public opinion index, the access data and the emotion characteristics into a preset information influence prediction model, and outputting the stability parameters of the target client.
In one embodiment, the monitoring the plurality of target information issued by the network platform includes: monitoring original information issued by a network platform; word segmentation processing is carried out on the original information to obtain an information tag corresponding to each piece of original information; acquiring a plurality of target keywords, and identifying whether the information tag contains the target keywords; if yes, marking the corresponding original information as target information.
In one embodiment, the calculating, based on the target information, a public opinion index corresponding to the product identifier includes: splitting the target information to obtain a plurality of short texts; extracting a product identifier from the short text, and associating the product identifier with the corresponding short text; calculating emotion indexes corresponding to each short text by using a preset public opinion analysis model; determining influence weights corresponding to the short texts respectively; and calculating the public opinion index corresponding to the corresponding product identifier according to the emotion index and the influence weight of the associated short text.
In one embodiment, the statistics of the access data of the target client to the target information in the monitoring period includes: acquiring an identification field of a target client, and sending the identification field to the network platform; receiving associated access data returned by the network platform according to the identification field; extracting an information access field, an information collection field and an information forwarding field from the associated access data; and counting the information access amount, the information collection amount and the information forwarding amount of the target client in the monitoring period based on the information access field, the information collection field and the information forwarding field.
In one embodiment, the determining the emotion feature of the target client according to the client data includes: acquiring a plurality of sub-models, and determining weight factors corresponding to the sub-models respectively; generating a first model according to the plurality of sub-models and the weight factors corresponding to the sub-models respectively; acquiring client sample data and corresponding emotion labels; inputting the client sample data into the first model to obtain an intermediate emotion analysis result; calculating the difference between the intermediate emotion analysis result and the emotion label, and adjusting the first model according to the difference to obtain a second model; and inputting the client data into the second model, and outputting the emotion characteristics of the target client.
In one embodiment, the information influence prediction model is obtained by training a deep neural network model; inputting the public opinion index, the access data and the emotion characteristics into a preset information influence prediction model, and outputting the stability parameters of the target client, wherein the method comprises the following steps: preprocessing the public opinion index, the access data and the emotion characteristics to obtain a client characteristic matrix; obtaining an input layer node sequence according to the client feature matrix; projecting the input layer node sequence to obtain a hidden layer node sequence corresponding to a first hidden layer, and taking the first hidden layer as a current processing hidden layer; acquiring the weight and deviation of each neuron node corresponding to the current processing hidden layer; according to the hidden layer node sequence corresponding to the current processing hidden layer and the weight and deviation of each neuron node, nonlinear mapping is adopted to obtain the hidden layer node sequence of the next hidden layer; iterating the next hidden layer serving as the current processing hidden layer until the next hidden layer is an output layer; and acquiring the stability parameters corresponding to the target clients output by the output layer.
A customer stability prediction apparatus, the apparatus comprising: the information analysis module is used for monitoring a plurality of target information issued by the network platform; acquiring client data corresponding to a target client; the customer data includes a product identifier; calculating a public opinion index corresponding to the product identifier based on the target information; the client analysis module is used for counting the access data of the target client to the target information in the monitoring period; determining emotion characteristics of the target client according to the client data; and the influence prediction module is used for inputting the public opinion index, the access data and the emotion characteristics into a preset information influence prediction model and outputting the stability parameters of the target client.
A computer device comprising a memory storing a computer program and a processor implementing the steps of the method of predicting customer stability provided in any one of the embodiments of the present application when the computer program is executed.
A computer readable storage medium having stored thereon a computer program which when executed by a processor implements the steps of the client stability prediction method provided in any one of the embodiments of the present application.
The method, the device, the computer equipment and the storage medium for predicting the stability of the client monitor a plurality of target information issued by a network platform in real time; one or more product identifiers can be obtained according to the client data corresponding to the target client to be analyzed; according to the monitored target information, public opinion indexes corresponding to different product identifiers can be calculated; the stability parameters of the target client can be obtained based on the information influence prediction model by counting the access data of the target client to the target information in the monitoring period and determining the emotion characteristics of the target client according to the client data. The influence of the target information on the product is predicted, so that the public opinion index corresponding to the product identification is obtained, the understanding degree of the target client on the target information is adopted, the emotion characteristics of whether the client is easily influenced by the target information are further considered, and the prediction accuracy of the information influence can be improved by comprehensively considering a plurality of factors; the prediction result can be obtained by directly inputting the calculated multiple factor data into a preset information influence prediction model, so that the stability of the client can be predicted in time and the prediction accuracy can be improved.
Drawings
FIG. 1 is an application scenario diagram of a method of predicting client stability in one embodiment;
FIG. 2 is a flow diagram of a method for predicting client stability in one embodiment;
FIG. 3 is a flowchart illustrating a step of calculating a product public opinion index according to one embodiment;
FIG. 4 is a flowchart illustrating steps for determining a client emotion feature in one embodiment;
FIG. 5 is a block diagram of a client stability prediction apparatus in one embodiment;
Fig. 6 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
The method for predicting the stability of the client provided by the application can be applied to an application environment shown in figure 1. Wherein the terminal 102 communicates with the server 104 via a network. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smartphones, tablet computers, and portable wearable devices, and the server 104 may be implemented by a stand-alone server or a server cluster composed of a plurality of servers. The server 104 is monitored for original information released by the network platform in a monitoring period, and target information related to enterprise business is screened from the original information. The server 104 predicts whether the target information will affect the stability of the target client according to the information influence prediction request sent by the terminal 102 or according to a preset time frequency. Specifically, the server 104 obtains client data corresponding to the target client. The customer data includes a product identification. The server 104 calculates a public opinion index corresponding to the product identification based on the target information, i.e., predicts the influence of the target information on the price, etc. of the product corresponding to the product identification. The server 104 counts access data of the target client to the target information, such as behavior data of browsing, forwarding, collecting, etc. of the target information in the monitoring period. Server 104 determines the emotional characteristics of the target client, i.e., the extent to which it is susceptible to changes due to external influences, based on the client data. The server 104 pre-stores the information impact prediction model. The information influence prediction model may be trained using a deep neural network model. The server 104 inputs the public opinion index, the access data and the emotion characteristics into the information influence prediction model, and outputs the stability parameters of the target client. According to the information influence prediction process, the influence of the target information on the product is predicted, the public opinion index corresponding to the product identification is obtained, the knowledge degree of the target client on the target information is adopted, the emotion characteristics of whether the client is easily influenced by the target information are further considered, and the information influence prediction accuracy can be improved by comprehensively considering a plurality of factors.
In one embodiment, as shown in fig. 2, a method for predicting the stability of a client is provided, and the method is applied to the server in fig. 1 for illustration, and includes the following steps:
step 202, monitoring a plurality of target information issued by a network platform.
The network platform can be a pre-designated variety of information readers, such as UC browser, QQ browser, today's headings, daily flash newspaper, etc. The network platform can also be a pre-specified social platform such as WeChat, microblog, QQ, bar, forum or knowledge. The target information has a corresponding influencing object. The target information can influence the emotion attitude of people, so that benign or malignant influence, such as passenger source loss, resource devaluation and the like, is caused on the influence object. The influencing object type may be a communication device, a property building, a virtual resource, other products, etc. Wherein the virtual resource may be a product or the like. The monitoring period refers to the time frequency of the server for collecting the target information, and can be freely set according to the requirement, such as 1 week, etc., which is not limited. The time frequency with which the server crawls the target information from different network platforms may be different.
In one embodiment, monitoring a plurality of target information published by a network platform includes: monitoring original information issued by a network platform; word segmentation is carried out on the original information, and an information label corresponding to each original information is obtained; acquiring a plurality of target keywords, and identifying whether the information tag contains the target keywords; if yes, marking the corresponding original information as target information.
The network platform may send out multiple pieces of original information at intervals during the monitoring period. But not all of the original information affects the stability of the client and the server screens the original information related to the business of the enterprise. Specifically, the server performs word segmentation, stop word removal, named entity replacement and other processing on each piece of acquired original information to obtain a plurality of information keywords. The server calculates TF-IDF values (termfrequency-inverse document frequency, word frequency-reverse file frequency) for each information keyword. Specifically, the server calculates TF values (term frequency) of the corresponding information keywords by counting the number of times each information keyword appears in all the information keywords; the IDF value (inverse document frequency, reverse file frequency) of the corresponding information keyword is calculated by counting the ratio of the total number of information sentences in the original information to the number of information sentences containing a certain information keyword. And the server calculates the TF-IDF value of the corresponding information keyword according to the TF value and the IDF value of each information keyword. The TF-IDF value may reflect traffic class discrimination capability of the information keyword. The server screens the information keywords according to the TF-IDF values to obtain the information keywords with high preset quantity of TF-IDF values.
And the server generates an information label corresponding to the original information by using the information keywords obtained through screening. The server pre-stores a plurality of target keywords related to the business. And the server matches the information tag with the target keyword and identifies whether the information tag contains the target keyword. If the information tag contains a target keyword, the original information is related to the enterprise business, and the server marks the original information as target information. According to the method and the device for the public opinion analysis, the collected information is screened, so that the server can only conduct public opinion analysis on target information related to enterprise business, the data size required to be processed by the server can be reduced, further occupation of server resources is reduced, and prediction efficiency of information influence is improved.
Step 204, obtaining client data corresponding to a target client; the customer data includes a product identification.
The target client may be an existing client or a potential client. For potential clients, the server may obtain corresponding client data by way of a web crawler. The client data comprises inherent attributes such as gender, age, occupation and the like, and also comprises dynamic behavior data such as web browsing, information forwarding, dynamic release and the like, and can be text, voice, video or pictures and the like. For existing clients, the server may directly obtain corresponding client data at the business system. The customer data of the existing customer includes an identification of the product that has been purchased or is ready to be purchased (hereinafter referred to as "product identification"). For example, the customer has purchased and is currently active, purchased but has failed, and the product identification of the insurance product is ready for purchase.
And 206, calculating the public opinion index corresponding to the product identifier based on the target information.
And splitting the target information by the server to obtain a plurality of short texts. The server extracts keywords of an object (hereinafter referred to as an 'influence object') which can represent the possible influence of the object from the target information, and determines the type of the influence object corresponding to the target object according to the extracted keywords. The server stores a plurality of influence object types in advance, and public opinion factors and public opinion analysis models corresponding to each influence object type. According to the target information, the server acquires public opinion factors corresponding to the corresponding influence object types, extracts target keywords from a plurality of short texts according to the public opinion factors, inputs the extracted target keywords into a public opinion analysis model corresponding to the influence object types, and calculates an emotion index corresponding to the target information. The server may analyze target information of various influencing object types, such as virtual resources, communication devices, etc. When the type of the influence object is a virtual resource, the server is also used for measuring and calculating the public opinion index corresponding to the virtual resource according to the emotion index. The public opinion index can characterize the degree of influence of target information on the price, audience group, popularity, etc. of different products.
And step 208, counting access data of the target clients to the target information in the monitoring period.
The server collects access data of the target client to the target information on the network platform based on the client identification. The access data includes a browse record for one or more pieces of logo information, such as browse time, forward, collect, browse duration, comment information, etc. And the server counts the information access quantity, the information collection quantity and the information forwarding quantity of the target client in the monitoring period according to the access data.
In one embodiment, counting access data of a target client to target information during a monitoring period includes: acquiring an identification field of a target client, and sending the identification field to a network platform; receiving associated access data returned by the network platform according to the identification field; extracting an information access field, an information collection field and an information forwarding field from the associated access data; and counting the information access amount, the information collection amount and the information forwarding amount of the target client in the monitoring period based on the information access field, the information collection field and the information forwarding field.
The target client has a corresponding identification field. The server extracts the base identification field from the identity information maintained by the target client at the enterprise. The identification field may be an identification field of the target client, as well as the relatives or friends of the target client (hereinafter "associated object"). The identification field comprises a name, an identification card number, a mobile phone number, a mailbox account number, a network account number, common equipment information and the like. The common device information may be IMEI (International Mobile Equipment Identity ), IP address, device fingerprint, operating system version number, serial number, etc.
Different network platforms are running on different internet servers. The target client will leave access records in the corresponding internet server when using the network platform inside and outside the various device access mechanisms. The access record may be stored in the form of a log or file or the like. And the server generates a data extraction request according to the basic identification field of the target client and sends the data extraction request to the Internet server. The internet server searches the access record containing the identification field, and returns the searched access record to the server. The access record refers to behavior data (hereinafter referred to as "associated access data") of a target client based on information access behaviors (such as browsing behaviors, comment behaviors, forwarding behaviors, collection behaviors, and the like) occurring in a mobile terminal, an automobile, an intelligent robot, an intelligent wearable device, and the like. The server extracts information access fields, information collection fields, information forwarding fields and the like from the associated access data, and performs statistical analysis on the extracted field information to obtain information access quantity, information collection quantity, information forwarding quantity and the like of target information by a target client in a monitoring period.
Step 210, determining emotion characteristics of the target client according to the client data.
The server trains and obtains analysis models of different monitoring periods by using client data of the historical monitoring periods. Each monitoring cycle produces an analytical model, each classification model having a corresponding model identification (which may be a date, such as 201708, etc.) and a weighting factor W. The server selects a preset number of analysis models based on a preset sliding window function, and constructs an initial machine learning model according to the analysis models obtained by screening and the corresponding weight factors. And training a preset machine learning model based on the client data of the current monitoring period by the server to obtain an emotion characteristic model. And when the monitoring period is up, the server inputs the client data of multiple dimensions of the target client into the emotion feature model to obtain emotion features of the target client. The emotional characteristic may be a quantitative or qualitative parameter value that characterizes the extent to which a customer is susceptible to social public opinion, i.e., the degree of influence of social public opinion on the customer, to produce an emotional fluctuation. For example, -100 can be used to characterize the emotional characteristics of the client, the larger the absolute value of the parameter is, the larger the influence of the social public opinion on the client is, the larger the parameter value is, the larger the positive influence of the social public opinion on the client is, and the smaller the parameter value is, the larger the negative influence of the social public opinion on the client is.
And 212, inputting the public opinion index, the access data and the emotion characteristics into a preset information influence prediction model, and outputting the stability parameters of the target client.
The server pre-stores the information impact prediction model. The information influence prediction model may be trained using a deep neural network model. Specifically, the information impact prediction model includes an input layer and an output layer. Also included between the input layer and the output layer are a plurality of hidden layers. The layers are fully connected. Each layer comprises a plurality of neurons, and the input parameters of the neurons of the same layer are the same.
In one embodiment, the information influence prediction model is obtained by training a deep neural network model; inputting public opinion indexes, access data and emotion characteristics into a preset information influence prediction model, and outputting stability parameters of a target client, wherein the method comprises the following steps: preprocessing public opinion indexes, access data and emotion characteristics to obtain a client characteristic matrix; obtaining an input layer node sequence according to the client feature matrix; projecting the input layer node sequence to obtain a hidden layer node sequence corresponding to the first hidden layer, and taking the first hidden layer as the current processing hidden layer; acquiring weight and deviation of each neuron node corresponding to the hidden layer in current processing; according to the hidden layer node sequence corresponding to the hidden layer which is currently processed and the weight and deviation of each neuron node, nonlinear mapping is adopted to obtain the hidden layer node sequence of the next hidden layer; iterating the next hidden layer serving as the current processing hidden layer until the next hidden layer is an output layer; and acquiring the stability parameters corresponding to the target clients output by the output layer.
In the embodiment, a plurality of target information issued by a network platform in a monitoring period is monitored in real time; one or more product identifiers can be obtained according to the client data corresponding to the target client to be analyzed; according to the monitored target information, public opinion indexes corresponding to different product identifiers can be calculated; the stability parameters of the target clients can be obtained based on the information influence prediction model by counting the access data of the target clients to the target information in the monitoring period and determining the emotion characteristics of the target clients according to the client data. The influence of the target information on the product is predicted, so that the public opinion index corresponding to the product identification is obtained, the understanding degree of the target client on the target information is adopted, the emotion characteristics of whether the client is easily influenced by the target information are further considered, and the prediction accuracy of the information influence can be improved by comprehensively considering a plurality of factors; the prediction result can be obtained by directly inputting the calculated multiple factor data into a preset information influence prediction model, so that the stability of the client can be predicted in time and the prediction accuracy can be improved.
In one embodiment, as shown in fig. 3, the step of calculating the public opinion index corresponding to the product identifier based on the target information, that is, calculating the product public opinion index, includes:
And step 302, splitting the target information to obtain a plurality of short texts.
And step 304, extracting the product identification from the short text, and associating the product identification with the corresponding short text.
And 306, calculating the emotion index corresponding to each short text by using a preset public opinion analysis model.
In step 308, the impact weights corresponding to the short texts are determined.
Step 310, calculating the public opinion index corresponding to the corresponding product identifier according to the emotion index and influence weight of the associated short text.
The target information may be text, voice, video, or pictures, etc. If the target information is voice, video or picture, the target information is firstly converted into text. The converted target information is a long text including a plurality of split identifiers. The server determines the position of each splitting identifier as a splitting position, and splits at each splitting position of the long text to obtain a plurality of short texts. The split identifier may be a sentence ending symbol such as a period, an exclamation mark, or the like.
The server extracts the intermediate keywords at each short text. Specifically, the server performs word segmentation on the short text, and if the word segmentation includes stop words or punctuation marks, filters a plurality of word segments, and deletes the stop words and the punctuation marks so as to save the storage space of the server. And the server performs synonym replacement and named entity replacement on the filtered multiple participles. The server pre-stores the synonym table and named entities. The synonym replacement can unify multiple expression modes of the same concept, so that key concepts of the short text are not highlighted, the difficulty of public opinion analysis according to intermediate keywords by a server is reduced, and the public opinion analysis efficiency and accuracy can be improved. The substitution of named entities can reduce the granularity of public opinion analysis and can further improve the efficiency of public opinion analysis. And according to the public opinion factors respectively corresponding to the prestored multiple types of influence objects, the server determines one or more replaced segmentation words as intermediate keywords. The public opinion factor refers to a factor which possibly affects the emotion attitude of the user in the target information.
The server stores a public opinion analysis model in advance. The public opinion analysis model can be obtained by training a machine learning classification model. Specifically, the server converts the plurality of intermediate keywords into corresponding word vectors based on a word2vec model, and adds corresponding classification labels to each word vector. The word vectors and the corresponding classification labels form a training set, and the machine learning classification model is trained based on the training set to obtain the public opinion analysis model. The machine learning classification model may be GBDT model or XGBOOST model, or the like. And the server inputs the extracted intermediate keywords into a public opinion analysis model corresponding to the corresponding influence object types, and calculates and obtains emotion indexes corresponding to the target information.
Each target information has corresponding profile information such as distribution time, distribution media, distribution author, etc. The server calculates an influence weight of each target information based on profile information of the target information. For example, the influence weight may be an accumulated sum of a time weight, a media weight, an author weight, and the like. It is easy to understand that influence weights corresponding to a plurality of short texts obtained by splitting the same target information are the same.
The emotion indexes of the target information calculated by the server comprise emotion indexes corresponding to a plurality of short texts respectively. The server extracts the product identification in the short text by a trie algorithm. The product identification may be a product name or product number, etc. In other words, the intermediate keywords extracted by the server in some short text include product identification. The server may extract the same or different product identifications in different short texts. The server associates the product identification with the corresponding short text. It is readily understood that the same product identification may be associated with multiple short texts from multiple target information. And the server calculates the target public opinion index corresponding to the corresponding product according to the emotion index of the short text corresponding to the product identifier and the corresponding influence weight. For example, the public opinion index corresponding to each product identifier may be a weighted sum of emotion indexes of all short texts associated with the product identifier, such as product a public opinion index = short text 11 x influence weight 11+short text 12 x influence weight 12+ + short text 21 x influence weight 21.
In this embodiment, the influence of different target information on different products, that is, the public opinion index, is calculated by combining the influence weights of the target information, so that the accuracy of public opinion analysis can be improved.
In one embodiment, as shown in fig. 4, the step of determining the emotion characteristics of the target client, i.e., determining the emotion characteristics of the client, according to the client data includes:
Step 402, obtaining a plurality of sub-models, and determining weight factors corresponding to the plurality of sub-models respectively.
Step 404, generating a first model according to the plurality of sub-models and the weight factors respectively corresponding to the sub-models.
Step 406, obtaining client sample data and corresponding emotion tags.
Step 408, inputting the customer sample data into the first model to obtain an intermediate emotion analysis result.
Step 410, calculating the difference between the intermediate emotion analysis result and the emotion label, and adjusting the first model according to the difference to obtain a second model.
Step 412, inputting the client data into the second model, and outputting the emotion characteristics of the target client.
And the server constructs an emotion characteristic analysis model of the corresponding monitoring period every other monitoring period. The time length of the monitoring period can be freely set according to the requirement, such as 1 year. The emotion characteristic analysis model corresponding to the current monitoring period can be constructed by using emotion characteristic analysis models of a plurality of historical monitoring periods. For convenience of description, the emotion feature analysis model of the history monitoring period is referred to as a "submodel". The initial sub-model may be a model that the server trains to the initial model using a large amount of customer sample data.
The server obtains client sample data for a plurality of historical periods. The history period corresponds to the history monitoring period described above. The server adds a corresponding quality tag to the customer sample data for each customer. In order to reduce the complexity of manual labeling, a server establishes a customer portrait according to customer sample data, and automatically generates quality labels of corresponding customers based on the customer portrait. Specifically, the server performs processing such as cleaning on the client sample data to obtain a plurality of attribute labels corresponding to the client, such as age, gender, occupation, marital status, cultural degree, occupation, property guarantee, health status and the like of the user. The server composes the acquired plurality of attribute tags into a text vector, and takes the composed text vector as a customer portrait of the customer. Customer portraits, which are virtual representations of real customers, are often built from products and markets, reflecting the characteristics and needs of real customers. The server prestores a plurality of attribute label combinations and quality labels corresponding to each combination respectively. And the server converts the customer portrait based on the corresponding relation between the prestored attribute label combination and the quality label to obtain the quality label corresponding to the corresponding customer identifier. The quality label may be a quantitative index such as a score, or a qualitative index such as a quality score, or a quality score.
The server trains the initial model based on a large number of client sample data and corresponding quality labels to obtain corresponding sub-models. The initial model can be obtained by fitting a feature classification model and a feature fusion model. The initial model includes a plurality of customer metrics, each customer metric having a corresponding plurality of customer attributes, e.g., the customer attribute corresponding to the customer metric "gender" may be "male" or "female". The server calculates the entropy gain corresponding to each client index. The formula for calculating the entropy gain may be:
Wherein GA represents the entropy gain of the calculated client index A; m represents the probability that the client emotion characteristic index reaches a threshold value; ai represents the ratio of the number of the client attributes i corresponding to the client index a to the total number of the client attributes in the client sample data, ai represents the probability that the client quality emotion feature index of the client attribute i based on the number of the client index a reaches a threshold value, and n represents the number of the client attributes corresponding to the client index a. And the server performs weighted summation on the entropy gain of the plurality of client indexes to obtain the entropy gain corresponding to the corresponding client index combination. And training the first preset model through a feature classification algorithm according to the entropy gain and the quality label respectively corresponding to each client index combination by the server to obtain a feature classification model. The feature classification algorithm may be a combination of GBDT (Gradient Boost Decision Tree, gradient-lifted tree algorithm) and (Logistic Regression, logistic regression algorithm).
The server trains and obtains a characteristic fusion model based on the client sample data. Specifically, if the client data is obtained by crawling through the network platform, naming modes of the same client index by different network platforms may be different, and in order to reduce influence of naming differences on model training, the server performs synonymous expansion processing on each client index to obtain expansion index combinations respectively corresponding to each client index combination. And the server respectively acquires synonyms corresponding to the individual word segments in the client indexes, and forms an expansion word set by the word segments and the corresponding synonyms. Each word segment has a corresponding set of expansion words, if the client index combination a is { a, b, c }, each client index in the client index combination has a corresponding set of expansion words, if the set of expansion words corresponding to the client index a is { a, a1, a2}. The server randomly selects one word from the expansion word sets corresponding to the client indexes according to the appearance sequence of the client indexes in the client index combination, and sequentially forms an expansion index set. When different words are selected from the expansion word sets, different expansion index sets are formed, and the expansion index sets form expansion index combinations. And training the second preset model by the server through a feature fusion algorithm according to each expansion index combination and the emotion analysis results corresponding to each expansion index combination to obtain a feature fusion model. The feature fusion algorithm may be a random forest algorithm or the like.
The method comprises the steps of firstly forming an expansion word set corresponding to each client index, then forming expansion index combinations corresponding to each client index combination through the expansion word set, greatly improving the expansion degree of the client indexes, expressing the meaning identical or similar to that of the original client index by each client index after expansion, and improving the effective coverage range of the client index, so that after a trained feature fusion model is input subsequently, the emotion feature analysis accuracy can be improved.
And fitting the feature classification model and the feature fusion model by the server to obtain a corresponding sub-model. In a specific embodiment, the server performs a linear fit to the logistic regression model, GBDT (Gradient Boost Decision Tree, nonlinear model), (Logistic Regression, logistic regression model), random forest model to obtain the sub-model. For example, submodel = logistic regression model w1+ GBDT w2+lr w3+random forest model w4. Wherein Wi is a weight factor. The ROC (receiver operating characteristic curve, subject working characteristics) variability exists among different types of models, and the accuracy of the emotion characteristic analysis of the client can be improved by fitting the different types of models.
Each sub-model has a corresponding time stamp. The time stamp may be generated based on a build period of the submodel, such as 2017,20170317, etc. The server decays the function according to timeAnd determining the contribution rate of each sub-model, namely determining the weight factors corresponding to the plurality of sub-models respectively. Wherein Δt is the time difference between the time tag and the current time; t is the optimal length of time. For example, the time difference Δt=1 between the year of the time tag "2017" and the current time "2018"; t may be the time span corresponding to the sliding window function, i.e. the number of sub-models obtained by screening. It is readily understood that the sub-model further from the current time period is trained using earlier historical customer data, the less the reference meaning (i.e., contribution rate) for analyzing the customer affective characteristics of the current time period. In other words, the larger the time difference, the smaller the contribution rate of the corresponding sub-model to the quality analysis, so that the weighting factors of the plurality of sub-models can be determined based on the time decay function.
The server performs linear regression operation based on the plurality of sub-models and the weight factors corresponding to the sub-models respectively to obtain a first model. In order to improve the accuracy of the first model, the server performs training reinforcement on the first model. Specifically, the server obtains client sample data of a plurality of clients in a current monitoring period. The customer sample data has a corresponding class label. The customer sample data includes information for multiple dimensions of the customer, such as age, occupation, family members, and the like. And the server inputs the client sample data of the current monitoring period into the first model to obtain an intermediate classification result. The server calculates the difference between the intermediate classification result and the classification label, and adjusts the first model according to the difference to obtain a second model.
In the embodiment, since the emotion feature analysis model is pre-built by using the client sample data, the emotion features of the corresponding clients can be obtained quickly by taking the client data as the input parameters based on the analysis model, and the information of a plurality of dimensions of the clients can be comprehensively considered, so that the emotion feature analysis efficiency of the clients is improved, and the emotion feature analysis accuracy of the clients is also improved.
It should be understood that, although the steps in the flowcharts of fig. 2 to 4 are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps of fig. 2-4 may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the sub-steps or stages are performed necessarily occur in sequence, but may be performed alternately or alternately with at least a portion of other steps or sub-steps or stages of other steps.
In one embodiment, as shown in fig. 5, there is provided a client stability prediction apparatus, including: an information analysis module 502, a customer analysis module 504, and an impact prediction module 506, wherein:
The information analysis module 502 is configured to monitor a plurality of target information published by the network platform; acquiring client data corresponding to a target client; the customer data includes a product identifier; and calculating the public opinion index corresponding to the product identifier based on the target information.
A client analysis module 504, configured to count access data of the target client to the target information in the monitoring period; and determining the emotion characteristics of the target client according to the client data.
The influence prediction module 506 is configured to input the public opinion index, the access data, and the emotion feature into a preset information influence prediction model, and output stability parameters of the target client.
In one embodiment, the information analysis module 502 is further configured to monitor the original information published by the network platform; word segmentation is carried out on the original information, and an information label corresponding to each original information is obtained; acquiring a plurality of target keywords, and identifying whether the information tag contains the target keywords; if yes, marking the corresponding original information as target information.
In one embodiment, the information analysis module 502 is further configured to split the target information to obtain a plurality of short texts; extracting a product identifier from the short text, and associating the product identifier with the corresponding short text; calculating emotion indexes corresponding to each short text by using a preset public opinion analysis model; determining influence weights corresponding to the short texts respectively; and calculating the public opinion index corresponding to the corresponding product identifier according to the emotion index and the influence weight of the associated short text.
In one embodiment, the client analysis module 504 is further configured to obtain an identification field of the target client, and send the identification field to the network platform; receiving associated access data returned by the network platform according to the identification field; extracting an information access field, an information collection field and an information forwarding field from the associated access data; and counting the information access amount, the information collection amount and the information forwarding amount of the target client in the monitoring period based on the information access field, the information collection field and the information forwarding field.
In one embodiment, the client analysis module 504 is further configured to obtain a plurality of sub-models, and determine weight factors corresponding to the plurality of sub-models respectively; generating a first model according to the plurality of sub-models and the weight factors corresponding to the sub-models respectively; acquiring client sample data and corresponding emotion labels; inputting the customer sample data into a first model to obtain an intermediate emotion analysis result; calculating the difference between the intermediate emotion analysis result and the emotion label, and adjusting the first model according to the difference to obtain a second model; and inputting the client data into the second model, and outputting the emotion characteristics of the target client.
In one embodiment, the information influence prediction model is obtained by training a deep neural network model; the influence prediction module 506 is further configured to preprocess the public opinion index, the access data and the emotion feature to obtain a client feature matrix; obtaining an input layer node sequence according to the client feature matrix; projecting the input layer node sequence to obtain a hidden layer node sequence corresponding to the first hidden layer, and taking the first hidden layer as the current processing hidden layer; acquiring weight and deviation of each neuron node corresponding to the hidden layer in current processing; according to the hidden layer node sequence corresponding to the hidden layer which is currently processed and the weight and deviation of each neuron node, nonlinear mapping is adopted to obtain the hidden layer node sequence of the next hidden layer; iterating the next hidden layer serving as the current processing hidden layer until the next hidden layer is an output layer; and acquiring the stability parameters corresponding to the target clients output by the output layer.
For specific limitations on the means for predicting the stability of a customer, reference may be made to the above limitations on the method for predicting the stability of a customer, and will not be described in detail herein. The respective modules in the above-described client stability prediction apparatus may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 6. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store customer data. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program, when executed by a processor, implements a method of predicting customer stability.
It will be appreciated by those skilled in the art that the structure shown in FIG. 6 is merely a block diagram of some of the structures associated with the present inventive arrangements and is not limiting of the computer device to which the present inventive arrangements may be applied, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.
A computer readable storage medium having stored thereon a computer program which when executed by a processor implements the steps of the client stability prediction method provided in any one of the embodiments of the present application.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (SYNCHLINK) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the application, which are described in detail and are not to be construed as limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of protection of the present application is to be determined by the appended claims.
Claims (10)
1. A method of predicting customer stability, the method comprising:
Monitoring a plurality of target information issued by a network platform;
Acquiring client data corresponding to a target client; the customer data includes a product identifier;
calculating a public opinion index corresponding to the product identifier based on the target information;
Counting access data of a target client to target information in a monitoring period;
determining emotion characteristics of the target client according to the client data, wherein the emotion characteristics are used for representing quantitative or qualitative parameter values of the degree of emotion fluctuation caused by the influence of social public opinion;
preprocessing the public opinion index, the access data and the emotion characteristics to obtain a client characteristic matrix;
obtaining an input layer node sequence according to the client feature matrix;
projecting the input layer node sequence to obtain a hidden layer node sequence corresponding to a first hidden layer, and taking the first hidden layer as a current processing hidden layer;
Acquiring the weight and deviation of each neuron node corresponding to the current processing hidden layer; according to the hidden layer node sequence corresponding to the current processing hidden layer and the weight and deviation of each neuron node, nonlinear mapping is adopted to obtain the hidden layer node sequence of the next hidden layer;
Iterating the next hidden layer serving as the current processing hidden layer until the next hidden layer is an output layer; and acquiring the stability parameters corresponding to the target clients output by the output layer.
2. The method of claim 1, wherein monitoring the plurality of target information published by the network platform comprises:
Monitoring original information issued by a network platform;
Word segmentation processing is carried out on the original information to obtain an information tag corresponding to each piece of original information;
acquiring a plurality of target keywords, and identifying whether the information tag contains the target keywords;
If yes, marking the corresponding original information as target information.
3. The method of claim 1, wherein the calculating the public opinion index corresponding to the product identification based on the objective information comprises:
splitting the target information to obtain a plurality of short texts;
extracting a product identifier from the short text, and associating the product identifier with the corresponding short text;
calculating emotion indexes corresponding to each short text by using a preset public opinion analysis model;
Determining influence weights corresponding to the short texts respectively;
and calculating the public opinion index corresponding to the corresponding product identifier according to the emotion index and the influence weight of the associated short text.
4. The method of claim 1, wherein the counting access data of the target client to the target information during the monitoring period comprises:
Acquiring an identification field of a target client, and sending the identification field to the network platform;
receiving associated access data returned by the network platform according to the identification field;
extracting an information access field, an information collection field and an information forwarding field from the associated access data;
And counting the information access amount, the information collection amount and the information forwarding amount of the target client in the monitoring period based on the information access field, the information collection field and the information forwarding field.
5. The method of claim 1, wherein said determining emotional characteristics of said target client based on said client data comprises:
Acquiring a plurality of sub-models, and determining weight factors corresponding to the sub-models respectively;
Generating a first model according to the plurality of sub-models and the weight factors corresponding to the sub-models respectively;
acquiring client sample data and corresponding emotion labels;
inputting the client sample data into the first model to obtain an intermediate emotion analysis result;
calculating the difference between the intermediate emotion analysis result and the emotion label, and adjusting the first model according to the difference to obtain a second model;
And inputting the client data into the second model, and outputting the emotion characteristics of the target client.
6. A client stability prediction apparatus, the apparatus comprising:
the information analysis module is used for monitoring a plurality of target information issued by the network platform; acquiring client data corresponding to a target client; the customer data includes a product identifier; calculating a public opinion index corresponding to the product identifier based on the target information;
The client analysis module is used for counting the access data of the target client to the target information in the monitoring period; determining emotion characteristics of the target client according to the client data, wherein the emotion characteristics are used for representing quantitative or qualitative parameter values of the degree of emotion fluctuation caused by the influence of social public opinion;
The influence prediction module is used for preprocessing the public opinion index, the access data and the emotion characteristics to obtain a client characteristic matrix; obtaining an input layer node sequence according to the client feature matrix; projecting the input layer node sequence to obtain a hidden layer node sequence corresponding to a first hidden layer, and taking the first hidden layer as a current processing hidden layer; acquiring the weight and deviation of each neuron node corresponding to the current processing hidden layer; according to the hidden layer node sequence corresponding to the current processing hidden layer and the weight and deviation of each neuron node, nonlinear mapping is adopted to obtain the hidden layer node sequence of the next hidden layer; iterating the next hidden layer serving as the current processing hidden layer until the next hidden layer is an output layer; and acquiring the stability parameters corresponding to the target clients output by the output layer.
7. The apparatus of claim 6, wherein the information analysis module is further configured to monitor raw information published by a network platform; word segmentation is carried out on the original information, and an information label corresponding to each original information is obtained; acquiring a plurality of target keywords, and identifying whether the information tag contains the target keywords; if yes, marking the corresponding original information as target information.
8. The apparatus of claim 6, wherein the information analysis module is further to:
Splitting the target information to obtain a plurality of short texts; extracting a product identifier from the short text, and associating the product identifier with the corresponding short text; calculating emotion indexes corresponding to each short text by using a preset public opinion analysis model; determining influence weights corresponding to the short texts respectively; and calculating the public opinion index corresponding to the corresponding product identifier according to the emotion index and the influence weight of the associated short text.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any one of claims 1 to 5 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811025974.XA CN109376237B (en) | 2018-09-04 | 2018-09-04 | Client stability prediction method, device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811025974.XA CN109376237B (en) | 2018-09-04 | 2018-09-04 | Client stability prediction method, device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109376237A CN109376237A (en) | 2019-02-22 |
| CN109376237B true CN109376237B (en) | 2024-05-28 |
Family
ID=65405116
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811025974.XA Active CN109376237B (en) | 2018-09-04 | 2018-09-04 | Client stability prediction method, device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109376237B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110083759A (en) * | 2019-03-15 | 2019-08-02 | 深圳壹账通智能科技有限公司 | Public opinion information crawler method, apparatus, computer equipment and storage medium |
| RU2744032C2 (en) * | 2019-04-15 | 2021-03-02 | Общество С Ограниченной Ответственностью "Яндекс" | Method and system for determining result of task execution in crowdsourced environment |
| CN110348914A (en) * | 2019-07-19 | 2019-10-18 | 中国银行股份有限公司 | Customer churn data analysing method and device |
| CN110688553A (en) * | 2019-08-13 | 2020-01-14 | 平安科技（深圳）有限公司 | Information pushing method and device based on data analysis, computer equipment and storage medium |
| CN110956303A (en) * | 2019-10-12 | 2020-04-03 | 未鲲(上海)科技服务有限公司 | Information prediction method, device, terminal and readable storage medium |
| CN111158732B (en) * | 2019-12-23 | 2024-04-02 | 中国平安人寿保险股份有限公司 | Access data processing method, device, computer equipment and storage medium |
| RU2020107002A (en) | 2020-02-14 | 2021-08-16 | Общество С Ограниченной Ответственностью «Яндекс» | METHOD AND SYSTEM FOR RECEIVING A LABEL FOR A DIGITAL PROBLEM PERFORMED IN A CROWDSORING ENVIRONMENT |
| CN111506727B (en) * | 2020-04-16 | 2023-10-03 | 腾讯科技（深圳）有限公司 | Text content category acquisition method, apparatus, computer device and storage medium |
| CN113657635B (en) * | 2020-05-12 | 2023-10-27 | 中国移动通信集团湖南有限公司 | Method for predicting loss of communication user and electronic equipment |
| CN111950623B (en) * | 2020-08-10 | 2023-11-14 | 中国平安人寿保险股份有限公司 | Data stability monitoring method, device, computer equipment and medium |
| CN113643060A (en) * | 2021-08-12 | 2021-11-12 | 工银科技有限公司 | Product price prediction method and device |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544255A (en) * | 2013-10-15 | 2014-01-29 | 常州大学 | Text semantic relativity based network public opinion information analysis method |
| CN104462096A (en) * | 2013-09-13 | 2015-03-25 | 北大方正集团有限公司 | Public opinion monitoring and analysis method and device |
| CN106097114A (en) * | 2016-06-27 | 2016-11-09 | 江苏易乐网络科技有限公司 | A kind of game public sentiment monitoring and player's behavior analysis cloud service platform |
| CN106126558A (en) * | 2016-06-16 | 2016-11-16 | 东软集团股份有限公司 | A kind of public sentiment monitoring method and device |
| CN106649270A (en) * | 2016-12-19 | 2017-05-10 | 四川长虹电器股份有限公司 | Public opinion monitoring and analyzing method |
| CN107038178A (en) * | 2016-08-03 | 2017-08-11 | 平安科技（深圳）有限公司 | The analysis of public opinion method and apparatus |
| CN107341270A (en) * | 2017-07-28 | 2017-11-10 | 东北大学 | Towards the user feeling influence power analysis method of social platform |
| CN107391493A (en) * | 2017-08-04 | 2017-11-24 | 青木数字技术股份有限公司 | A kind of public feelings information extracting method, device, terminal device and storage medium |
| CN107908619A (en) * | 2017-11-15 | 2018-04-13 | 中国平安人寿保险股份有限公司 | Processing method, device, terminal and computer-readable storage medium based on public sentiment monitoring |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013059290A1 (en) * | 2011-10-17 | 2013-04-25 | Metavana, Inc. | Sentiment and influence analysis of twitter tweets |
| US20160162582A1 (en) * | 2014-12-09 | 2016-06-09 | Moodwire, Inc. | Method and system for conducting an opinion search engine and a display thereof |
| CA2923600A1 (en) * | 2015-03-12 | 2016-09-12 | Staples, Inc. | Review sentiment analysis |
-
2018
- 2018-09-04 CN CN201811025974.XA patent/CN109376237B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104462096A (en) * | 2013-09-13 | 2015-03-25 | 北大方正集团有限公司 | Public opinion monitoring and analysis method and device |
| CN103544255A (en) * | 2013-10-15 | 2014-01-29 | 常州大学 | Text semantic relativity based network public opinion information analysis method |
| CN106126558A (en) * | 2016-06-16 | 2016-11-16 | 东软集团股份有限公司 | A kind of public sentiment monitoring method and device |
| CN106097114A (en) * | 2016-06-27 | 2016-11-09 | 江苏易乐网络科技有限公司 | A kind of game public sentiment monitoring and player's behavior analysis cloud service platform |
| CN107038178A (en) * | 2016-08-03 | 2017-08-11 | 平安科技（深圳）有限公司 | The analysis of public opinion method and apparatus |
| CN106649270A (en) * | 2016-12-19 | 2017-05-10 | 四川长虹电器股份有限公司 | Public opinion monitoring and analyzing method |
| CN107341270A (en) * | 2017-07-28 | 2017-11-10 | 东北大学 | Towards the user feeling influence power analysis method of social platform |
| CN107391493A (en) * | 2017-08-04 | 2017-11-24 | 青木数字技术股份有限公司 | A kind of public feelings information extracting method, device, terminal device and storage medium |
| CN107908619A (en) * | 2017-11-15 | 2018-04-13 | 中国平安人寿保险股份有限公司 | Processing method, device, terminal and computer-readable storage medium based on public sentiment monitoring |
Non-Patent Citations (1)
| Title |
|---|
| 基于内容和信任度的舆情扩散研究;陈振春;刘学军;李斌;《计算机应用与软件》;第34卷(第10期);第59-65页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109376237A (en) | 2019-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109376237B (en) | Client stability prediction method, device, computer equipment and storage medium | |
| CN109241427B (en) | Information pushing method, device, computer equipment and storage medium | |
| CN108874992B (en) | Public opinion analysis method, system, computer equipment and storage medium | |
| CN109165840B (en) | Risk prediction processing method, risk prediction processing device, computer equipment and medium | |
| US20230334254A1 (en) | Fact checking | |
| Alam et al. | Processing social media images by combining human and machine computing during crises | |
| CN109272396B (en) | Customer risk early warning method, device, computer equipment and medium | |
| CN108563722B (en) | Industry classification method, system, computer device and storage medium for text information | |
| CN109829629B (en) | Risk analysis report generation method, apparatus, computer device and storage medium | |
| WO2021027317A1 (en) | Relationship network-based attribute information processing method and device, computer apparatus, and storage medium | |
| CN108509424B (en) | System information processing method, apparatus, computer device and storage medium | |
| CN109543925B (en) | Risk prediction method and device based on machine learning, computer equipment and storage medium | |
| CN108460582B (en) | System information processing method, apparatus, computer device and storage medium | |
| CN109582876B (en) | Tourist industry user portrait construction method and device and computer equipment | |
| CN110880006B (en) | User classification method, apparatus, computer device and storage medium | |
| Nawrocka et al. | Application of machine learning in recommendation systems | |
| CN111259985A (en) | Classification model training method and device based on business safety and storage medium | |
| CN112784168B (en) | Information push model training method and device, information push method and device | |
| CN114238573A (en) | Information pushing method and device based on text countermeasure sample | |
| CN113869931A (en) | Advertisement putting strategy determining method and device, computer equipment and storage medium | |
| CN112288279A (en) | Business risk assessment method and device based on natural language processing and linear regression | |
| Komalavalli et al. | Challenges in big data analytics techniques: a survey | |
| CN116049379A (en) | Knowledge recommendation method, knowledge recommendation device, electronic equipment and storage medium | |
| CN110968780B (en) | Page content recommendation method and device, computer equipment and storage medium | |
| WO2021081914A1 (en) | Pushing object determination method and apparatus, terminal device and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |