CN113591881B - Intention recognition method and device based on model fusion, electronic equipment and medium - Google Patents
Intention recognition method and device based on model fusion, electronic equipment and medium Download PDFInfo
- Publication number
- CN113591881B CN113591881B CN202110913907.7A CN202110913907A CN113591881B CN 113591881 B CN113591881 B CN 113591881B CN 202110913907 A CN202110913907 A CN 202110913907A CN 113591881 B CN113591881 B CN 113591881B
- Authority
- CN
- China
- Prior art keywords
- standard
- model
- fusion
- scene
- data set
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/24323—Tree-organised classifiers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Abstract
The invention relates to an intelligent decision technology, and discloses an intention recognition method based on model fusion, which comprises the following steps: and distributing the standard feature set to a single-side user and a multi-side user, carrying out missing value filling and abnormal value processing on the distributed data to obtain standard single-side data and standard multi-side data, respectively constructing and testing according to the standard single-side data and the standard multi-side data to obtain a standard single-scene model and a standard multi-scene model, respectively carrying out model fusion on a preset number of standard single-scene models and standard multi-scene models to obtain a single-scene fusion model and a multi-scene fusion model, and calling a corresponding model according to the user category of the data to be identified to carry out intention identification to obtain an intention identification result. Furthermore, the present invention relates to blockchain techniques, the set of timing characteristics may be stored at nodes of the blockchain. The invention also provides an intention recognition device based on model fusion, electronic equipment and a computer readable storage medium. The invention can solve the problem of low accuracy of intention recognition.
Description
Technical Field
The invention relates to the technical field of intelligent decision making, in particular to an intention recognition method, an intention recognition device, electronic equipment and a computer readable storage medium based on model fusion.
Background
At the moment of rapid development of information technology, digitization and science and technology are widely used in the financial field, wherein business oriented to terminal retail is more obvious in digital science and technology transformation, and consumers need more flexible and convenient services. The subsequent feedback and intent of the consumer is often the focus of attention, so that willingness recognition based on the behavior data of the consumer is needed, and the recognition result is helpful to find potential products of interest to more consumers, and better improvement is performed on the service.
The existing intention recognition method generally utilizes a single model to perform independent recognition, and does not consider other scenes and dimensions, so that the accuracy of intention recognition is low.
Disclosure of Invention
The invention provides an intention recognition method and device based on model fusion and a computer readable storage medium, and mainly aims to solve the problem of low accuracy of intention recognition.
In order to achieve the above object, the present invention provides an intention recognition method based on model fusion, including:
Acquiring a service data set of a user, and extracting features of the service data set to obtain a service feature set;
performing time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and summarizing the time sequence feature set and the service feature set to obtain a standard feature set;
classifying the users to obtain single-side users and multi-side users, and distributing the standard feature sets to the single-side users and the multi-side users to obtain single-side data sets corresponding to the single-side users and multi-side data sets corresponding to the multi-side users;
performing missing value filling and abnormal value processing on the unilateral data set and the multilateral data set to obtain a standard unilateral data set and a standard multilateral data set;
dividing the standard unilateral data set into a training data set and a test data set, constructing a single scene model based on the training data set, performing test processing on the single scene model by utilizing the test data set, and outputting a standard single scene model according to an obtained first test result;
constructing a multi-scene model according to the standard polygonal data set, testing the multi-scene model, and outputting a standard multi-scene model according to an obtained second test result;
Carrying out model fusion on a preset number of standard single scene models to obtain single scene fusion models, and carrying out model fusion on a preset number of standard multi-scene models to obtain multi-scene fusion models, wherein the single scene fusion models and the multi-scene fusion models both comprise fusion formulas;
and acquiring data to be identified, identifying a user category corresponding to the data to be identified, and calling the single scene fusion model or the multi-scene fusion model according to the user category to perform intention identification processing on the data to be identified to obtain an intention identification result.
Optionally, the constructing a single scene model based on the training dataset includes:
constructing an initial decision tree by using the standard feature set, and performing decision tree adding processing on the initial decision tree to obtain an updated decision tree;
inputting the training data set into the updating decision tree to obtain a predicted value set, and calculating a loss value of the updating decision tree according to the predicted value set and a preset kitchen loss function;
and when the loss value is greater than or equal to a preset loss threshold value, executing the operation of adding the decision tree to the initial decision tree again until the loss value is smaller than the loss threshold value, and outputting the current updated decision tree as a single scene model.
Optionally, the constructing an initial decision tree using the standard feature set includes:
classifying and labeling the standard feature set to obtain labels corresponding to the standard feature set;
selecting one label at will as a segmentation point, and taking the segmentation point as a root node of an original decision tree;
generating the child nodes of the cut points and distributing the standard feature set to the child nodes to obtain an initial decision tree.
Optionally, the preset cookability loss function is:
FL(p t )＝-α t (1-p t ) γ log(p t )
wherein FL (p) t ) For the loss value, p, of the updated decision tree t Alpha is the predicted value in the predicted value set t And gamma is a preset second weight for the preset first weight.
Optionally, the performing missing value filling and outlier processing on the unilateral data set to obtain a standard unilateral data set includes:
detecting the unilateral data set according to a pre-constructed missing detection statement, and filling the missing part by using a preset filling value;
sorting the unilateral data sets according to the sequence from big to small, and screening the medians positioned at the middle positions from the sorted unilateral data sets;
respectively calculating the similarity between the median and a plurality of single-side data in the single-side data set;
Judging the single-side data with the similarity larger than a preset abnormal threshold value as an abnormal value, and deleting the abnormal value to obtain a standard single-side data set.
Optionally, the performing time sequence calculation on the service data in the service data set according to the service feature set to obtain a time sequence feature set includes:
extracting service data corresponding to the service feature set in the service data set;
and calculating time sequence characteristics of the service data corresponding to the service characteristic set according to a preset time domain characteristic calculation formula to obtain the time sequence characteristic set.
Optionally, the feature extraction of the service data set to obtain a service feature set includes:
acquiring a plurality of preset feature dimensions and feature word libraries corresponding to the feature dimensions;
and screening the data corresponding to the feature words in the feature word library in the service data set to obtain a service feature set.
In order to solve the above problems, the present invention further provides an intention recognition device based on model fusion, the device comprising:
the feature extraction module is used for acquiring a service data set of a user, extracting features of the service data set to obtain a service feature set, performing time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and summarizing the time sequence feature set and the service feature set to obtain a standard feature set;
The data processing module is used for classifying the users to obtain single-side users and multi-side users, distributing the standard feature sets to the single-side users and the multi-side users to obtain single-side data sets corresponding to the single-side users and multi-side data sets corresponding to the multi-side users, and carrying out missing value filling and abnormal value processing on the single-side data sets and the multi-side data sets to obtain standard single-side data sets and standard multi-side data sets;
the standard single scene model construction module is used for dividing the standard single-side data set into a training data set and a test data set, constructing a single scene model based on the training data set, carrying out test processing on the single scene model by utilizing the test data set, and outputting a standard single scene model according to an obtained first test result;
the standard multi-scene model construction module is used for constructing a multi-scene model according to the standard polygonal data set, testing the multi-scene model and outputting a standard multi-scene model according to the obtained second test result;
the model fusion module is used for carrying out model fusion on a preset number of standard single scene models to obtain single scene fusion models, carrying out model fusion on a preset number of standard multi-scene models to obtain multi-scene fusion models, wherein the single scene fusion models and the multi-scene fusion models both comprise fusion formulas;
The intention recognition module is used for acquiring data to be recognized, recognizing a user category corresponding to the data to be recognized, and calling the single scene fusion model or the multi-scene fusion model according to the user category to perform intention recognition processing on the data to be recognized to obtain an intention recognition result.
In order to solve the above-mentioned problems, the present invention also provides an electronic apparatus including:
a memory storing at least one instruction; a kind of electronic device with high-pressure air-conditioning system
And the processor executes the instructions stored in the memory to realize the intention recognition method based on the model fusion.
In order to solve the above-mentioned problems, the present invention also provides a computer-readable storage medium having stored therein at least one instruction that is executed by a processor in an electronic device to implement the above-mentioned model fusion-based intent recognition method.
In the embodiment of the invention, the service feature set is obtained by carrying out feature extraction on the service data set, wherein the service feature set comprises data with a plurality of feature dimensions, the service data in the service data set is subjected to time sequence calculation according to the service feature set to obtain the time sequence feature set, the floating change of the service data and the characteristics in time domain can be reflected through the time sequence feature set, so that the preference of the service data can be predicted more accurately, the unilateral data set and the polygonal data set are subjected to missing value filling and outlier processing, the accuracy of the standard unilateral data set and the standard polygonal data set obtained through processing is ensured, the data redundancy is avoided, the single scene model is constructed based on the training data set, the test data set is utilized to carry out test processing on the single scene model, the standard single scene model with a preset number is subjected to model fusion according to the obtained test result, and the single scene fusion model is obtained. The multi-scene fusion model is obtained by carrying out model fusion on a preset number of standard multi-scene models, and the multi-user scene is considered, so that the recognition range is widened. In the embodiment of the invention, the user category corresponding to the data to be identified is identified, the data to be identified is subjected to the intention identification processing according to the user category single scene fusion model or the multi-scene fusion model, the intention identification result is obtained, and the accuracy of the intention identification is improved. Therefore, the intention recognition method, the intention recognition device, the electronic equipment and the computer readable storage medium based on the model fusion can solve the problem of low accuracy of intention recognition.
Drawings
FIG. 1 is a flow chart of a method for identifying intent based on model fusion according to an embodiment of the present invention;
FIG. 2 is a functional block diagram of an intent recognition device based on model fusion according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of an electronic device implementing the intent recognition method based on model fusion according to an embodiment of the present invention.
The achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.
Detailed Description
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 invention.
The embodiment of the application provides an intention recognition method based on model fusion. The execution subject of the intent recognition method based on model fusion includes, but is not limited to, at least one of a server, a terminal, and the like, which can be configured to execute the method provided by the embodiment of the application. In other words, the model fusion-based intention recognition method may be performed by software or hardware installed in a terminal device or a server device, and the software may be a blockchain platform. The service end includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, and the like.
Referring to fig. 1, a flow chart of a method for identifying intent based on model fusion according to an embodiment of the present invention is shown. In this embodiment, the method for identifying intent based on model fusion includes:
s1, acquiring a service data set of a user, and extracting features of the service data set to obtain a service feature set.
In the embodiment of the present invention, the service data set of the user refers to related service data generated around a service target in a service scene, for example, in the financial field, the service data set of the user is user data of a group company and user data of a family of an automobile.
Specifically, the feature extraction of the service data set to obtain a service feature set includes:
acquiring a plurality of preset feature dimensions and feature word libraries corresponding to the feature dimensions;
and screening the data corresponding to the feature words in the feature word library in the service data set to obtain a service feature set.
In detail, in the embodiment of the present invention, the feature dimensions refer to user information, product preferences, login behavior, and the like, and the feature word libraries corresponding to the feature dimensions refer to keywords related to the feature dimensions, for example, when the feature dimensions are product preferences, the corresponding feature word libraries include words such as "preferred", "buyback", and the like.
S2, performing time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and summarizing the time sequence feature set and the service feature set to obtain a standard feature set.
In the embodiment of the present invention, the performing time sequence calculation on the service data in the service data set according to the service feature set to obtain a time sequence feature set includes:
extracting service data corresponding to the service feature set in the service data set;
and calculating time sequence characteristics of the service data corresponding to the service characteristic set according to a preset time domain characteristic calculation formula to obtain the time sequence characteristic set.
In detail, the time sequence features corresponding to the service data are calculated according to a preset time domain feature calculation formula, and the time sequence features comprise, but are not limited to, mean difference, variance, cosine similarity and the like, and formulas and methods for calculating the mean difference, the variance, the cosine similarity and the like are different, so that the corresponding formulas are selected according to the time sequence features to be calculated for calculation.
For example, when the timing characteristic is variance, the time domain characteristic calculation formula is:
wherein S is a time sequence feature, M is an average value of a service data set, and x 1 、x 2 …x n And n is the number of the service data in the service data set.
And S3, classifying the users to obtain single-side users and multi-side users, and distributing the standard feature sets to the single-side users and the multi-side users to obtain single-side data sets corresponding to the single-side users and multi-side data sets corresponding to the multi-side users.
In the embodiment of the invention, different users have corresponding scenes, the users are divided into single-side users and multi-side users, and corresponding data sets are identified in the standard feature set according to different users, so that the single-side data sets corresponding to the single-side users and the multi-side data sets corresponding to the multi-side users are obtained.
Wherein the single-sided user refers to a user belonging to one single category, the multi-sided user refers to a user belonging to a plurality of categories, for example, the single-sided user may exist in company a and also exist in group B, in the embodiment of the present invention, the single-sided user may refer to a group user, and the multi-sided user refers to multiple users of banks and groups.
S4, carrying out missing value filling and abnormal value processing on the unilateral data set and the polygonal data set to obtain a standard unilateral data set and a standard polygonal data set.
In the embodiment of the present invention, the step of performing missing value filling and outlier processing on the single-side data set to obtain a standard single-side data set includes:
detecting the unilateral data set according to a pre-constructed missing detection statement, and filling the missing part by using a preset filling value;
sorting the unilateral data sets according to the sequence from big to small, and screening the medians positioned at the middle positions from the sorted unilateral data sets;
respectively calculating the similarity between the median and a plurality of single-side data in the single-side data set;
judging the single-side data with the similarity larger than a preset abnormal threshold value as an abnormal value, and deleting the abnormal value to obtain a standard single-side data set.
The pre-constructed deletion detection statement may be a java statement, and the preset filling value may be user data in a preset user database.
Specifically, the calculating the similarity between the median and the plurality of single-side data in the single-side data set respectively includes:
wherein J (α, β) is a similarity between the median and the single-side data, α is the median, β is the single-side data, |α| is a modular length of the median, |β| is a modular length of the single-side data, and|α n β| is a modular length of an intersection of the median and the single-side data.
In detail, the process of performing missing value filling and outlier processing on the polygon data set is consistent with the process of performing missing value filling and outlier processing on the single-side data set, and will not be described herein.
S5, dividing the standard unilateral data set into a training data set and a test data set, constructing a single scene model based on the training data set, testing the single scene model by using the test data set, and outputting a standard single scene model according to the obtained first test result.
In the embodiment of the invention, the standard unilateral data set is divided into the training data set and the testing data set by using a preset dividing ratio.
Preferably, the preset dividing ratio may be 7:3, that is, the ratio of the standard unilateral data set is divided into a training data set and a testing data set. The training can be performed by using more data as much as possible, so that the training is more accurate, meanwhile, a part of data is used for testing, the model obtained through training is further optimized, and the accuracy of the model obtained through training is improved.
Wherein the training data set is a data sample for model fitting and the test data set is used to evaluate the generalization ability of the model final model.
Specifically, the constructing a single scene model based on the training data set includes:
constructing an initial decision tree by using the standard feature set, and performing decision tree adding processing on the initial decision tree to obtain an updated decision tree;
inputting the training data set into the updating decision tree to obtain a predicted value set, and calculating a loss value of the updating decision tree according to the predicted value set and a preset kitchen loss function;
and when the loss value is greater than or equal to a preset loss threshold value, executing the operation of adding the decision tree to the initial decision tree again until the loss value is smaller than the loss threshold value, and outputting the current updated decision tree as a single scene model.
In detail, in the embodiment of the invention, the single scene model can be an XGboost model, the XGboost model can explain complex multidimensional relations, the prediction capability is strong, and the model can achieve good prediction performance results on training data.
Further, the constructing an initial decision tree by using the standard feature set includes:
classifying and labeling the standard feature set to obtain labels corresponding to the standard feature set;
Selecting one label at will as a segmentation point, and taking the segmentation point as a root node of an original decision tree;
generating the child nodes of the cut points and distributing the standard feature set to the child nodes to obtain an initial decision tree.
Specifically, the step of adding the decision tree to the initial decision tree to obtain an updated decision tree includes:
acquiring preset adding times;
and splitting the root node of the initial decision tree for corresponding times based on the adding times, and distributing the standard feature set to the split root node to obtain an updated decision tree.
In detail, the training data set is input into the updated decision tree, resulting in a set of predictors, i.e. a summary of the predictors at each leaf node of the updated decision tree into which the training data set is input.
Further, the preset cookability loss function is as follows:
FL(p t )＝-α t (1-p t ) γ log(p t )
wherein FL (p) t ) For the loss value, p, of the updated decision tree t Alpha is the predicted value in the predicted value set t And gamma is a preset second weight for the preset first weight.
Specifically, the magnitude between the loss value and a preset loss threshold is judged, when the loss value is larger than or equal to the preset loss threshold, the operation of adding the decision tree to the initial decision tree is executed again until the loss value is smaller than the loss threshold, and the current updated decision tree is output as a single scene model.
Further, the test data set is utilized to carry out test processing on the single scene model, the test data set is input into the single scene model, a corresponding prediction score can be obtained, an error value between the prediction score and a preset true score is calculated, the size between the error value and a preset error threshold is judged, when the error value is larger than the error threshold, model parameters of the single scene model are adjusted, and the test data set is input into the single scene model after the model parameters are adjusted again until the error value is smaller than the error threshold, and a standard single scene model is output.
S6, constructing a multi-scene model according to the standard polygonal data set, testing the multi-scene model, and outputting the standard multi-scene model according to the obtained second test result.
In the embodiment of the present invention, the process of constructing and testing the standard polygonal scene model according to the standard polygonal data set is the same as the process of constructing and testing the standard unilateral scene model according to the standard unilateral data set, and will not be described herein.
S7, carrying out model fusion on the standard single scene models with the preset number to obtain single scene fusion models, and carrying out model fusion on the standard multi-scene models with the preset number to obtain multi-scene fusion models, wherein the single scene fusion models and the multi-scene fusion models both comprise fusion formulas.
In the embodiment of the invention, the single scene fusion model is formed by sharing an embedded layer by a plurality of sub-models, and takes a fusion formula as an output layer.
In detail, the fusion formula in the fusion model can calculate the fusion value of a plurality of standard single scene models fused together, the fusion formula is used for carrying out model fusion, and the fusion value calculated according to the fusion formula in the fusion model can be compared with a preset threshold value, so that an intention recognition result is obtained.
For example, in the embodiment of the present invention, if four standard single scene models are obtained, the fusion formula is:
wherein, psi is the fusion value, alpha 1 、α 2 、α 3 、α 4 Respectively preset weights corresponding to four standard single scene models, p 1 、p 2 、p 3 、p 4 The scores corresponding to the four standard single scene models are respectively obtained.
Specifically, the model fusion process of the standard multi-scene model is the same as that of the standard single-scene model, and will not be described herein.
S8, acquiring data to be identified, identifying a user category corresponding to the data to be identified, and calling the single scene fusion model or the multi-scene fusion model according to the user category to perform intention identification processing on the data to be identified to obtain an intention identification result.
In the embodiment of the invention, the user corresponding to the data to be identified is identified as a single user or multiple users, if the user is a single user, a single scene fusion model is called to process the data to be identified, a willingness degree value is obtained, the size between the willingness degree value and a preset threshold value is judged, and then the willingness of the user is judged.
For example, when the willingness value is smaller than a preset threshold, the intention recognition result of the corresponding user is judged to be unwilling, and when the willingness value is larger than or equal to the preset threshold, the intention recognition result of the corresponding user is judged to be willing.
Similarly, if the user is a multi-user, a multi-scene fusion model is called, intent recognition is carried out on the multi-user by utilizing the multi-scene fusion model, a willingness value corresponding to the multi-user is obtained, the magnitude between the willingness value corresponding to the multi-user and the preset threshold is judged, when the willingness value corresponding to the multi-user is smaller than the preset threshold, the intent recognition result of the multi-user is judged to be unwilling, and when the willingness value is larger than or equal to the preset threshold, the intent recognition result of the multi-user is judged to be willing.
In the embodiment of the invention, the service feature set is obtained by carrying out feature extraction on the service data set, wherein the service feature set comprises data with a plurality of feature dimensions, the service data in the service data set is subjected to time sequence calculation according to the service feature set to obtain the time sequence feature set, the floating change of the service data and the characteristics in time domain can be reflected through the time sequence feature set, so that the preference of the service data can be predicted more accurately, the unilateral data set and the polygonal data set are subjected to missing value filling and outlier processing, the accuracy of the standard unilateral data set and the standard polygonal data set obtained through processing is ensured, the data redundancy is avoided, the single scene model is constructed based on the training data set, the test data set is utilized to carry out test processing on the single scene model, the standard single scene model with a preset number is subjected to model fusion according to the obtained test result, and the single scene fusion model is obtained. The multi-scene fusion model is obtained by carrying out model fusion on a preset number of standard multi-scene models, and the multi-user scene is considered, so that the recognition range is widened. In the embodiment of the invention, the user category corresponding to the data to be identified is identified, the data to be identified is subjected to the intention identification processing according to the user category single scene fusion model or the multi-scene fusion model, the intention identification result is obtained, and the accuracy of the intention identification is improved. Therefore, the intention recognition method based on model fusion can solve the problem of low accuracy of intention recognition.
Fig. 2 is a functional block diagram of an intent recognition device based on model fusion according to an embodiment of the present invention.
The intent recognition device 100 based on model fusion according to the present invention may be installed in an electronic apparatus. Depending on the functions implemented, the model fusion based intent recognition device 100 may include a feature extraction module 101, a data processing module 102, a standard single scene model construction module 103, a standard multiple scene model construction module 104, a model fusion module 105, and an intent recognition module 106. The module of the invention, which may also be referred to as a unit, refers to a series of computer program segments, which are stored in the memory of the electronic device, capable of being executed by the processor of the electronic device and of performing a fixed function.
In the present embodiment, the functions concerning the respective modules/units are as follows:
the feature extraction module 101 is configured to obtain a service data set of a user, perform feature extraction on the service data set to obtain a service feature set, perform time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and aggregate the time sequence feature set and the service feature set to obtain a standard feature set;
The data processing module 102 is configured to classify the users to obtain a single-side user and a multi-side user, allocate the standard feature set to the single-side user and the multi-side user to obtain a single-side data set corresponding to the single-side user and a multi-side data set corresponding to the multi-side user, and perform missing value filling and outlier processing on the single-side data set and the multi-side data set to obtain a standard single-side data set and a standard multi-side data set;
the standard single scene model construction module 103 is configured to divide the standard single-side dataset into a training dataset and a test dataset, construct a single scene model based on the training dataset, perform test processing on the single scene model by using the test dataset, and output a standard single scene model according to an obtained first test result;
the standard multi-scene model construction module 104 is configured to construct a multi-scene model according to the standard polygonal data set, perform test processing on the multi-scene model, and output a standard multi-scene model according to an obtained second test result;
the model fusion module 105 is configured to perform model fusion on a preset number of standard single scene models to obtain single scene fusion models, and perform model fusion on a preset number of standard multi-scene models to obtain multi-scene fusion models, where the single scene fusion models and the multi-scene fusion models both include fusion formulas;
The intention recognition module 106 is configured to obtain data to be recognized, recognize a user category corresponding to the data to be recognized, and invoke the single scene fusion model or the multi-scene fusion model according to the user category to perform intention recognition processing on the data to be recognized, so as to obtain an intention recognition result.
In detail, the specific implementation manner of each module of the intent recognition device 100 based on model fusion is as follows:
step one, acquiring a service data set of a user, and extracting features of the service data set to obtain a service feature set.
In the embodiment of the present invention, the service data set of the user refers to related service data generated around a service target in a service scene, for example, in the financial field, the service data set of the user is user data of a group company and user data of a family of an automobile.
Specifically, the feature extraction of the service data set to obtain a service feature set includes:
acquiring a plurality of preset feature dimensions and feature word libraries corresponding to the feature dimensions;
and screening the data corresponding to the feature words in the feature word library in the service data set to obtain a service feature set.
In detail, in the embodiment of the present invention, the feature dimensions refer to user information, product preferences, login behavior, and the like, and the feature word libraries corresponding to the feature dimensions refer to keywords related to the feature dimensions, for example, when the feature dimensions are product preferences, the corresponding feature word libraries include words such as "preferred", "buyback", and the like.
And step two, carrying out time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and summarizing the time sequence feature set and the service feature set to obtain a standard feature set.
In the embodiment of the present invention, the performing time sequence calculation on the service data in the service data set according to the service feature set to obtain a time sequence feature set includes:
extracting service data corresponding to the service feature set in the service data set;
and calculating time sequence characteristics of the service data corresponding to the service characteristic set according to a preset time domain characteristic calculation formula to obtain the time sequence characteristic set.
In detail, the time sequence features corresponding to the service data are calculated according to a preset time domain feature calculation formula, and the time sequence features comprise, but are not limited to, mean difference, variance, cosine similarity and the like, and formulas and methods for calculating the mean difference, the variance, the cosine similarity and the like are different, so that the corresponding formulas are selected according to the time sequence features to be calculated for calculation.
For example, when the timing characteristic is variance, the time domain characteristic calculation formula is:
wherein S is a time sequence feature, M is an average value of a service data set, and x 1 、x 2 …x n And n is the number of the service data in the service data set.
And thirdly, classifying the users to obtain single-side users and multi-side users, and distributing the standard feature sets to the single-side users and the multi-side users to obtain single-side data sets corresponding to the single-side users and multi-side data sets corresponding to the multi-side users.
In the embodiment of the invention, different users have corresponding scenes, the users are divided into single-side users and multi-side users, and corresponding data sets are identified in the standard feature set according to different users, so that the single-side data sets corresponding to the single-side users and the multi-side data sets corresponding to the multi-side users are obtained.
Wherein the single-sided user refers to a user belonging to one single category, the multi-sided user refers to a user belonging to a plurality of categories, for example, the single-sided user may exist in company a and also exist in group B, in the embodiment of the present invention, the single-sided user may refer to a group user, and the multi-sided user refers to multiple users of banks and groups.
And fourthly, performing missing value filling and abnormal value processing on the unilateral data set and the polygonal data set to obtain a standard unilateral data set and a standard polygonal data set.
In the embodiment of the present invention, the step of performing missing value filling and outlier processing on the single-side data set to obtain a standard single-side data set includes:
detecting the unilateral data set according to a pre-constructed missing detection statement, and filling the missing part by using a preset filling value;
sorting the unilateral data sets according to the sequence from big to small, and screening the medians positioned at the middle positions from the sorted unilateral data sets;
respectively calculating the similarity between the median and a plurality of single-side data in the single-side data set;
judging the single-side data with the similarity larger than a preset abnormal threshold value as an abnormal value, and deleting the abnormal value to obtain a standard single-side data set.
The pre-constructed deletion detection statement may be a java statement, and the preset filling value may be user data in a preset user database.
Specifically, the calculating the similarity between the median and the plurality of single-side data in the single-side data set respectively includes:
Wherein J (α, β) is a similarity between the median and the single-side data, α is the median, β is the single-side data, |α| is a modular length of the median, |β| is a modular length of the single-side data, and|α n β| is a modular length of an intersection of the median and the single-side data.
In detail, the process of performing missing value filling and outlier processing on the polygon data set is consistent with the process of performing missing value filling and outlier processing on the single-side data set, and will not be described herein.
And fifthly, dividing the standard unilateral data set into a training data set and a testing data set, constructing a single scene model based on the training data set, testing the single scene model by utilizing the testing data set, and outputting a standard single scene model according to the obtained first testing result.
In the embodiment of the invention, the standard unilateral data set is divided into the training data set and the testing data set by using a preset dividing ratio.
Preferably, the preset dividing ratio may be 7:3, that is, the ratio of the standard unilateral data set is divided into a training data set and a testing data set. The training can be performed by using more data as much as possible, so that the training is more accurate, meanwhile, a part of data is used for testing, the model obtained through training is further optimized, and the accuracy of the model obtained through training is improved.
Wherein the training data set is a data sample for model fitting and the test data set is used to evaluate the generalization ability of the model final model.
Specifically, the constructing a single scene model based on the training data set includes:
constructing an initial decision tree by using the standard feature set, and performing decision tree adding processing on the initial decision tree to obtain an updated decision tree;
inputting the training data set into the updating decision tree to obtain a predicted value set, and calculating a loss value of the updating decision tree according to the predicted value set and a preset kitchen loss function;
and when the loss value is greater than or equal to a preset loss threshold value, executing the operation of adding the decision tree to the initial decision tree again until the loss value is smaller than the loss threshold value, and outputting the current updated decision tree as a single scene model.
In detail, in the embodiment of the invention, the single scene model can be an XGboost model, the XGboost model can explain complex multidimensional relations, the prediction capability is strong, and the model can achieve good prediction performance results on training data.
Further, the constructing an initial decision tree by using the standard feature set includes:
Classifying and labeling the standard feature set to obtain labels corresponding to the standard feature set;
selecting one label at will as a segmentation point, and taking the segmentation point as a root node of an original decision tree;
generating the child nodes of the cut points and distributing the standard feature set to the child nodes to obtain an initial decision tree.
Specifically, the step of adding the decision tree to the initial decision tree to obtain an updated decision tree includes:
acquiring preset adding times;
and splitting the root node of the initial decision tree for corresponding times based on the adding times, and distributing the standard feature set to the split root node to obtain an updated decision tree.
In detail, the training data set is input into the updated decision tree, resulting in a set of predictors, i.e. a summary of the predictors at each leaf node of the updated decision tree into which the training data set is input.
Further, the preset cookability loss function is as follows:
FL(p t )＝-α t (1-p t ) γ log(p t )
wherein FL (p) t ) For the loss value, p, of the updated decision tree t Alpha is the predicted value in the predicted value set t And gamma is a preset second weight for the preset first weight.
Specifically, the magnitude between the loss value and a preset loss threshold is judged, when the loss value is larger than or equal to the preset loss threshold, the operation of adding the decision tree to the initial decision tree is executed again until the loss value is smaller than the loss threshold, and the current updated decision tree is output as a single scene model.
Further, the test data set is utilized to carry out test processing on the single scene model, the test data set is input into the single scene model, a corresponding prediction score can be obtained, an error value between the prediction score and a preset true score is calculated, the size between the error value and a preset error threshold is judged, when the error value is larger than the error threshold, model parameters of the single scene model are adjusted, and the test data set is input into the single scene model after the model parameters are adjusted again until the error value is smaller than the error threshold, and a standard single scene model is output.
And step six, constructing a multi-scene model according to the standard polygonal data set, testing the multi-scene model, and outputting the standard multi-scene model according to the obtained second test result.
In the embodiment of the present invention, the process of constructing and testing the standard polygonal scene model according to the standard polygonal data set is the same as the process of constructing and testing the standard unilateral scene model according to the standard unilateral data set, and will not be described herein.
And seventhly, carrying out model fusion on the standard single scene models with the preset number to obtain single scene fusion models, and carrying out model fusion on the standard multi-scene models with the preset number to obtain multi-scene fusion models, wherein the single scene fusion models and the multi-scene fusion models both comprise fusion formulas.
In the embodiment of the invention, the single scene fusion model is formed by sharing an embedded layer by a plurality of sub-models, and takes a fusion formula as an output layer.
In detail, the fusion formula in the fusion model can calculate the fusion value of a plurality of standard single scene models fused together, the fusion formula is used for carrying out model fusion, and the fusion value calculated according to the fusion formula in the fusion model can be compared with a preset threshold value, so that an intention recognition result is obtained.
For example, in the embodiment of the present invention, if four standard single scene models are obtained, the fusion formula is:
Wherein, psi is the fusion value, alpha 1 、α 2 、α 3 、α 4 Respectively preset weights corresponding to four standard single scene models, p 1 、p 2 、p 3 、p 4 The scores corresponding to the four standard single scene models are respectively obtained.
Specifically, the model fusion process of the standard multi-scene model is the same as that of the standard single-scene model, and will not be described herein.
And step eight, acquiring data to be identified, identifying a user category corresponding to the data to be identified, and calling the single scene fusion model or the multi-scene fusion model according to the user category to perform intention identification processing on the data to be identified to obtain an intention identification result.
In the embodiment of the invention, the user corresponding to the data to be identified is identified as a single user or multiple users, if the user is a single user, a single scene fusion model is called to process the data to be identified, a willingness degree value is obtained, the size between the willingness degree value and a preset threshold value is judged, and then the willingness of the user is judged.
For example, when the willingness value is smaller than a preset threshold, the intention recognition result of the corresponding user is judged to be unwilling, and when the willingness value is larger than or equal to the preset threshold, the intention recognition result of the corresponding user is judged to be willing.
Similarly, if the user is a multi-user, a multi-scene fusion model is called, intent recognition is carried out on the multi-user by utilizing the multi-scene fusion model, a willingness value corresponding to the multi-user is obtained, the magnitude between the willingness value corresponding to the multi-user and the preset threshold is judged, when the willingness value corresponding to the multi-user is smaller than the preset threshold, the intent recognition result of the multi-user is judged to be unwilling, and when the willingness value is larger than or equal to the preset threshold, the intent recognition result of the multi-user is judged to be willing.
In the embodiment of the invention, the service feature set is obtained by carrying out feature extraction on the service data set, wherein the service feature set comprises data with a plurality of feature dimensions, the service data in the service data set is subjected to time sequence calculation according to the service feature set to obtain the time sequence feature set, the floating change of the service data and the characteristics in time domain can be reflected through the time sequence feature set, so that the preference of the service data can be predicted more accurately, the unilateral data set and the polygonal data set are subjected to missing value filling and outlier processing, the accuracy of the standard unilateral data set and the standard polygonal data set obtained through processing is ensured, the data redundancy is avoided, the single scene model is constructed based on the training data set, the test data set is utilized to carry out test processing on the single scene model, the standard single scene model with a preset number is subjected to model fusion according to the obtained test result, and the single scene fusion model is obtained. The multi-scene fusion model is obtained by carrying out model fusion on a preset number of standard multi-scene models, and the multi-user scene is considered, so that the recognition range is widened. In the embodiment of the invention, the user category corresponding to the data to be identified is identified, the data to be identified is subjected to the intention identification processing according to the user category single scene fusion model or the multi-scene fusion model, the intention identification result is obtained, and the accuracy of the intention identification is improved. Therefore, the intention recognition device based on model fusion can solve the problem of low accuracy of intention recognition.
Fig. 3 is a schematic structural diagram of an electronic device for implementing a model fusion-based intent recognition method according to an embodiment of the present invention.
The electronic device may comprise a processor 10, a memory 11, a communication interface 12 and a bus 13, and may further comprise a computer program stored in the memory 11 and executable on the processor 10, such as an intent recognition program based on model fusion.
The memory 11 includes at least one type of readable storage medium, including flash memory, a mobile hard disk, a multimedia card, a card memory (e.g., SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc. The memory 11 may in some embodiments be an internal storage unit of the electronic device, such as a mobile hard disk of the electronic device. The memory 11 may in other embodiments also be an external storage device of the electronic device, such as a plug-in mobile hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card) or the like, which are provided on the electronic device. Further, the memory 11 may also include both an internal storage unit and an external storage device of the electronic device. The memory 11 may be used not only for storing application software installed in an electronic device and various types of data, such as codes of an intention recognition program based on model fusion, but also for temporarily storing data that has been output or is to be output.
The processor 10 may be comprised of integrated circuits in some embodiments, for example, a single packaged integrated circuit, or may be comprised of multiple integrated circuits packaged with the same or different functions, including one or more central processing units (Central Processing unit, CPU), microprocessors, digital processing chips, graphics processors, combinations of various control chips, and the like. The processor 10 is a Control Unit (Control Unit) of the electronic device, connects various components of the entire electronic device using various interfaces and lines, executes or executes programs or modules (e.g., an intention recognition program based on model fusion, etc.) stored in the memory 11, and invokes data stored in the memory 11 to perform various functions of the electronic device and process data.
The communication interface 12 is used for communication between the electronic device and other devices, including network interfaces and user interfaces. Optionally, the network interface may include a wired interface and/or a wireless interface (e.g., WI-FI interface, bluetooth interface, etc.), typically used to establish a communication connection between the electronic device and other electronic devices. The user interface may be a Display (Display), an input unit such as a Keyboard (Keyboard), or alternatively a standard wired interface, a wireless interface. Alternatively, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, or the like. The display may also be referred to as a display screen or display unit, as appropriate, for displaying information processed in the electronic device and for displaying a visual user interface.
The bus 13 may be a peripheral component interconnect standard (peripheral component interconnect, PCI) bus, or an extended industry standard architecture (extended industry standard architecture, EISA) bus, among others. The bus 13 may be classified into an address bus, a data bus, a control bus, and the like. The bus 13 is arranged to enable a connection communication between the memory 11 and at least one processor 10 etc.
Fig. 3 shows only an electronic device with components, and it will be understood by those skilled in the art that the structure shown in fig. 3 is not limiting of the electronic device and may include fewer or more components than shown, or may combine certain components, or a different arrangement of components.
For example, although not shown, the electronic device may further include a power source (such as a battery) for supplying power to the respective components, and preferably, the power source may be logically connected to the at least one processor 10 through a power management device, so that functions of charge management, discharge management, power consumption management, and the like are implemented through the power management device. The power supply may also include one or more of any of a direct current or alternating current power supply, recharging device, power failure detection circuit, power converter or inverter, power status indicator, etc. The electronic device may further include various sensors, bluetooth modules, wi-Fi modules, etc., which are not described herein.
Further, the electronic device may also include a network interface, optionally, the network interface may include a wired interface and/or a wireless interface (e.g., WI-FI interface, bluetooth interface, etc.), typically used to establish a communication connection between the electronic device and other electronic devices.
Optionally, the electronic device may further comprise a user interface, which may be a Display, an input unit, such as a Keyboard (Keyboard), or a standard wired interface, a wireless interface. Alternatively, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch, or the like. The display may also be referred to as a display screen or display unit, as appropriate, for displaying information processed in the electronic device and for displaying a visual user interface.
It should be understood that the embodiments described are for illustrative purposes only and are not limited to this configuration in the scope of the patent application.
The model fusion based intent recognition program stored in the memory 11 in the electronic device is a combination of instructions that, when executed in the processor 10, may implement:
Acquiring a service data set of a user, and extracting features of the service data set to obtain a service feature set;
performing time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and summarizing the time sequence feature set and the service feature set to obtain a standard feature set;
classifying the users to obtain single-side users and multi-side users, and distributing the standard feature sets to the single-side users and the multi-side users to obtain single-side data sets corresponding to the single-side users and multi-side data sets corresponding to the multi-side users;
performing missing value filling and abnormal value processing on the unilateral data set and the multilateral data set to obtain a standard unilateral data set and a standard multilateral data set;
dividing the standard unilateral data set into a training data set and a test data set, constructing a single scene model based on the training data set, performing test processing on the single scene model by utilizing the test data set, and outputting a standard single scene model according to an obtained first test result;
constructing a multi-scene model according to the standard polygonal data set, testing the multi-scene model, and outputting a standard multi-scene model according to an obtained second test result;
Carrying out model fusion on a preset number of standard single scene models to obtain single scene fusion models, and carrying out model fusion on a preset number of standard multi-scene models to obtain multi-scene fusion models, wherein the single scene fusion models and the multi-scene fusion models both comprise fusion formulas;
and acquiring data to be identified, identifying a user category corresponding to the data to be identified, and calling the single scene fusion model or the multi-scene fusion model according to the user category to perform intention identification processing on the data to be identified to obtain an intention identification result.
Specifically, the specific implementation method of the above instructions by the processor 10 may refer to the description of the relevant steps in the corresponding embodiment of fig. 1, which is not repeated herein.
Further, the electronic device integrated modules/units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. The computer readable storage medium may be volatile or nonvolatile. For example, the computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM).
The present invention also provides a computer readable storage medium storing a computer program which, when executed by a processor of an electronic device, can implement:
acquiring a service data set of a user, and extracting features of the service data set to obtain a service feature set;
performing time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and summarizing the time sequence feature set and the service feature set to obtain a standard feature set;
classifying the users to obtain single-side users and multi-side users, and distributing the standard feature sets to the single-side users and the multi-side users to obtain single-side data sets corresponding to the single-side users and multi-side data sets corresponding to the multi-side users;
performing missing value filling and abnormal value processing on the unilateral data set and the multilateral data set to obtain a standard unilateral data set and a standard multilateral data set;
dividing the standard unilateral data set into a training data set and a test data set, constructing a single scene model based on the training data set, performing test processing on the single scene model by utilizing the test data set, and outputting a standard single scene model according to an obtained first test result;
Constructing a multi-scene model according to the standard polygonal data set, testing the multi-scene model, and outputting a standard multi-scene model according to an obtained second test result;
carrying out model fusion on a preset number of standard single scene models to obtain single scene fusion models, and carrying out model fusion on a preset number of standard multi-scene models to obtain multi-scene fusion models, wherein the single scene fusion models and the multi-scene fusion models both comprise fusion formulas;
and acquiring data to be identified, identifying a user category corresponding to the data to be identified, and calling the single scene fusion model or the multi-scene fusion model according to the user category to perform intention identification processing on the data to be identified to obtain an intention identification result.
In the several embodiments provided in the present invention, it should be understood that the disclosed apparatus, device and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the modules is merely a logical function division, and there may be other manners of division when actually implemented.
The modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical units, may be located in one place, or may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional module in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units can be realized in a form of hardware or a form of hardware and a form of software functional modules.
It will be evident to those skilled in the art that the invention is not limited to the details of the foregoing illustrative embodiments, and that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof.
The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. Any reference signs in the claims shall not be construed as limiting the claim concerned.
The blockchain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm and the like. The Blockchain (Blockchain), which is essentially a decentralised database, is a string of data blocks that are generated by cryptographic means in association, each data block containing a batch of information of network transactions for verifying the validity of the information (anti-counterfeiting) and generating the next block. The blockchain may include a blockchain underlying platform, a platform product services layer, an application services layer, and the like.
Furthermore, it is evident that the word "comprising" does not exclude other elements or steps, and that the singular does not exclude a plurality. A plurality of units or means recited in the system claims can also be implemented by means of software or hardware by means of one unit or means. The terms second, etc. are used to denote a name, but not any particular order.
Finally, it should be noted that the above-mentioned embodiments are merely for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.
Claims (10)
1. An intention recognition method based on model fusion, which is characterized by comprising the following steps:
acquiring a service data set of a user, and extracting features of the service data set to obtain a service feature set;
performing time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and summarizing the time sequence feature set and the service feature set to obtain a standard feature set;
classifying the users to obtain single-side users and multi-side users, and distributing the standard feature sets to the single-side users and the multi-side users to obtain single-side data sets corresponding to the single-side users and multi-side data sets corresponding to the multi-side users;
Performing missing value filling and abnormal value processing on the unilateral data set and the multilateral data set to obtain a standard unilateral data set and a standard multilateral data set;
dividing the standard unilateral data set into a training data set and a test data set, constructing a single scene model based on the training data set, performing test processing on the single scene model by utilizing the test data set, and outputting a standard single scene model according to an obtained first test result;
constructing a multi-scene model according to the standard polygonal data set, testing the multi-scene model, and outputting a standard multi-scene model according to an obtained second test result;
carrying out model fusion on a preset number of standard single scene models to obtain single scene fusion models, and carrying out model fusion on a preset number of standard multi-scene models to obtain multi-scene fusion models, wherein the single scene fusion models and the multi-scene fusion models both comprise fusion formulas;
and acquiring data to be identified, identifying a user category corresponding to the data to be identified, and calling the single scene fusion model or the multi-scene fusion model according to the user category to perform intention identification processing on the data to be identified to obtain an intention identification result.
2. The model fusion-based intent recognition method as recited in claim 1, wherein said constructing a single scene model based on said training dataset includes:
constructing an initial decision tree by using the standard feature set, and performing decision tree adding processing on the initial decision tree to obtain an updated decision tree;
inputting the training data set into the updating decision tree to obtain a predicted value set, and calculating a loss value of the updating decision tree according to the predicted value set and a preset kitchen loss function;
and when the loss value is greater than or equal to a preset loss threshold value, executing the operation of adding the decision tree to the initial decision tree again until the loss value is smaller than the loss threshold value, and outputting the current updated decision tree as a single scene model.
3. The model fusion-based intent recognition method as recited in claim 2, wherein constructing an initial decision tree using the set of standard features includes:
classifying and labeling the standard feature set to obtain labels corresponding to the standard feature set;
selecting one label at will as a segmentation point, and taking the segmentation point as a root node of an original decision tree;
Generating the child nodes of the cut points and distributing the standard feature set to the child nodes to obtain an initial decision tree.
4. The model fusion-based intent recognition method of claim 2, wherein the preset focal loss function is:
FL(p t )＝-α t (1-p t ) γ log(p t )
wherein FL (p) t ) For the loss value, p, of the updated decision tree t Alpha is the predicted value in the predicted value set t And gamma is a preset second weight for the preset first weight.
5. The method for identifying intent based on model fusion as claimed in claim 1, wherein said performing missing value filling and outlier processing on said single-sided dataset to obtain a standard single-sided dataset includes:
detecting the unilateral data set according to a pre-constructed missing detection statement, and filling the missing part by using a preset filling value;
sorting the unilateral data sets according to the sequence from big to small, and screening the medians positioned at the middle positions from the sorted unilateral data sets;
respectively calculating the similarity between the median and a plurality of single-side data in the single-side data set;
judging the single-side data with the similarity larger than a preset abnormal threshold value as an abnormal value, and deleting the abnormal value to obtain a standard single-side data set.
6. The method for identifying intent based on model fusion as claimed in any one of claims 1 to 5, wherein the performing time sequence calculation on the service data in the service data set according to the service feature set to obtain a time sequence feature set includes:
extracting service data corresponding to the service feature set in the service data set;
and calculating time sequence characteristics of the service data corresponding to the service characteristic set according to a preset time domain characteristic calculation formula to obtain the time sequence characteristic set.
7. The method for identifying intent based on model fusion as claimed in any one of claims 1 to 5, wherein the feature extraction of the service data set to obtain a service feature set includes:
acquiring a plurality of preset feature dimensions and feature word libraries corresponding to the feature dimensions;
and screening the data corresponding to the feature words in the feature word library in the service data set to obtain a service feature set.
8. An intent recognition device based on model fusion, the device comprising:
the feature extraction module is used for acquiring a service data set of a user, extracting features of the service data set to obtain a service feature set, performing time sequence calculation on service data in the service data set according to the service feature set to obtain a time sequence feature set, and summarizing the time sequence feature set and the service feature set to obtain a standard feature set;
The data processing module is used for classifying the users to obtain single-side users and multi-side users, distributing the standard feature sets to the single-side users and the multi-side users to obtain single-side data sets corresponding to the single-side users and multi-side data sets corresponding to the multi-side users, and carrying out missing value filling and abnormal value processing on the single-side data sets and the multi-side data sets to obtain standard single-side data sets and standard multi-side data sets;
the standard single scene model construction module is used for dividing the standard single-side data set into a training data set and a test data set, constructing a single scene model based on the training data set, carrying out test processing on the single scene model by utilizing the test data set, and outputting a standard single scene model according to an obtained first test result;
the standard multi-scene model construction module is used for constructing a multi-scene model according to the standard polygonal data set, testing the multi-scene model and outputting a standard multi-scene model according to the obtained second test result;
the model fusion module is used for carrying out model fusion on a preset number of standard single scene models to obtain single scene fusion models, carrying out model fusion on a preset number of standard multi-scene models to obtain multi-scene fusion models, wherein the single scene fusion models and the multi-scene fusion models both comprise fusion formulas;
The intention recognition module is used for acquiring data to be recognized, recognizing a user category corresponding to the data to be recognized, and calling the single scene fusion model or the multi-scene fusion model according to the user category to perform intention recognition processing on the data to be recognized to obtain an intention recognition result.
9. An electronic device, the electronic device comprising:
at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,
the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the model fusion-based intent recognition method as claimed in any one of claims 1 to 7.
10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the model fusion based intent recognition method as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110913907.7A CN113591881B (en) | 2021-08-10 | 2021-08-10 | Intention recognition method and device based on model fusion, electronic equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110913907.7A CN113591881B (en) | 2021-08-10 | 2021-08-10 | Intention recognition method and device based on model fusion, electronic equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113591881A CN113591881A (en) | 2021-11-02 |
| CN113591881B true CN113591881B (en) | 2023-06-02 |
Family
ID=78256855
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110913907.7A Active CN113591881B (en) | 2021-08-10 | 2021-08-10 | Intention recognition method and device based on model fusion, electronic equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113591881B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116173484A (en) * | 2023-03-03 | 2023-05-30 | 乐渊网络科技(上海)有限公司 | Motion data processing method and device and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112036550A (en) * | 2020-09-04 | 2020-12-04 | 平安科技（深圳）有限公司 | Client intention identification method and device based on artificial intelligence and computer equipment |
| CN112988963A (en) * | 2021-02-19 | 2021-06-18 | 平安科技（深圳）有限公司 | User intention prediction method, device, equipment and medium based on multi-process node |
| CN113204971A (en) * | 2021-03-26 | 2021-08-03 | 南京邮电大学 | Scene self-adaptive Attention multi-intention identification method based on deep learning |
-
2021
- 2021-08-10 CN CN202110913907.7A patent/CN113591881B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112036550A (en) * | 2020-09-04 | 2020-12-04 | 平安科技（深圳）有限公司 | Client intention identification method and device based on artificial intelligence and computer equipment |
| CN112988963A (en) * | 2021-02-19 | 2021-06-18 | 平安科技（深圳）有限公司 | User intention prediction method, device, equipment and medium based on multi-process node |
| CN113204971A (en) * | 2021-03-26 | 2021-08-03 | 南京邮电大学 | Scene self-adaptive Attention multi-intention identification method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113591881A (en) | 2021-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022141861A1 (en) | Emotion classification method and apparatus, electronic device, and storage medium | |
| CN113449187B (en) | Product recommendation method, device, equipment and storage medium based on double images | |
| CN113051356B (en) | Open relation extraction method and device, electronic equipment and storage medium | |
| CN113157927B (en) | Text classification method, apparatus, electronic device and readable storage medium | |
| US8793201B1 (en) | System and method for seeding rule-based machine learning models | |
| CN115018588A (en) | Product recommendation method and device, electronic equipment and readable storage medium | |
| CN113658002B (en) | Transaction result generation method and device based on decision tree, electronic equipment and medium | |
| CN113434542B (en) | Data relationship identification method and device, electronic equipment and storage medium | |
| CN113591881B (en) | Intention recognition method and device based on model fusion, electronic equipment and medium | |
| CN113268665A (en) | Information recommendation method, device and equipment based on random forest and storage medium | |
| CN112579781A (en) | Text classification method and device, electronic equipment and medium | |
| CN112269875A (en) | Text classification method and device, electronic equipment and storage medium | |
| CN115146653B (en) | Dialogue scenario construction method, device, equipment and storage medium | |
| CN113705692B (en) | Emotion classification method and device based on artificial intelligence, electronic equipment and medium | |
| CN116340516A (en) | Entity relation cluster extraction method, device, equipment and storage medium | |
| CN113656586B (en) | Emotion classification method, emotion classification device, electronic equipment and readable storage medium | |
| CN115114408A (en) | Multi-modal emotion classification method, device, equipment and storage medium | |
| CN114219367A (en) | User scoring method, device, equipment and storage medium | |
| CN113268614A (en) | Label system updating method and device, electronic equipment and readable storage medium | |
| CN110262906B (en) | Interface label recommendation method and device, storage medium and electronic equipment | |
| CN113706207A (en) | Order transaction rate analysis method, device, equipment and medium based on semantic analysis | |
| CN113343102A (en) | Data recommendation method and device based on feature screening, electronic equipment and medium | |
| CN113626605A (en) | Information classification method and device, electronic equipment and readable storage medium | |
| CN114723488B (en) | Course recommendation method and device, electronic equipment and storage medium | |
| CN113592606B (en) | Product recommendation method, device, equipment and storage medium based on multiple decisions |
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 |