CN111127179A - Information pushing method and device, computer equipment and storage medium - Google Patents
Information pushing method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN111127179A CN111127179A CN201911277050.3A CN201911277050A CN111127179A CN 111127179 A CN111127179 A CN 111127179A CN 201911277050 A CN201911277050 A CN 201911277050A CN 111127179 A CN111127179 A CN 111127179A
- Authority
- CN
- China
- Prior art keywords
- sample
- processed
- label
- probability
- samples
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/12—Accounting
- G06Q40/123—Tax preparation or submission
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The embodiment of the application provides an information pushing method, an information pushing device, computer equipment and a storage medium, wherein the method comprises the steps of obtaining a sample set to be processed; generating a sample characteristic vector for each sample to be processed in a sample set to be processed, and predicting the initial probability of the sample to be processed under each sample label in a preset sample label set; determining part of samples to be processed from the sample set to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set; and selecting a target label for each sample to be processed from the sample label set and pushing the target label based on the sample feature vector corresponding to each sample to be processed, the real labels corresponding to partial samples to be processed and the label probability prediction model corresponding to each sample label in the sample label set.
Description
Technical Field
The present application relates to the field of data processing technologies, and in particular, to an information pushing method, an information pushing apparatus, a computer device, and a storage medium.
Background
In model training, labels are generally required to be labeled for samples of input models, the labeled contents are usually phrases of several words and can cover subjects with meaningful texts, such as questions about tax return in customer service, and can be labeled as tax return forms, tax compensation subjects and the like.
The process of labeling the sample is generally realized in a manual labeling mode, time and labor are consumed, when the sample size is large, a labeling error phenomenon may occur in manual labeling, and the accuracy of the finally obtained sample label is reduced.
Disclosure of Invention
In view of the above, an object of the embodiments of the present application is to provide an information pushing method, an information pushing apparatus, a computer device, and a storage medium, so as to improve efficiency of pushing information.
The first invention, an embodiment of the present application provides an information pushing apparatus, including:
the acquisition module is used for acquiring a sample set to be processed;
the first processing module is used for generating a sample characteristic vector for each sample to be processed in the sample set to be processed and predicting the initial probability of the sample to be processed under each sample label in a preset sample label set;
the determination module is used for determining part of samples to be processed from the sample set to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set;
and the second processing module is used for selecting a target label for each sample to be processed from the sample label set and pushing the target label based on the sample feature vector corresponding to each sample to be processed, the real labels corresponding to partial samples to be processed and the label probability prediction model corresponding to each sample label in the sample label set.
In one embodiment, the first processing module is configured to generate a sample feature vector for the sample to be processed according to the following steps:
for each sample to be processed, performing word segmentation processing on the sample to be processed to obtain a word set corresponding to the sample to be processed;
generating a vocabulary vector for each vocabulary in the vocabulary set corresponding to the sample to be processed;
and carrying out weighting processing on the vocabulary vector corresponding to each vocabulary in the vocabulary set corresponding to the sample to be processed to obtain the sample characteristic vector corresponding to the sample to be processed.
In one embodiment, the first processing module is configured to predict an initial probability of the sample to be processed under each sample label in a preset sample label set according to the following steps:
and for each sample label in the sample label set, inputting the sample feature vector of the sample to be processed into a first label probability prediction model corresponding to the sample label, and predicting to obtain the initial probability of the sample to be processed under the sample label.
In one embodiment, the first processing module is further configured to:
determining the maximum initial probability as a target initial probability corresponding to a sample to be processed for each sample to be processed in a sample set to be processed, and taking a sample label corresponding to the target initial probability as a first label selected for the sample to be processed;
determining a first number of samples to be processed belonging to the same sample label based on the first label selected for each sample to be processed;
and adjusting the model parameters of the first label probability prediction model corresponding to each sample label based on the initial probability corresponding to each sample to be processed and the ratio of the first number corresponding to each sample label in the total number of samples to obtain the adjusted model parameters corresponding to each sample label.
In one embodiment, the second processing module is configured to select a target label for each specimen to be processed from the set of specimen labels according to the following steps:
for each sample label, taking the sample feature vector of each sample to be processed as the model input of a label probability prediction model corresponding to the sample label, and predicting to obtain a first probability of each sample to be processed under the sample label;
adjusting model parameters of a label probability prediction model corresponding to each sample label based on the first probability of each sample to be processed under each sample label and the real labels of part of samples to be processed;
aiming at each sample label, taking the adjusted model parameter corresponding to the sample label as the final model parameter of the label probability prediction model;
taking the sample characteristic vector of each sample to be processed as the model input of a label probability prediction model corresponding to the sample label, and predicting to obtain the final probability of each sample to be processed under the sample label;
and for each sample to be processed, determining the sample label corresponding to the maximum final probability as the target sample label corresponding to the sample to be processed.
In one embodiment, the second processing module is configured to adjust model parameters of a label probability prediction model corresponding to each sample label according to the following steps:
determining the maximum first probability as a target first probability corresponding to the sample to be processed for each sample to be processed in the sample set to be processed, and taking a sample label corresponding to the target first probability as a second label selected for the sample to be processed;
determining a second number of partial samples to be processed belonging to the same sample label based on the second label selected for each sample to be processed;
and adjusting the model parameters of the label probability prediction model corresponding to each sample label based on the real labels respectively corresponding to part of samples to be processed and the real probabilities under the corresponding real labels, the second probabilities respectively corresponding to other processed samples except part of samples to be processed in the sample set to be processed, and the ratio of the second number corresponding to each sample label to the total number of the samples.
In one embodiment, the determining module is configured to determine the portion of the sample to be processed according to the following steps:
for each sample to be processed, sequencing the initial probabilities corresponding to the samples to be processed according to the sequence of the probabilities from large to small;
determining the difference value between the first initial probability and the second initial probability in the initial probability sequence corresponding to the sample to be processed;
and determining the sample to be processed corresponding to the difference value smaller than the preset threshold value as part of the sample to be processed.
In a second aspect, an embodiment of the present application provides an information pushing method, where the method includes:
acquiring a sample set to be processed;
generating a sample characteristic vector for each sample to be processed in a sample set to be processed, and predicting the initial probability of the sample to be processed under each sample label in a preset sample label set;
determining part of samples to be processed from the sample set to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set;
and selecting a target label for each sample to be processed from the sample label set and pushing the target label based on the sample feature vector corresponding to each sample to be processed, the real labels corresponding to partial samples to be processed and the label probability prediction model corresponding to each sample label in the sample label set.
In a third aspect, an embodiment of the present application provides a computer device, including: the data processing method comprises a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, when a computer device runs, the processor and the storage medium communicate through the bus, and the processor executes the machine-readable instructions to execute the steps of the data processing method.
In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to perform the steps of the data processing method.
The information pushing method provided by the embodiment of the application obtains a sample set to be processed, generates a sample feature vector for each sample to be processed in the sample set to be processed, predicts the initial probability of the sample to be processed under each sample label in a preset sample label set, determines part of samples to be processed from the sample set to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set, selects a target label for each sample to be processed from the sample label set based on the sample feature vector corresponding to each sample to be processed and the real label corresponding to part of samples to be processed respectively, and a label probability prediction model corresponding to each sample label in the sample label set, so that when determining the target label for the sample to be processed, the real label and the sample feature vector of the sample to be processed are considered, the accuracy of the sample label selected for each sample to be processed is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.
Fig. 1 illustrates a first flowchart of an information pushing method provided by an embodiment of the present application;
fig. 2 is a schematic structural diagram illustrating an information pushing apparatus according to an embodiment of the present application;
fig. 3 shows a schematic structural diagram of a computer device provided in an embodiment of the present application.
Detailed Description
In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it should be understood that the drawings in the present application are for illustrative and descriptive purposes only and are not used to limit the scope of protection of the present application. Additionally, it should be understood that the schematic drawings are not necessarily drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of the present application. It should be understood that the operations of the flow diagrams may be performed out of order, and steps without logical context may be performed in reverse order or simultaneously. One skilled in the art, under the guidance of this application, may add one or more other operations to, or remove one or more operations from, the flowchart.
In addition, the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application.
It should be noted that in the embodiments of the present application, the term "comprising" is used to indicate the presence of the features stated hereinafter, but does not exclude the addition of further features.
In the related art, the text clustering methods include a neighbor algorithm (KNN) method and a K-means clustering algorithm (K-means) method, however, these two clustering methods involve a large amount of calculation, and cannot be applied to the calculation of data sets with large data volume or large dimensionality, and these two methods cannot well capture the distribution of data, and cannot provide other information about the data, so that the efficiency is not high.
Based on this, the information push method provided in the embodiment of the present application obtains a sample set to be processed, generates a sample feature vector for each sample to be processed in the sample set to be processed, predicts an initial probability of the sample to be processed under each sample label in a preset sample label set, determines a part of samples to be processed from the sample set to be processed based on the initial probability of each sample label in the sample label set, selects a target label for each sample to be processed from the sample label set based on a sample feature vector corresponding to each sample to be processed and a real label corresponding to each part of samples to be processed, and a label probability prediction model corresponding to each sample label in the sample label set, so that when determining the target label for the sample to be processed, the real label and the sample feature vector of the sample to be processed are considered, the target label is automatically determined for the samples to be processed, improving the accuracy of the sample label selected for each sample to be processed.
The information pushing method of the embodiment of the application can be applied to a server and can also be applied to any other computing equipment with a processing function. In some embodiments, the server or computing device may include a processor. The processor may process information and/or data related to the service request to perform one or more of the functions described herein.
An embodiment of the present application provides an information push method, as shown in fig. 1, the method specifically includes the following steps:
s101, obtaining a sample set to be processed;
s102, aiming at each sample to be processed in a sample set to be processed, generating a sample characteristic vector for the sample to be processed, and predicting the initial probability of the sample to be processed under each sample label in a preset sample label set;
s103, determining part of samples to be processed from the sample set to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set;
and S104, selecting a target label for each sample to be processed from the sample label set and pushing the target label based on the sample feature vector corresponding to each sample to be processed, the real labels corresponding to partial samples to be processed and the label probability prediction model corresponding to each sample label in the sample label set.
In S101, the to-be-processed sample set includes a plurality of to-be-processed samples, where the to-be-processed samples may be text samples or chat records in different fields, the text samples may be articles, paragraphs in articles, and the like, and the text samples may be text samples in a communication field, a financial field, an education field, and an accounting field; the chat records can be conversation records among different users, and the like; the pending sample set may be sent by the requesting peer.
In S102, the sample feature vector characterizes a feature vector of the meaning of the sample; the sample label set is preset, the initial probability represents the probability that the sample to be processed is any sample label in the sample label set, the larger the probability is, the higher the probability that the sample to be processed is any sample label is.
Generating a sample feature vector for each sample to be processed in the set of samples to be processed, may include the following steps:
for each sample to be processed, performing word segmentation processing on the sample to be processed to obtain a word set corresponding to the sample to be processed;
generating a vocabulary vector for each vocabulary in the vocabulary set corresponding to the sample to be processed;
and carrying out weighting processing on the vocabulary vector corresponding to each vocabulary in the vocabulary set corresponding to the sample to be processed to obtain the sample characteristic vector corresponding to the sample to be processed.
Here, the tool for performing word segmentation processing on each sample to be processed includes a final word segmentation tool, an nltk word segmentation tool, and the like; the vocabulary set comprises a plurality of vocabularies; the lexical vector characterizes a vector of lexical meanings.
In the specific implementation process, a word segmentation tool is used for carrying out word segmentation on each sample to be processed to obtain a word set corresponding to each sample to be processed.
For each sample to be processed, generating a vocabulary vector for each vocabulary in the vocabulary set corresponding to the sample to be processed by using a vocabulary vector generation model, weighting and averaging each vocabulary vector, namely, calculating the product of each vocabulary vector and the corresponding weight, and further calculating the sum of the vocabulary vectors after each product, thereby obtaining the sample vector corresponding to each sample to be processed. The vocabulary vector generation model can be a GloVe model, when the vocabulary vectors in the vocabulary set are weighted and averaged, the weight corresponding to each vocabulary vector can be determined according to the frequency of the occurrence of the vocabulary corresponding to the vocabulary vector in the sample to be processed, and the higher the frequency is, the larger the weight corresponding to the vocabulary vector is.
When the GloVe model generates the vocabulary vectors for the vocabularies in the vocabulary set, the context relationship between the vocabularies is considered, the more the vocabularies appear, the closer the relationship between the vocabularies is, and therefore, the vocabulary vectors generated by the GloVe model for each vocabulary have the context relationship between the vocabularies.
After obtaining the sample feature vector of each sample to be processed, predicting the initial probability of the sample to be processed under each sample label in a preset sample label set aiming at each sample to be processed in the sample set to be processed, including:
for each sample label in the sample label set, inputting the sample feature vector of the sample to be processed into a first label probability prediction model corresponding to the sample label, and predicting to obtain the initial probability of the sample to be processed under the sample label;
determining the maximum initial probability as a target initial probability corresponding to a sample to be processed for each sample to be processed in a sample set to be processed, and taking a sample label corresponding to the target initial probability as a first label selected for the sample to be processed;
determining a first number of samples to be processed belonging to the same sample label based on the first label selected for each sample to be processed;
and adjusting the model parameters of the first label probability prediction model corresponding to each sample label based on the initial probability corresponding to each sample to be processed and the ratio of the first number corresponding to each sample label in the total number of samples to obtain the adjusted model parameters corresponding to each sample label.
Here, the first label probability prediction model may be a generalized hyperbolic distribution model, and model parameters of the first label probability prediction model are set according to historical experience; the first label is an initial specimen label determined for the specimen to be processed.
In the related art, a traditional distributed model has strong interpretability, however, gaussian distribution in the related art requires high symmetry and cannot be compatible with defects such as abnormal values, and generally processed text data has a certain distribution, which is also a main assumption of all natural language processing. However, the general distribution (e.g., gaussian distribution) cannot well describe the distribution of text data after being converted into a digital vector, and therefore, a more flexible distribution is needed to describe the text data, so that a generalized hyperbolic distribution model is proposed, which has excellent distribution and compatibility, can be compatible with symmetric and asymmetric distributions, and has a very high degree of containment for abnormal values.
In a specific implementation process, for each sample label in a sample label set, the sample feature vector of the sample to be processed is input into a first label probability prediction model corresponding to the sample label, and the initial probability of the sample to be processed under the sample label is obtained through prediction.
The model parameters in the first label probability prediction model may be adjusted in the following manner to improve the accuracy of the predicted initial probability.
After the initial probability of each sample to be processed under each sample label is obtained, for each sample to be processed, the maximum initial probability is determined as the target initial probability corresponding to the sample to be processed, and further, the sample label corresponding to the target initial probability is used as the first label selected for the sample to be processed.
For example, the set of samples to be processed includes two samples to be processed, which are S1 and S2, respectively, the preset sample tag set includes two sample tags, which are T1 and T2, respectively, the initial probabilities of the sample S1 under T1 and T2 are α 1 and α 2, the initial probabilities of the sample S2 under T1 and T2 are α 3 and α 4, respectively, the maximum initial probability corresponding to the sample S1 is α 1, the maximum initial probability corresponding to the sample S2 is α 4, the maximum initial probability corresponding to α 1 is T1, then T1 is the first tag of the sample S1, the sample tag corresponding to the maximum initial probability α 4 is T2, and then T2 is the first tag of the sample S2.
Counting the first number of the to-be-processed samples corresponding to each sample label in the sample label set, calculating the ratio of the first number corresponding to each sample label to the total number of the samples in the to-be-processed sample set, taking the ratio as an occupation ratio, and further determining the first total loss value of the first label probability prediction model corresponding to each sample label by using the corresponding initial probability of each to-be-processed sample, the occupation ratio corresponding to each sample label and the probability function corresponding to each sample label, so that the first total loss value maximally adjusts the model parameter of the first label probability prediction model corresponding to each sample label.
The first loss value satisfies the following equation:
wherein, L is a first total loss value of the first label probability prediction model corresponding to each sample label; pgThe ratio of the first number of the samples to be processed corresponding to the g-th sample label to the total number of the samples; f. ofg(xi| θ) is a model probability distribution function corresponding to the ith sample to be processed corresponding to the g-th sample label when the model parameter θ corresponding to the g-th sample label is the model parameter θ; zigThe initial probability of the ith sample to be processed under the g sample label is obtained; n is the total number of samples in the sample set to be processed; g is the total number of sample tags in the sample tag set.
In S103, the number of partial samples to be processed may be determined from the history data.
When determining a part of the samples to be processed from the sample set to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set, the method may include the following steps:
for each sample to be processed, sequencing the initial probabilities corresponding to the samples to be processed according to the sequence of the probabilities from large to small;
determining the difference value between the first initial probability and the second initial probability in the initial probability sequence corresponding to the sample to be processed;
and determining the sample to be processed corresponding to the difference value smaller than the preset threshold value as part of the sample to be processed.
Here, the preset threshold may be determined based on historical data, and may be determined based on actual conditions.
In a specific implementation process, after the initial probability of each sample to be processed under each sample label is obtained, the initial probabilities corresponding to the samples to be processed are ranked according to a descending order for each sample to be processed, a difference value between a first initial probability and a second initial probability in the initial probability ranking corresponding to the samples to be processed is calculated, and the samples to be processed corresponding to the difference value smaller than a preset threshold value are selected as part of the samples to be processed.
For example, ten samples to be processed are included in the sample set to be processed, which are respectively S1 and S2 … … S10, the preset sample tag set includes two sample tags, which are respectively T1, T2 and T3, the initial probabilities of the sample S1 under T1, T2 and T3 are sorted in descending order into α 1, α 2 and α 3, the difference between α 1 and α 2 is calculated as P1, the calculation manner of the other samples S2 to S10 is the same as that of the sample S1, the differences corresponding to the samples S2, S3 and … … S10 are respectively P2 and P3 … … P10, the preset threshold is O, and the differences smaller than the preset threshold are P1, P2 and P5, so that part of the samples to be processed are S1, S2 and S5.
In S104, the real label is a label manually labeled for the sample to be processed, that is, an actual label of the sample to be processed; the label probability prediction model is a generalized hyperbolic distribution model.
When selecting a target label for each sample to be processed from the sample label set based on the sample feature vector corresponding to each sample to be processed, the real labels corresponding to a part of samples to be processed, and the label probability prediction model corresponding to each sample label in the sample label set, the method may include the following steps:
for each sample label, taking the sample feature vector of each sample to be processed as the model input of a label probability prediction model corresponding to the sample label, and predicting to obtain a first probability of each sample to be processed under the sample label;
adjusting model parameters of a label probability prediction model corresponding to each sample label based on the first probability of each sample to be processed under each sample label and the real labels of part of samples to be processed;
aiming at each sample label, taking the adjusted model parameter corresponding to the sample label as the final model parameter of the label probability prediction model;
taking the sample characteristic vector of each sample to be processed as the model input of a label probability prediction model corresponding to the sample label, and predicting to obtain the final probability of each sample to be processed under the sample label;
and for each sample to be processed, determining the sample label corresponding to the maximum final probability as the target sample label corresponding to the sample to be processed.
Here, the first probability is the probability of the sample to be processed under the sample label, which is obtained by prediction of the label probability prediction model, and the greater the probability, the higher the possibility that the sample to be processed is the sample label is;
in a specific implementation process, the initial model parameters of the label probability prediction model may be determined according to historical data, or may be adjusted model parameters corresponding to the first label probability prediction model. In the actual implementation process, in order to improve the accuracy of the probability obtained by the prediction of the label probability prediction model, the above adjusted model parameters are used as the initial model parameters of the label probability prediction model.
And for each sample label, taking the sample feature vector of each sample to be processed as the model input of the label probability prediction model corresponding to the sample label, thereby predicting and obtaining the first probability of each sample to be processed under the sample label.
Based on the first probability of each sample to be processed under each sample label and the real labels of part of samples to be processed, the initial model parameters of the label probability prediction model corresponding to each sample label are adjusted so as to improve the prediction accuracy of the label probability prediction model.
When the model parameters of the label probability prediction model corresponding to each sample label are adjusted based on the second probability of each sample to be processed under each sample label and the real label of a part of samples to be processed, the method may include the following steps:
determining the maximum first probability as a target first probability corresponding to the sample to be processed for each sample to be processed in the sample set to be processed, and taking a sample label corresponding to the target first probability as a second label selected for the sample to be processed;
determining a second number of samples to be processed belonging to the same sample label based on the second label selected for each sample to be processed;
and adjusting the model parameters of the label probability prediction model corresponding to each sample label based on the real labels respectively corresponding to part of the samples to be processed and the real probabilities under the corresponding real labels, the second probabilities respectively corresponding to other processed samples except part of the samples to be processed in the sample set to be processed, and the ratio of the number of the other processed samples corresponding to each sample label to the total number of the samples.
Here, the first probability is the probability of the sample to be processed under the corresponding sample label, which is obtained by the label probability prediction model; the second label refers to a sample label selected for the sample to be treated; the second number is the total number of samples to be processed corresponding to the sample label;
in a specific implementation process, for each sample to be processed in the sample set to be processed, the maximum first probability is determined as the target first probability corresponding to the sample to be processed, that is, the maximum first probability is selected as the target first probability from the first probabilities under different sample labels corresponding to the sample to be processed, and further, the sample label corresponding to the target first probability is used as the second label selected for the sample to be processed. In this way, each sample to be processed corresponds to one sample label, the second number of the samples to be processed corresponding to each sample label is counted, the ratio of the second number corresponding to each sample label to the total number of the samples is calculated, and the ratio is used as the first proportion; and calculating a third number of other samples to be processed except for part of the samples to be processed corresponding to each sample label, calculating a ratio of the third number corresponding to each sample label to the total number of the samples, and taking the ratio as a second ratio.
And determining a second total loss value of the label probability prediction model corresponding to each sample label based on the real labels respectively corresponding to part of samples to be processed and the real probabilities under the corresponding real labels, as well as second probabilities respectively corresponding to other samples to be processed except part of samples to be processed in the sample set to be processed, and the ratio of the second number to the total number of the samples and the ratio of the third number to the total number of the samples, which correspond to each sample label, so that the second total loss value is maximum, and adjusting the model parameters of the label probability prediction model corresponding to each sample label.
The second loss value satisfies the following equation:
wherein S is a second total loss value, O, of the label probability prediction model corresponding to each sample labelgThe ratio of the second number of the part of the samples to be processed corresponding to the g sample label to the total number of the samples; ggThe ratio of the third number of the samples to be processed except the part of the samples to be processed corresponding to the g sample label to the total number of the samples; f. ofg(xi| θ) is a model probability distribution function corresponding to the ith sample to be processed (other samples to be processed except part of samples to be processed) corresponding to the g-th sample label when the model parameter θ corresponding to the g-th sample label is the model parameter θ; f. of0g(xi| θ) is a model probability distribution function corresponding to the ith part to-be-processed sample corresponding to the g-th sample label when the model parameter θ corresponding to the g-th sample label is the model parameter θ; a. theigThe true probability of the ith part of the sample to be processed under the g sample label is obtained; kigThe first probability of the ith sample to be processed except part of the sample to be processed under the g sample label is obtained; n is the total number of samples in the sample set to be processed; g is the total number of sample labels in the sample label setNumber: k is the total number of partial samples to be processed.
And aiming at each sample label, taking the adjusted model parameter corresponding to the sample label as the final model parameter of the label probability prediction model, further taking the sample feature vector of each sample to be processed as the model input of the label probability prediction model corresponding to the sample label, and predicting to obtain the final probability of each sample to be processed under the sample label. And for each sample to be processed, determining the sample label corresponding to the maximum final probability as the target sample label corresponding to the sample to be processed.
For example, the set of samples to be processed includes ten samples to be processed, which are respectively S1 and S2 … … S10, the preset sample label set includes three sample labels, which are respectively T1, T2 and T3, the final probabilities of the sample to be processed S1 under T1, T2 and T3 are α 10, α 20 and α 30, where α 10 is the maximum final probability, T1 is the label of the sample to be processed S1, and the calculation manners of the other samples to be processed S2 to S10 are the same as the calculation manner of the sample to be processed S1, which is not described herein again.
Based on the same inventive concept, an information pushing device corresponding to the information pushing method is further provided in the embodiment of the present application, and as the principle of solving the problem of the method in the embodiment of the present application is similar to that of the information pushing method in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and repeated details are not repeated.
An embodiment of the present application provides an information pushing apparatus, as shown in fig. 2, the apparatus includes:
an obtaining module 21, configured to obtain a sample set to be processed;
the first processing module 22 is configured to generate a sample feature vector for each sample to be processed in the sample set to be processed, and predict an initial probability of the sample to be processed under each sample label in a preset sample label set;
a determining module 23, configured to determine, based on an initial probability of each sample to be processed under each sample label in a sample label set, a part of samples to be processed from the sample set to be processed;
and the second processing module 24 is configured to select a target label for each sample to be processed from the sample label set based on the sample feature vector corresponding to each sample to be processed, the real labels corresponding to a part of the samples to be processed, and the label probability prediction model corresponding to each sample label in the sample label set, and push the target label.
Optionally, the first processing module 22 is configured to generate a sample feature vector for the sample to be processed according to the following steps:
for each sample to be processed, performing word segmentation processing on the sample to be processed to obtain a word set corresponding to the sample to be processed;
generating a vocabulary vector for each vocabulary in the vocabulary set corresponding to the sample to be processed;
and carrying out weighting processing on the vocabulary vector corresponding to each vocabulary in the vocabulary set corresponding to the sample to be processed to obtain the sample characteristic vector corresponding to the sample to be processed.
Optionally, the first processing module 22 is configured to predict an initial probability of the sample to be processed under each sample label in a preset sample label set according to the following steps:
and for each sample label in the sample label set, inputting the sample feature vector of the sample to be processed into a first label probability prediction model corresponding to the sample label, and predicting to obtain the initial probability of the sample to be processed under the sample label.
Optionally, the first processing module 22 is further configured to:
determining the maximum initial probability as a target initial probability corresponding to a sample to be processed for each sample to be processed in a sample set to be processed, and taking a sample label corresponding to the target initial probability as a first label selected for the sample to be processed;
determining a first number of samples to be processed belonging to the same sample label based on the first label selected for each sample to be processed;
and adjusting the model parameters of the first label probability prediction model corresponding to each sample label based on the initial probability corresponding to each sample to be processed and the ratio of the first number corresponding to each sample label in the total number of samples to obtain the adjusted model parameters corresponding to each sample label.
Optionally, the second processing module 24 is configured to select a target label for each sample to be processed from the sample label set according to the following steps:
for each sample label, taking the sample feature vector of each sample to be processed as the model input of a label probability prediction model corresponding to the sample label, and predicting to obtain a first probability of each sample to be processed under the sample label;
adjusting model parameters of a label probability prediction model corresponding to each sample label based on the first probability of each sample to be processed under each sample label and the real labels of part of samples to be processed;
aiming at each sample label, taking the adjusted model parameter corresponding to the sample label as the final model parameter of the label probability prediction model;
taking the sample characteristic vector of each sample to be processed as the model input of a label probability prediction model corresponding to the sample label, and predicting to obtain the final probability of each sample to be processed under the sample label;
and for each sample to be processed, determining the sample label corresponding to the maximum final probability as the target sample label corresponding to the sample to be processed.
Optionally, the second processing module 24 is configured to adjust model parameters of the label probability prediction model corresponding to each sample label according to the following steps:
determining the maximum first probability as a target first probability corresponding to the sample to be processed for each sample to be processed in the sample set to be processed, and taking a sample label corresponding to the target first probability as a second label selected for the sample to be processed;
determining a second number of partial samples to be processed belonging to the same sample label based on the second label selected for each sample to be processed;
and adjusting the model parameters of the label probability prediction model corresponding to each sample label based on the real labels respectively corresponding to part of samples to be processed and the real probabilities under the corresponding real labels, the second probabilities respectively corresponding to other processed samples except part of samples to be processed in the sample set to be processed, and the ratio of the second number corresponding to each sample label to the total number of the samples.
Optionally, the determining module 23 is configured to determine the part of the sample to be processed according to the following steps:
for each sample to be processed, sequencing the initial probabilities corresponding to the samples to be processed according to the sequence of the probabilities from large to small;
determining the difference value between the first initial probability and the second initial probability in the initial probability sequence corresponding to the sample to be processed;
and determining the sample to be processed corresponding to the difference value smaller than the preset threshold value as part of the sample to be processed.
Corresponding to the information pushing method in fig. 1, an embodiment of the present application further provides a computer device 300, as shown in fig. 3, the device includes a memory 301, a processor 302, and a computer program stored on the memory 301 and executable on the processor 302, where the processor 302 implements the information pushing method when executing the computer program.
Specifically, the memory 301 and the processor 302 can be general memories and processors, which are not specifically limited herein, and when the processor 302 runs a computer program stored in the memory 301, the information push method can be executed to solve the problem of low information push accuracy in the prior art, the present application obtains a sample set to be processed, generates a sample feature vector for each sample to be processed in the sample set to be processed, and predicts an initial probability of the sample to be processed under each sample label in a preset sample label set, determines a part of samples to be processed from the sample set to be processed based on the initial probability of the sample to be processed under each sample label in the sample label set, determines a real label corresponding to each sample to be processed and a sample feature vector corresponding to each sample to be processed, and a label probability prediction model corresponding to each sample label in the sample label set, and selecting a target label for each sample to be processed from the sample label set, so that when the target label is determined for the sample to be processed, the real label and the sample characteristic vector of the sample to be processed are considered, and the accuracy of the sample label selected for each sample to be processed is improved.
Corresponding to the information pushing method in fig. 1, an embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and the computer program is executed by a processor to perform the steps of the information pushing method.
Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, and the like, and when executed, the storage medium can execute the above information pushing method to solve the problem of low information pushing accuracy in the prior art, the present application obtains a set of samples to be processed, generates a sample feature vector for each sample to be processed in the set of samples to be processed, and predicts an initial probability of the sample to be processed under each sample label in a preset sample label set, determines a part of samples to be processed from the set of samples to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set, determines a true label corresponding to each sample to be processed and a sample feature vector corresponding to each sample to be processed, and a label probability prediction model corresponding to each sample label in the sample label set, and selecting a target label for each sample to be processed from the sample label set, so that when the target label is determined for the sample to be processed, the real label and the sample characteristic vector of the sample to be processed are considered, and the accuracy of the sample label selected for each sample to be processed is improved.
It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the system and the apparatus described above may refer to corresponding processes in the method embodiments, and are not described in detail in this application. In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. The above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a logical division, and there may be other divisions in actual implementation, and for example, a plurality of modules or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or modules through some communication interfaces, and may be in an electrical, mechanical or other form.
The modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical units, may be located in one place, or may be distributed on a plurality of road network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a U disk, a removable hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. An information pushing apparatus, comprising:
the acquisition module is used for acquiring a sample set to be processed;
the first processing module is used for generating a sample characteristic vector for each sample to be processed in the sample set to be processed and predicting the initial probability of the sample to be processed under each sample label in a preset sample label set;
the determination module is used for determining part of samples to be processed from the sample set to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set;
and the second processing module is used for selecting a target label for each sample to be processed from the sample label set and pushing the target label based on the sample feature vector corresponding to each sample to be processed, the real labels corresponding to partial samples to be processed and the label probability prediction model corresponding to each sample label in the sample label set.
2. The apparatus of claim 1, wherein the first processing module is configured to generate a sample feature vector for the sample to be processed according to the following steps:
for each sample to be processed, performing word segmentation processing on the sample to be processed to obtain a word set corresponding to the sample to be processed;
generating a vocabulary vector for each vocabulary in the vocabulary set corresponding to the sample to be processed;
and carrying out weighting processing on the vocabulary vector corresponding to each vocabulary in the vocabulary set corresponding to the sample to be processed to obtain the sample characteristic vector corresponding to the sample to be processed.
3. The apparatus of claim 1, wherein the first processing module is configured to predict the initial probability of the sample to be processed under each sample label in a predetermined sample label set according to the following steps:
and for each sample label in the sample label set, inputting the sample feature vector of the sample to be processed into a first label probability prediction model corresponding to the sample label, and predicting to obtain the initial probability of the sample to be processed under the sample label.
4. The apparatus of claim 3, wherein the first processing module is further to:
determining the maximum initial probability as a target initial probability corresponding to a sample to be processed for each sample to be processed in a sample set to be processed, and taking a sample label corresponding to the target initial probability as a first label selected for the sample to be processed;
determining a first number of samples to be processed belonging to the same sample label based on the first label selected for each sample to be processed;
and adjusting the model parameters of the first label probability prediction model corresponding to each sample label based on the initial probability corresponding to each sample to be processed and the ratio of the first number corresponding to each sample label in the total number of samples to obtain the adjusted model parameters corresponding to each sample label.
5. The apparatus of claim 1, wherein the second processing module is configured to select a target label for each specimen to be processed from the set of specimen labels according to the following steps:
for each sample label, taking the sample feature vector of each sample to be processed as the model input of a label probability prediction model corresponding to the sample label, and predicting to obtain a first probability of each sample to be processed under the sample label;
adjusting model parameters of a label probability prediction model corresponding to each sample label based on the first probability of each sample to be processed under each sample label and the real labels of part of samples to be processed;
aiming at each sample label, taking the adjusted model parameter corresponding to the sample label as the final model parameter of the label probability prediction model;
taking the sample characteristic vector of each sample to be processed as the model input of a label probability prediction model corresponding to the sample label, and predicting to obtain the final probability of each sample to be processed under the sample label;
and for each sample to be processed, determining the sample label corresponding to the maximum final probability as the target sample label corresponding to the sample to be processed.
6. The apparatus of claim 1, wherein the second processing module is configured to adjust model parameters of the label probability prediction model corresponding to each sample label according to the following steps:
determining the maximum first probability as a target first probability corresponding to the sample to be processed for each sample to be processed in the sample set to be processed, and taking a sample label corresponding to the target first probability as a second label selected for the sample to be processed;
determining a second number of partial samples to be processed belonging to the same sample label based on the second label selected for each sample to be processed;
and adjusting the model parameters of the label probability prediction model corresponding to each sample label based on the real labels respectively corresponding to part of samples to be processed and the real probabilities under the corresponding real labels, the second probabilities respectively corresponding to other processed samples except part of samples to be processed in the sample set to be processed, and the ratio of the second number corresponding to each sample label to the total number of the samples.
7. The apparatus of claim 1, wherein the determination module is configured to determine the portion of the sample to be processed according to the following steps:
for each sample to be processed, sequencing the initial probabilities corresponding to the samples to be processed according to the sequence of the probabilities from large to small;
determining the difference value between the first initial probability and the second initial probability in the initial probability sequence corresponding to the sample to be processed;
and determining the sample to be processed corresponding to the difference value smaller than the preset threshold value as part of the sample to be processed.
8. An information pushing method, characterized in that the method comprises:
acquiring a sample set to be processed;
generating a sample characteristic vector for each sample to be processed in a sample set to be processed, and predicting the initial probability of the sample to be processed under each sample label in a preset sample label set;
determining part of samples to be processed from the sample set to be processed based on the initial probability of each sample to be processed under each sample label in the sample label set;
and selecting a target label for each sample to be processed from the sample label set and pushing the target label based on the sample feature vector corresponding to each sample to be processed, the real labels corresponding to partial samples to be processed and the label probability prediction model corresponding to each sample label in the sample label set.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method as claimed in claim 8 are implemented when the processor executes the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method as set forth in claim 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911277050.3A CN111127179B (en) | 2019-12-12 | 2019-12-12 | Information pushing method, device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911277050.3A CN111127179B (en) | 2019-12-12 | 2019-12-12 | Information pushing method, device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111127179A true CN111127179A (en) | 2020-05-08 |
| CN111127179B CN111127179B (en) | 2023-08-29 |
Family
ID=70498563
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911277050.3A Active CN111127179B (en) | 2019-12-12 | 2019-12-12 | Information pushing method, device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111127179B (en) |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040145478A1 (en) * | 2001-02-08 | 2004-07-29 | Frederick Thomas J. | Differentially coherent combining for electronic article surveillance systems |
| JP2009211693A (en) * | 2008-02-29 | 2009-09-17 | Fujitsu Ltd | Pattern identification device and pattern identification method |
| US20130226758A1 (en) * | 2011-08-26 | 2013-08-29 | Reincloud Corporation | Delivering aggregated social media with third party apis |
| CN103279868A (en) * | 2013-05-22 | 2013-09-04 | 兰亭集势有限公司 | Method and device for automatically identifying fraud order form |
| US20140279777A1 (en) * | 2013-03-15 | 2014-09-18 | Google Inc. | Signal processing systems |
| EP3171297A1 (en) * | 2015-11-18 | 2017-05-24 | CentraleSupélec | Joint boundary detection image segmentation and object recognition using deep learning |
| CN107066449A (en) * | 2017-05-09 | 2017-08-18 | 北京京东尚科信息技术有限公司 | Information-pushing method and device |
| US20180204082A1 (en) * | 2017-01-17 | 2018-07-19 | International Business Machines Corporation | Ensemble based labeling |
| CN108763194A (en) * | 2018-04-27 | 2018-11-06 | 广州优视网络科技有限公司 | Using mark stamp methods, device, storage medium and computer equipment |
| AU2018101514A4 (en) * | 2018-10-11 | 2018-11-15 | Chi, Henan Mr | An automatic text-generating program for Chinese Hip-hop lyrics |
| CN109344314A (en) * | 2018-08-20 | 2019-02-15 | 腾讯科技（深圳）有限公司 | A kind of data processing method, device and server |
| CN109710842A (en) * | 2018-12-17 | 2019-05-03 | 泰康保险集团股份有限公司 | Method for pushing, device and the readable storage medium storing program for executing of business information |
| CN109872242A (en) * | 2019-01-30 | 2019-06-11 | 北京字节跳动网络技术有限公司 | Information-pushing method and device |
| CN110163301A (en) * | 2019-05-31 | 2019-08-23 | 北京金山云网络技术有限公司 | A kind of classification method and device of image |
| CN110223114A (en) * | 2019-05-29 | 2019-09-10 | 恩亿科(北京)数据科技有限公司 | The method and apparatus for managing the characteristic information of user |
| CN110472154A (en) * | 2019-08-26 | 2019-11-19 | 秒针信息技术有限公司 | A kind of resource supplying method, apparatus, electronic equipment and readable storage medium storing program for executing |
| CN110535974A (en) * | 2019-09-27 | 2019-12-03 | 恩亿科(北京)数据科技有限公司 | Method for pushing, driving means, equipment and the storage medium of resource to be put |
-
2019
- 2019-12-12 CN CN201911277050.3A patent/CN111127179B/en active Active
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040145478A1 (en) * | 2001-02-08 | 2004-07-29 | Frederick Thomas J. | Differentially coherent combining for electronic article surveillance systems |
| JP2009211693A (en) * | 2008-02-29 | 2009-09-17 | Fujitsu Ltd | Pattern identification device and pattern identification method |
| US20130226758A1 (en) * | 2011-08-26 | 2013-08-29 | Reincloud Corporation | Delivering aggregated social media with third party apis |
| US20140279777A1 (en) * | 2013-03-15 | 2014-09-18 | Google Inc. | Signal processing systems |
| CN103279868A (en) * | 2013-05-22 | 2013-09-04 | 兰亭集势有限公司 | Method and device for automatically identifying fraud order form |
| US20140351109A1 (en) * | 2013-05-22 | 2014-11-27 | Light In The Box Limited | Method and apparatus for automatically identifying a fraudulent order |
| EP3171297A1 (en) * | 2015-11-18 | 2017-05-24 | CentraleSupélec | Joint boundary detection image segmentation and object recognition using deep learning |
| US20180204082A1 (en) * | 2017-01-17 | 2018-07-19 | International Business Machines Corporation | Ensemble based labeling |
| CN107066449A (en) * | 2017-05-09 | 2017-08-18 | 北京京东尚科信息技术有限公司 | Information-pushing method and device |
| CN108763194A (en) * | 2018-04-27 | 2018-11-06 | 广州优视网络科技有限公司 | Using mark stamp methods, device, storage medium and computer equipment |
| CN109344314A (en) * | 2018-08-20 | 2019-02-15 | 腾讯科技（深圳）有限公司 | A kind of data processing method, device and server |
| AU2018101514A4 (en) * | 2018-10-11 | 2018-11-15 | Chi, Henan Mr | An automatic text-generating program for Chinese Hip-hop lyrics |
| CN109710842A (en) * | 2018-12-17 | 2019-05-03 | 泰康保险集团股份有限公司 | Method for pushing, device and the readable storage medium storing program for executing of business information |
| CN109872242A (en) * | 2019-01-30 | 2019-06-11 | 北京字节跳动网络技术有限公司 | Information-pushing method and device |
| CN110223114A (en) * | 2019-05-29 | 2019-09-10 | 恩亿科(北京)数据科技有限公司 | The method and apparatus for managing the characteristic information of user |
| CN110163301A (en) * | 2019-05-31 | 2019-08-23 | 北京金山云网络技术有限公司 | A kind of classification method and device of image |
| CN110472154A (en) * | 2019-08-26 | 2019-11-19 | 秒针信息技术有限公司 | A kind of resource supplying method, apparatus, electronic equipment and readable storage medium storing program for executing |
| CN110535974A (en) * | 2019-09-27 | 2019-12-03 | 恩亿科(北京)数据科技有限公司 | Method for pushing, driving means, equipment and the storage medium of resource to be put |
Non-Patent Citations (5)
| Title |
|---|
| 姚明海;林宣民;王宪保;: "基于局部敏感Hash的半监督直推SVM增量学习算法", 浙江工业大学学报, no. 02 * |
| 张俊丽等: "标签传播算法理论及其应用研究综述", 《计算机应用研究》 * |
| 张俊丽等: "标签传播算法理论及其应用研究综述", 《计算机应用研究》, no. 01, 15 January 2013 (2013-01-15) * |
| 李红梅等: "面向隐式反馈的标签感知推荐方法", 《计算机科学》 * |
| 李红梅等: "面向隐式反馈的标签感知推荐方法", 《计算机科学》, no. 04, 15 April 2019 (2019-04-15) * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111127179B (en) | 2023-08-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108804641B (en) | Text similarity calculation method, device, equipment and storage medium | |
| CN110909165B (en) | Data processing method, device, medium and electronic equipment | |
| CN108446741B (en) | Method, system and storage medium for evaluating importance of machine learning hyper-parameter | |
| US11915104B2 (en) | Normalizing text attributes for machine learning models | |
| WO2017193121A1 (en) | Agent aptitude prediction | |
| CN111783993A (en) | Intelligent labeling method and device, intelligent platform and storage medium | |
| US8832015B2 (en) | Fast binary rule extraction for large scale text data | |
| CN111178537B (en) | Feature extraction model training method and device | |
| CN114492363B (en) | Small sample fine adjustment method, system and related device | |
| CN111127364A (en) | Image data enhancement strategy selection method and face recognition image data enhancement method | |
| CN115688760B (en) | Intelligent diagnosis guiding method, device, equipment and storage medium | |
| CN112148986B (en) | Top-N service re-recommendation method and system based on crowdsourcing | |
| CN113537630A (en) | Training method and device of business prediction model | |
| Swishchuk et al. | General semi-Markov model for limit order books | |
| CN113159213A (en) | Service distribution method, device and equipment | |
| US20100296728A1 (en) | Discrimination Apparatus, Method of Discrimination, and Computer Program | |
| CN111179055A (en) | Credit limit adjusting method and device and electronic equipment | |
| CN111127179A (en) | Information pushing method and device, computer equipment and storage medium | |
| CN115936184A (en) | Load prediction matching method suitable for multi-user types | |
| CN112200488B (en) | Risk identification model training method and device for business object | |
| CN115547514A (en) | Pathogenic gene sequencing method, pathogenic gene sequencing device, electronic equipment and medium | |
| CN111191049B (en) | Information pushing method and device, computer equipment and storage medium | |
| CN113761875A (en) | Event extraction method and device, electronic equipment and storage medium | |
| CN112906785A (en) | Zero-sample object type identification method, device and equipment based on fusion | |
| CN113011689A (en) | Software development workload assessment method and device and computing equipment |
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 |