CN117455618A

CN117455618A - Object processing method and device and electronic equipment

Info

Publication number: CN117455618A
Application number: CN202311436439.4A
Authority: CN
Inventors: 刘桓硕; 侯金鑫; 徐明成; 孙从阳; 薛涛
Original assignee: Tianyi Electronic Commerce Co Ltd
Current assignee: Tianyi Electronic Commerce Co Ltd
Priority date: 2023-10-31
Filing date: 2023-10-31
Publication date: 2024-01-26

Abstract

The application discloses an object processing method and device and electronic equipment. Relates to the technical field of artificial intelligence, and the method comprises the following steps: acquiring a plurality of first objects and target data information of each first object; predicting the category of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial category information corresponding to each first object; predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object; and determining target category information corresponding to each first object according to the first initial category information and the second initial category information. According to the method and the device, the problem that in the related art, the accuracy of classifying the users is low due to the fact that the users are classified manually is solved.

Description

Object processing method and device and electronic equipment

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a method and a device for processing an object and electronic equipment.

Background

In the recommendation of telecommunication products, accurate analysis and prediction are required for users to be able to make more effective recommendation strategies. However, the conventional recommendation method is often formulated based on experience and intuition, that is, users are classified by manual method into users with high intention to purchase telecommunication products and users with low intention to purchase telecommunication products, and then recommendation strategies of telecommunication products are formulated according to classification results.

Aiming at the problem that the classification of users is low in accuracy caused by manually classifying the users in the related technology, no effective solution is proposed at present.

Disclosure of Invention

The main purpose of the application is to provide a method and a device for processing an object and an electronic device, so as to solve the problem that in the related art, the accuracy of classifying users is low due to the fact that the users are classified manually.

In order to achieve the above object, according to one aspect of the present application, there is provided a method of processing an object. The method comprises the following steps: acquiring a plurality of first objects and target data information of each first object, wherein the target data information at least comprises historical telecommunication product purchase information of the first object and identity information of the first object; predicting the class of the first object based on the target data information through a voting classifier in a fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value; predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object; and determining target category information corresponding to each first object according to the first initial category information and the second initial category information.

Further, predicting the class of the first object based on the target data information through a voting classifier in the fusion model, and obtaining first initial class information corresponding to each first object includes: extracting features of the target data information through a plurality of target machine learning models in the voting classifier to obtain feature vectors corresponding to each first object; predicting the class of the first object based on the feature vector through each target machine learning model to obtain a first probability value of the first object belonging to the first class and a second probability value of the first object belonging to the second class, which are output by each target machine learning model; average calculation is carried out according to the first probability value to obtain a first average probability value of the first object belonging to the first category, and average calculation is carried out according to the second probability value to obtain a second average probability value of the first object belonging to the second category; and obtaining the first initial category information corresponding to each first object according to the first average probability value and the second average probability value.

Further, predicting, by the virtual classifier in the fusion model, the class of the first object based on the target data information, and obtaining second initial class information corresponding to each first object includes: matching the target data information through a plurality of expert rules in the virtual classifier to obtain a matching result; and obtaining second initial category information corresponding to each first object according to the matching result.

Further, determining target category information corresponding to each first object according to the first initial category information and the second initial category information includes: judging whether the first initial category information and the second initial category information are the same or not; if the first initial category information is different from the second initial category information, determining a priority between the voting classifier and the virtual classifier; if the priority of the voting classifier is higher than that of the virtual classifier, determining the first initial class information as the target class information; and if the priority of the voting classifier is lower than that of the virtual classifier, determining the second initial category information as the target category information.

Further, before predicting the class of the first object based on the target data information by the voting classifier in the fusion model to obtain first initial class information corresponding to each first object, the method further includes: acquiring a training set, wherein the training set at least comprises a plurality of sample objects, sample data information corresponding to each sample object and real category information corresponding to each sample object; training a plurality of machine learning models according to the training set to obtain a plurality of target machine learning models, and carrying out model fusion on the plurality of target machine learning models in a voting classifier mode to obtain the voting classifier; determining a plurality of expert rules according to the training set, and obtaining the virtual classifier according to the expert rules; and carrying out model fusion according to the voting classifier and the virtual classifier to obtain the fusion model.

Further, performing model fusion on the plurality of target machine learning models by means of a voting classifier, and obtaining the voting classifier includes: constructing an initial voting classifier, and inputting the multiple target machine learning models into the initial voting classifier to obtain a processed initial voting classifier; and setting the attribute of the processed initial voting classifier to obtain the voting classifier.

Further, determining a plurality of expert rules from the training set includes: obtaining a plurality of characteristic information corresponding to each sample object according to the sample data information in the training set; calculating the information value of each piece of characteristic information to obtain the information value corresponding to each piece of characteristic information, and calculating the evidence weight of each piece of characteristic information to obtain the evidence weight corresponding to each piece of characteristic information; screening the plurality of characteristic information according to the information value and the evidence weight to obtain a plurality of target characteristic information; and determining the expert rules according to the target feature information.

Further, after determining the target category information corresponding to each first object according to the first initial category information and the second initial category information, the method further includes: and determining a plurality of second objects from the plurality of first objects according to the target category information, wherein the second objects are objects with the probability of purchasing the target telecommunication product being larger than the preset value.

In order to achieve the above object, according to another aspect of the present application, there is provided a processing apparatus of an object. The device comprises: a first obtaining unit, configured to obtain a plurality of first objects and target data information of each first object, where the target data information includes at least historical telecommunication product purchase information of the first object and identity information of the first object; the first prediction unit is used for predicting the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value; the second prediction unit is used for predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object; and the first determining unit is used for determining target category information corresponding to each first object according to the first initial category information and the second initial category information.

Further, the first prediction unit includes: the extraction module is used for extracting the characteristics of the target data information through a plurality of target machine learning models in the voting classifier to obtain a characteristic vector corresponding to each first object; the prediction module is used for predicting the class of the first object based on the feature vector through each target machine learning model to obtain a first probability value of the first object belonging to the first class and a second probability value of the first object belonging to the second class, which are output by each target machine learning model; the first calculation module is used for carrying out average calculation according to the first probability value to obtain a first average probability value of the first object belonging to the first category, and carrying out average calculation according to the second probability value to obtain a second average probability value of the first object belonging to the second category; and the first determining module is used for obtaining the first initial category information corresponding to each first object according to the first average probability value and the second average probability value.

Further, the second prediction unit includes: the matching module is used for matching the target data information through a plurality of expert rules in the virtual classifier to obtain a matching result; and the second determining module is used for obtaining second initial category information corresponding to each first object according to the matching result.

Further, the first determination unit includes: the judging module is used for judging whether the first initial category information and the second initial category information are the same or not; a third determining module, configured to determine a priority between the voting classifier and the virtual classifier if the first initial class information is different from the second initial class information; a fourth determining module, configured to determine the first initial category information as the target category information if the priority of the voting classifier is higher than the priority of the virtual classifier; and a fifth determining module, configured to determine the second initial category information as the target category information if the priority of the voting classifier is lower than the priority of the virtual classifier.

Further, the apparatus further comprises: the second obtaining unit is used for obtaining a training set before predicting the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the training set at least comprises a plurality of sample objects, sample data information corresponding to each sample object and real class information corresponding to each sample object; the training unit is used for training a plurality of machine learning models according to the training set to obtain a plurality of target machine learning models, and carrying out model fusion on the plurality of target machine learning models in a voting classifier mode to obtain the voting classifier; the second determining unit is used for determining a plurality of expert rules according to the training set and obtaining the virtual classifier according to the expert rules; and the fusion unit is used for carrying out model fusion according to the voting classifier and the virtual classifier to obtain the fusion model.

Further, the training unit includes: the construction module is used for constructing an initial voting classifier, inputting the target machine learning models into the initial voting classifier and obtaining a processed initial voting classifier; and the setting module is used for setting the attribute of the processed initial voting classifier to obtain the voting classifier.

Further, the second determination unit includes: a sixth determining module, configured to obtain a plurality of feature information corresponding to each sample object according to the sample data information in the training set; the second calculation module is used for calculating the information value of each piece of characteristic information to obtain the information value corresponding to each piece of characteristic information, and calculating the evidence weight of each piece of characteristic information to obtain the evidence weight corresponding to each piece of characteristic information; the screening module is used for screening the plurality of characteristic information according to the information value and the evidence weight to obtain a plurality of target characteristic information; and a seventh determining module, configured to determine the plurality of expert rules according to the plurality of target feature information.

Further, the apparatus further comprises: and the third determining unit is used for determining a plurality of second objects from the plurality of first objects according to the target category information after determining the target category information corresponding to each first object according to the first initial category information and the second initial category information, wherein the second objects are objects with the probability of purchasing the target telecommunication product being larger than the preset value.

To achieve the above object, according to another aspect of the present application, there is also provided an electronic device, including one or more processors and a memory for storing a processing method of the one or more processors for implementing the object described in any one of the above.

Through the application, the following steps are adopted: acquiring a plurality of first objects and target data information of each first object, wherein the target data information at least comprises historical telecommunication product purchase information of the first object and identity information of the first object; predicting the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value; predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object; the target category information corresponding to each first object is determined according to the first initial category information and the second initial category information, so that the problem that the accuracy of classifying the users is low due to the fact that the users are classified manually in the related technology is solved. In the scheme, the voting classifier in the fusion model predicts the category of the first object by utilizing the historical telecom product purchase information and the identity information of the first object to obtain first initial category information, and predicts the category of the first object by utilizing the historical telecom product purchase information and the identity information of the first object by utilizing the virtual classifier in the fusion model to obtain second initial category information, finally obtains the target category information of the first object according to the first initial category information and the second initial category information, classifies the user by the fusion model, avoids subjectivity of manual classification, improves accuracy of classification of the user, predicts the category twice by the voting classifier and the virtual classifier, improves accuracy of classification of the user, and can more reasonably formulate a telecom product recommendation strategy by improving accuracy of classification of the user, thereby achieving the effect of improving the recommendation success rate of the telecom product.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application, illustrate and explain the application and are not to be construed as limiting the application. In the drawings:

FIG. 1 is a flow chart of a method of processing an object provided in accordance with an embodiment of the present application;

FIG. 2 is a schematic diagram of a method of processing an object provided according to an embodiment of the present application;

FIG. 3 is a schematic diagram of a processing device for an object provided in accordance with an embodiment of the present application;

fig. 4 is a schematic diagram of an electronic device provided according to an embodiment of the present application.

Detailed Description

It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.

In order to make the present application solution better understood by those skilled in the art, the following description will be made in detail and with reference to the accompanying drawings in the embodiments of the present application, it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application.

It should be noted that the terms "first," "second," and the like in the description and claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate in order to describe the embodiments of the present application described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

It should be noted that, related information (including, but not limited to, user equipment information, user personal information, etc.) and data (including, but not limited to, data for presentation, analyzed data, etc.) related to the present disclosure are information and data authorized by a user or sufficiently authorized by each party. For example, an interface is provided between the system and the relevant user or institution, before acquiring the relevant information, the system needs to send an acquisition request to the user or institution through the interface, and acquire the relevant information after receiving the consent information fed back by the user or institution.

The invention will be described with reference to preferred implementation steps, and fig. 1 is a flowchart of a method for processing an object according to an embodiment of the present application, as shown in fig. 1, where the method includes the following steps:

step S101, acquiring a plurality of first objects and target data information of each first object, wherein the target data information at least comprises historical telecommunication product purchase information of the first object and identity information of the first object;

optionally, a plurality of first objects to be classified are explicitly required, and historical telecommunication product purchase information, identity information and the like of each first object are acquired from a database. From this information, the first object's propensity to purchase telecommunication products can be accurately assessed.

Step S102, predicting the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value;

Optionally, after the target data information is obtained, the target data information is input into a fusion model, and the class of the first object is predicted based on the target data information through a voting classifier in the fusion model, so as to obtain first initial class information corresponding to each first object.

It should be noted that the categories of the first object are a first category and a second category, the first object of the first category is an object with a probability of purchasing the target telecommunication product being greater than a preset value, and the first object of the second category is an object with a probability of purchasing the target telecommunication product being less than or equal to the preset value, and in short, the probability of purchasing the target telecommunication product of the first category is greater than the probability of purchasing the target telecommunication product of the second category.

It should be noted that the fusion model may be fused into a pmml (predictive model markup language) file, so as to facilitate the subsequent management and maintenance of the fusion model.

Step S103, predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object;

optionally, after the target data information is obtained, a virtual classifier in the fusion model is used for carrying out second prediction on the category of the first object based on the target data information, so as to obtain second initial category information corresponding to the first object. In an alternative embodiment, the virtual classifier may be a virtual classifier derived based on expert rules.

Step S104, determining target category information corresponding to each first object according to the first initial category information and the second initial category information.

Optionally, after the first initial category information and the second initial category information are obtained, determining target category information corresponding to each first object according to the first initial category information and the second initial category information.

In an alternative embodiment, if the first initial category information and the second initial category information are the same, the first initial category information or the second initial category information is directly determined as the target category information of the first object. If the first initial category information is different from the second initial category information, determining target category information of the first object according to the priorities of the first initial category information and the second initial category information.

In an alternative embodiment, after the target class information of the first object is obtained, a recommendation policy of the target telecommunication product may be formulated according to the target class information of the first object, for example, the target telecommunication product is recommended to the first class object.

In summary, the voting classifier in the fusion model predicts the category of the first object by using the historical telecom product purchase information and the identity information of the first object to obtain the first initial category information, and predicts the category of the first object by using the historical telecom product purchase information and the identity information of the first object by using the virtual classifier in the fusion model to obtain the second initial category information, finally obtains the target category information of the first object according to the first initial category information and the second initial category information, classifies the user by the fusion model, avoids the subjectivity of manual classification, improves the accuracy of classification of the user, and makes twice category predictions by the voting classifier and the virtual classifier, also improves the accuracy of classification of the user, and can reasonably recommend telecom products by improving the accuracy of classification of the user, thereby achieving the effect of improving the recommendation success rate of telecom products.

Optionally, in the method for processing an object provided in the embodiment of the present application, predicting, by a vote classifier in a fusion model, a class of a first object based on target data information, and obtaining first initial class information corresponding to each first object includes: extracting features of the target data information through a plurality of target machine learning models in the voting classifier to obtain feature vectors corresponding to each first object; predicting the class of the first object based on the feature vector through each target machine learning model to obtain a first probability value of the first object belonging to the first class and a second probability value of the first object belonging to the second class, which are output by each target machine learning model; average calculation is carried out according to the first probability value to obtain a first average probability value of the first object belonging to the first category, and average calculation is carried out according to the second probability value to obtain a second average probability value of the first object belonging to the second category; and obtaining first initial category information corresponding to each first object according to the first average probability value and the second average probability value.

Optionally, the above voting classifier is fused by a plurality of target machine learning models, so predicting the class of the first object based on the target data information by the voting classifier in the fused model includes: and carrying out feature extraction on the target data information through each target machine learning model to obtain a corresponding feature vector (namely the feature vector corresponding to each first object), then predicting the class of the first object through each target machine learning model based on the feature vector, obtaining a first probability value of the first object belonging to the first class and a second probability value of the first object belonging to the second class output by each target machine learning model, and carrying out average calculation on the first probability value and the second probability value to obtain a first average probability value of the first object belonging to the first class and a second average probability value of the first object belonging to the second class.

After the first average probability value and the second average probability value are obtained, first initial category information corresponding to each first object is obtained through the first average probability value and the second average probability value, for example, if the first average probability value is greater than a threshold (for example, 0.8), the first initial category information corresponding to the first object is determined to be the first category.

In an alternative embodiment, the plurality of target machine learning models may be composed of decision trees, support vector machines, and neural network models.

In summary, by fusing a plurality of machine learning models into the voting classifier, the advantages of various models can be fully utilized, and more accurate prediction results can be obtained, so that more effective product recommendation strategies can be formulated.

Optionally, in the method for processing an object provided in the embodiment of the present application, predicting, by a virtual classifier in a fusion model, a class of a first object based on target data information, and obtaining second initial class information corresponding to each first object includes: matching the target data information through a plurality of expert rules in the virtual classifier to obtain a matching result; and obtaining second initial category information corresponding to each first object according to the matching result.

In an alternative embodiment, predicting the class of the first object based on the target data information by means of a virtual classifier in the fusion model comprises the steps of: and matching the target data information through a plurality of expert rules in the virtual classifier, and then determining second initial category information corresponding to each first object according to the matching result. For example, expert rules may identify the class of an object as a first class for the object presence feature 1.

The second initial category system information of the first object can be directly and quickly obtained through a plurality of expert rules in the virtual classifier.

Optionally, in the method for processing an object provided in the embodiment of the present application, determining, according to the first initial category information and the second initial category information, target category information corresponding to each first object includes: judging whether the first initial category information and the second initial category information are the same or not; if the first initial category information is different from the second initial category information, determining the priority between the voting classifier and the virtual classifier; if the priority of the voting classifier is higher than that of the virtual classifier, determining the first initial class information as target class information; and if the priority of the voting classifier is lower than that of the virtual classifier, determining the second initial category information as target category information.

In an alternative embodiment, determining the target category information corresponding to each first object according to the first initial category information and the second initial category information includes the following steps: judging whether the first initial category information and the second initial category information are the same, and if the first initial category information and the second initial category information are the same, directly determining the first initial category information or the second initial category information as target category information of the first object.

If the first initial category information is not the same as the second initial category information, a priority between the voting classifier and the virtual classifier needs to be determined. It should be noted that, the priority between the voting classifier and the virtual classifier may be set according to the situation.

After the priority is obtained, if the priority of the voting classifier is higher than the priority of the virtual classifier, determining the first initial class information as target class information; if the priority of the voting classifier is lower than the priority of the virtual classifier, the second initial class information is determined as the target class information.

By setting the priority between the voting classifier and the virtual classifier and then determining the class of the first object according to the priority, the classification accuracy is improved.

Optionally, in the method for processing an object provided in the embodiment of the present application, before predicting, by a vote classifier in a fusion model, a class of a first object based on target data information, to obtain first initial class information corresponding to each first object, the method further includes: acquiring a training set, wherein the training set at least comprises a plurality of sample objects, sample data information corresponding to each sample object and real category information corresponding to each sample object; training a plurality of machine learning models according to the training set to obtain a plurality of target machine learning models, and carrying out model fusion on the plurality of target machine learning models in a voting classifier mode to obtain a voting classifier; determining a plurality of expert rules according to the training set, and obtaining a virtual classifier according to the plurality of expert rules; and carrying out model fusion according to the voting classifier and the virtual classifier to obtain a fusion model.

In an alternative embodiment, the fusion model described above is obtained using the following steps: and constructing the training set through historical telecommunication product recommendation results. The training set at least comprises a plurality of sample objects, sample data information corresponding to each sample object and real category information corresponding to each sample object. For example, according to historical telecommunication product recommendation result data, positive and negative samples of modeling and rule analysis are arranged, the real category information of users with purchase will is marked as 1, the real category information of users without purchase will is marked as 0, and transaction data information, identity information and the like of the users are obtained from a database.

After the training set is obtained, a plurality of machine learning models such as decision trees, random forests, logistic regression, lightGBM and the like are selected, and the machine learning models are trained and optimized through the training set, namely the plurality of machine learning models are trained, so that a plurality of target machine learning models are obtained.

After obtaining a plurality of target machine learning models, each model can be summarized and transferred to a voting classifier through a sklearn. And then determining a plurality of expert rules through the training set, and obtaining the virtual classifier according to the expert rules. And finally, carrying out model fusion according to the voting classifier and the virtual classifier to obtain a fusion model.

In an alternative embodiment, after obtaining a plurality of expert rules, a virtual classifier is created by using the DummyClassifier class in the scikit-learn library, all expert rules are written into the DummyClassifier, and finally input features and labels are transmitted into the virtual classifier for training by using a fit method, so that the virtual classifier fused with the expert rules is obtained.

In an alternative embodiment, performing model fusion according to the voting classifier and the virtual classifier, and obtaining a fusion model includes: integrating the VotingClassifier classifier fused with a plurality of machine learning models and the DummyClassifier classifier fused with a plurality of expert rules, inputting the two classifiers into sklearn2pmml.ensable.SelectFirstClassifier, and setting priority through business analysis to realize the fusion of the machine learning models and the expert rules. Finally, the obtained SelectFirstClassifier fusion classifier is converted into PMMLPipeline, and the n_output_value of the_final_counter attribute is set to be 1, so that the output of the classifier is ensured to be a single scalar value, namely, the corresponding identifier of the output category is ensured.

In an alternative embodiment, the fusion model described above may be converted to PMML pipeline, i.e. a plurality of different machine learning models and expert rules are fused into one PMML file. When the model is required to be updated subsequently, a plurality of models and expert rules can be updated only by updating one pmml file, the deployment flow of the models can be greatly simplified, the deployment and management of users are facilitated, and meanwhile, the portability and interoperability of the models can be improved by using the pmml file, and the pmml file is in a universal format and can be supported by a plurality of different software systems and programming languages.

In summary, fusing a plurality of different machine learning models with expert rules can help the models to better address specific scenarios and problems, improving the accuracy of classification of users.

Optionally, in the method for processing an object provided in the embodiment of the present application, model fusion is performed on a plurality of target machine learning models by means of a voting classifier, and obtaining the voting classifier includes: constructing an initial voting classifier, and inputting a plurality of target machine learning models into the initial voting classifier to obtain a processed initial voting classifier; and setting the attribute of the processed initial voting classifier to obtain the voting classifier.

In an alternative embodiment, the voting classifier described above is obtained using the following steps: the model pkl file trained in S5.2 (i.e., the target machine learning model described above) is imported using joblib. Load, and each model summary is passed to the initial vote classifier using sklearn. Ensable. Votingclassification, where the prediction of each model is treated as a probability of that class using Soft, and the final classification result is the average of all input model predictions.

After the processed initial vote classifier is obtained, setting the attribute of the processed initial vote classifier: the trained machine learning models are assigned to the estimator_attribute in the new voting classifier VotingClassifier and set to a pmmlpipeline object for direct use in subsequent prediction classification.

An encoder LabelEncoder object encoding the class tag [0,1] is initialized and assigned to the le_attribute of the voting classifier VotingClassiier.

The class_attribute of the voting classifier VotingClassifier is set to the class_attribute of LabelEncoder object le_to ensure that the voting classifier can be predicted directly without retraining.

The _final_ estimator, target _fields attribute of the voting classifier is set in an overlaying mode, and the prediction result of the voting classifier is controlled to be a model score, so that the voting classifier realizing a plurality of target machine learning models is obtained.

Through the steps, a plurality of different machine learning models are fused into one PMML file, so that the model deployment efficiency is improved.

Optionally, in the method for processing an object provided in the embodiment of the present application, determining, according to a training set, a plurality of expert rules includes: obtaining a plurality of characteristic information corresponding to each sample object according to sample data information in the training set; calculating the information value of each piece of characteristic information to obtain the information value corresponding to each piece of characteristic information, and calculating the evidence weight of each piece of characteristic information to obtain the evidence weight corresponding to each piece of characteristic information; screening the plurality of characteristic information according to the information value and the evidence weight to obtain a plurality of target characteristic information; and determining a plurality of expert rules according to the target characteristic information.

In an alternative embodiment, the plurality of expert rules described above are determined using the steps of: firstly, carrying out feature extraction according to sample data information in a training set to obtain a plurality of feature information corresponding to each sample object. In an alternative embodiment, after obtaining the above multiple feature information, the feature information may be further preprocessed, for example, string conversion, feature culling, and missing value filling. The feature information after preprocessing is analyzed, for example, IV values (i.e., the above-mentioned value information) and WOE values (i.e., the above-mentioned evidence weights) are calculated. And then, feature screening is carried out according to the IV value and the WOE value, and then, a plurality of expert rules are determined according to the target feature information.

Through the steps, the accuracy of expert rules can be effectively improved.

Optionally, in the method for processing an object provided in the embodiment of the present application, after determining target category information corresponding to each first object according to the first initial category information and the second initial category information, the method further includes: and determining a plurality of second objects from the plurality of first objects according to the target category information, wherein the second objects are objects with the probability of purchasing the target telecommunication product being larger than a preset value.

Optionally, after obtaining the target class information of the first object, determining a plurality of second objects from the plurality of first objects through the target class information, namely selecting the first class object, and finally recommending the target telecommunication product to the second objects.

In an alternative embodiment, the above fusion model may be obtained using a schematic diagram as described in fig. 2:

s1, data acquisition;

s2, preprocessing data;

s3, data analysis;

s4, extracting expert rules;

s5, training a machine learning model;

s6, multi-model fusion, wherein VotingClassifier in Sklearn is used;

s7, expert rule fusion is carried out, and DummyClassifier in Sklearn is used;

S8, fusing the model and the rule, and using a SelectFirstClassifier in Sklearn2 pmml;

and S9, outputting a fusion model result.

According to the object processing method, a plurality of first objects and target data information of each first object are obtained, wherein the target data information at least comprises historical telecommunication product purchase information of the first objects and identity information of the first objects; predicting the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value; predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object; the target category information corresponding to each first object is determined according to the first initial category information and the second initial category information, so that the problem that the accuracy of classifying the users is low due to the fact that the users are classified manually in the related technology is solved. In the scheme, the voting classifier in the fusion model predicts the category of the first object by utilizing the historical telecom product purchase information and the identity information of the first object to obtain first initial category information, and predicts the category of the first object by utilizing the historical telecom product purchase information and the identity information of the first object by utilizing the virtual classifier in the fusion model to obtain second initial category information, finally obtains the target category information of the first object according to the first initial category information and the second initial category information, classifies the user by the fusion model, avoids subjectivity of manual classification, improves accuracy of classification of the user, predicts the category twice by the voting classifier and the virtual classifier, improves accuracy of classification of the user, and can more reasonably formulate a telecom product recommendation strategy by improving accuracy of classification of the user, thereby achieving the effect of improving the recommendation success rate of the telecom product.

It should be noted that the steps illustrated in the flowcharts of the figures may be performed in a computer system such as a set of computer executable instructions, and that although a logical order is illustrated in the flowcharts, in some cases the steps illustrated or described may be performed in an order other than that illustrated herein.

The embodiment of the application also provides an object processing device, and it should be noted that the object processing device of the embodiment of the application may be used to execute the object processing method provided by the embodiment of the application. The following describes a processing device for an object provided in an embodiment of the present application.

Fig. 3 is a schematic diagram of a processing apparatus of an object according to an embodiment of the present application. As shown in fig. 3, the apparatus includes: a first acquisition unit 301, a first prediction unit 302, a second prediction unit 303, and a first determination unit 304.

A first obtaining unit 301, configured to obtain a plurality of first objects and target data information of each first object, where the target data information includes at least historical telecommunication product purchase information of the first object and identity information of the first object;

the first prediction unit 302 is configured to predict, by using a vote classifier in the fusion model, a class of a first object based on the target data information, to obtain first initial class information corresponding to each first object, where the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with a probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with a probability of purchasing the target telecommunication product being less than or equal to the preset value;

A second prediction unit 303, configured to predict, by using a virtual classifier in the fusion model, a class of the first object based on the target data information, so as to obtain second initial class information corresponding to each first object;

the first determining unit 304 is configured to determine target category information corresponding to each first object according to the first initial category information and the second initial category information.

The object processing device provided in the embodiment of the present application acquires, through the first acquiring unit 301, a plurality of first objects and target data information of each first object, where the target data information includes at least historical telecommunication product purchase information of the first object and identity information of the first object; the first prediction unit 302 predicts the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing the target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value; the second prediction unit 303 predicts the category of the first object based on the target data information through a virtual classifier in the fusion model, and obtains second initial category information corresponding to each first object; the first determining unit 304 determines the target category information corresponding to each first object according to the first initial category information and the second initial category information, so that the problem that the accuracy of classifying the users is low due to the fact that the users are classified manually in the related art is solved. In the scheme, the voting classifier in the fusion model predicts the category of the first object by utilizing the historical telecom product purchase information and the identity information of the first object to obtain first initial category information, and predicts the category of the first object by utilizing the historical telecom product purchase information and the identity information of the first object by utilizing the virtual classifier in the fusion model to obtain second initial category information, finally obtains the target category information of the first object according to the first initial category information and the second initial category information, classifies the user by the fusion model, avoids subjectivity of manual classification, improves accuracy of classification of the user, predicts the category twice by the voting classifier and the virtual classifier, improves accuracy of classification of the user, and can more reasonably formulate a telecom product recommendation strategy by improving accuracy of classification of the user, thereby achieving the effect of improving the recommendation success rate of the telecom product.

Optionally, in the processing apparatus for an object provided in the embodiment of the present application, the first prediction unit includes: the extraction module is used for extracting the characteristics of the target data information through a plurality of target machine learning models in the voting classifier to obtain a characteristic vector corresponding to each first object; the prediction module is used for predicting the class of the first object based on the feature vector through each target machine learning model to obtain a first probability value of the first object belonging to the first class and a second probability value of the first object belonging to the second class output by each target machine learning model; the first calculation module is used for carrying out average calculation according to the first probability value to obtain a first average probability value of the first object belonging to the first category, and carrying out average calculation according to the second probability value to obtain a second average probability value of the first object belonging to the second category; the first determining module is used for obtaining first initial category information corresponding to each first object according to the first average probability value and the second average probability value.

Optionally, in the processing apparatus for an object provided in the embodiment of the present application, the second prediction unit includes: the matching module is used for matching the target data information through a plurality of expert rules in the virtual classifier to obtain a matching result; and the second determining module is used for obtaining second initial category information corresponding to each first object according to the matching result.

Optionally, in the processing apparatus for an object provided in the embodiment of the present application, the first determining unit includes: the judging module is used for judging whether the first initial category information and the second initial category information are the same or not; the third determining module is used for determining the priority between the voting classifier and the virtual classifier if the first initial category information is different from the second initial category information; a fourth determining module, configured to determine the first initial category information as target category information if the priority of the voting classifier is higher than the priority of the virtual classifier; and a fifth determining module, configured to determine the second initial category information as the target category information if the priority of the voting classifier is lower than the priority of the virtual classifier.

Optionally, in the processing apparatus for an object provided in the embodiment of the present application, the apparatus further includes: the second acquisition unit is used for acquiring a training set before predicting the class of the first object based on the target data information through the voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the training set at least comprises a plurality of sample objects, sample data information corresponding to each sample object and real class information corresponding to each sample object; the training unit is used for training the plurality of machine learning models according to the training set to obtain a plurality of target machine learning models, and carrying out model fusion on the plurality of target machine learning models in a voting classifier mode to obtain a voting classifier; the second determining unit is used for determining a plurality of expert rules according to the training set and obtaining a virtual classifier according to the expert rules; and the fusion unit is used for carrying out model fusion according to the voting classifier and the virtual classifier to obtain a fusion model.

Optionally, in the processing device for an object provided in the embodiment of the present application, the training unit includes: the construction module is used for constructing an initial voting classifier, inputting a plurality of target machine learning models into the initial voting classifier and obtaining a processed initial voting classifier; the setting module is used for setting the attributes of the processed initial voting classifier to obtain the voting classifier.

Optionally, in the processing apparatus for an object provided in the embodiment of the present application, the second determining unit includes: a sixth determining module, configured to obtain, according to sample data information in the training set, a plurality of feature information corresponding to each sample object; the second calculation module is used for calculating the information value of each piece of characteristic information to obtain the information value corresponding to each piece of characteristic information, and calculating the evidence weight of each piece of characteristic information to obtain the evidence weight corresponding to each piece of characteristic information; the screening module is used for screening the plurality of characteristic information according to the information value and the evidence weight to obtain a plurality of target characteristic information; and the seventh determining module is used for determining a plurality of expert rules according to the plurality of target characteristic information.

Optionally, in the processing apparatus for an object provided in the embodiment of the present application, the apparatus further includes: and the third determining unit is used for determining a plurality of second objects from the plurality of first objects according to the target category information after determining the target category information corresponding to each first object according to the first initial category information and the second initial category information, wherein the second objects are objects with the probability of purchasing the target telecommunication product being larger than a preset value.

The processing apparatus for an object includes a processor and a memory, and the first acquisition unit 301, the first prediction unit 302, the second prediction unit 303, the first determination unit 304, and the like described above are stored as program units in the memory, and the processor executes the program units stored in the memory to realize corresponding functions.

The processor includes a kernel, and the kernel fetches the corresponding program unit from the memory. The kernel can be provided with one or more than one, and accurate classification of users is realized by adjusting kernel parameters.

The memory may include volatile memory, random Access Memory (RAM), and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM), among other forms in computer readable media, the memory including at least one memory chip.

The embodiment of the invention provides a computer-readable storage medium having a program stored thereon, which when executed by a processor, implements a method of processing an object.

The embodiment of the invention provides a processor, which is used for running a program, wherein the program runs a processing method for executing an object.

As shown in fig. 4, an embodiment of the present invention provides an electronic device, where the device includes a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, the following steps are implemented: acquiring a plurality of first objects and target data information of each first object, wherein the target data information at least comprises historical telecommunication product purchase information of the first object and identity information of the first object; predicting the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value; predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object; and determining target category information corresponding to each first object according to the first initial category information and the second initial category information.

Optionally, predicting, by the voting classifier in the fusion model, the class of the first object based on the target data information, and obtaining first initial class information corresponding to each first object includes: extracting features of the target data information through a plurality of target machine learning models in the voting classifier to obtain feature vectors corresponding to each first object; predicting the class of the first object based on the feature vector through each target machine learning model to obtain a first probability value of the first object belonging to the first class and a second probability value of the first object belonging to the second class, which are output by each target machine learning model; average calculation is carried out according to the first probability value to obtain a first average probability value of the first object belonging to the first category, and average calculation is carried out according to the second probability value to obtain a second average probability value of the first object belonging to the second category; and obtaining first initial category information corresponding to each first object according to the first average probability value and the second average probability value.

Optionally, predicting the class of the first object based on the target data information through a virtual classifier in the fusion model, and obtaining second initial class information corresponding to each first object includes: matching the target data information through a plurality of expert rules in the virtual classifier to obtain a matching result; and obtaining second initial category information corresponding to each first object according to the matching result.

Optionally, determining the target category information corresponding to each first object according to the first initial category information and the second initial category information includes: judging whether the first initial category information and the second initial category information are the same or not; if the first initial category information is different from the second initial category information, determining the priority between the voting classifier and the virtual classifier; if the priority of the voting classifier is higher than that of the virtual classifier, determining the first initial class information as target class information; and if the priority of the voting classifier is lower than that of the virtual classifier, determining the second initial category information as target category information.

Optionally, before predicting the class of the first object based on the target data information by the voting classifier in the fusion model to obtain first initial class information corresponding to each first object, the method further includes: acquiring a training set, wherein the training set at least comprises a plurality of sample objects, sample data information corresponding to each sample object and real category information corresponding to each sample object; training a plurality of machine learning models according to the training set to obtain a plurality of target machine learning models, and carrying out model fusion on the plurality of target machine learning models in a voting classifier mode to obtain a voting classifier; determining a plurality of expert rules according to the training set, and obtaining a virtual classifier according to the plurality of expert rules; and carrying out model fusion according to the voting classifier and the virtual classifier to obtain a fusion model.

Optionally, performing model fusion on the plurality of target machine learning models by means of a voting classifier, and obtaining the voting classifier includes: constructing an initial voting classifier, and inputting a plurality of target machine learning models into the initial voting classifier to obtain a processed initial voting classifier; and setting the attribute of the processed initial voting classifier to obtain the voting classifier.

Optionally, determining the plurality of expert rules from the training set includes: obtaining a plurality of characteristic information corresponding to each sample object according to sample data information in the training set; calculating the information value of each piece of characteristic information to obtain the information value corresponding to each piece of characteristic information, and calculating the evidence weight of each piece of characteristic information to obtain the evidence weight corresponding to each piece of characteristic information; screening the plurality of characteristic information according to the information value and the evidence weight to obtain a plurality of target characteristic information; and determining a plurality of expert rules according to the target characteristic information.

Optionally, after determining the target category information corresponding to each first object according to the first initial category information and the second initial category information, the method further includes: and determining a plurality of second objects from the plurality of first objects according to the target category information, wherein the second objects are objects with the probability of purchasing the target telecommunication product being larger than a preset value.

The device herein may be a server, PC, PAD, cell phone, etc.

The present application also provides a computer program product adapted to perform, when executed on a data processing device, a program initialized with the method steps of: acquiring a plurality of first objects and target data information of each first object, wherein the target data information at least comprises historical telecommunication product purchase information of the first object and identity information of the first object; predicting the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value; predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object; and determining target category information corresponding to each first object according to the first initial category information and the second initial category information.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, etc., such as Read Only Memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises an element.

It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The foregoing is merely exemplary of the present application and is not intended to limit the present application. Various modifications and changes may be made to the present application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. which are within the spirit and principles of the present application are intended to be included within the scope of the claims of the present application.

Claims

1. A method of processing an object, comprising:

acquiring a plurality of first objects and target data information of each first object, wherein the target data information at least comprises historical telecommunication product purchase information of the first object and identity information of the first object;

predicting the class of the first object based on the target data information through a voting classifier in a fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value;

Predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object;

and determining target category information corresponding to each first object according to the first initial category information and the second initial category information.

2. The method of claim 1, wherein predicting the class of the first object based on the target data information by a vote classifier in a fusion model, the obtaining first initial class information corresponding to each first object comprises:

extracting features of the target data information through a plurality of target machine learning models in the voting classifier to obtain feature vectors corresponding to each first object;

predicting the class of the first object based on the feature vector through each target machine learning model to obtain a first probability value of the first object belonging to the first class and a second probability value of the first object belonging to the second class, which are output by each target machine learning model;

average calculation is carried out according to the first probability value to obtain a first average probability value of the first object belonging to the first category, and average calculation is carried out according to the second probability value to obtain a second average probability value of the first object belonging to the second category;

And obtaining the first initial category information corresponding to each first object according to the first average probability value and the second average probability value.

3. The method of claim 1, wherein predicting, by a virtual classifier in the fusion model, the class of the first object based on the target data information, obtaining second initial class information corresponding to each first object comprises:

matching the target data information through a plurality of expert rules in the virtual classifier to obtain a matching result;

and obtaining second initial category information corresponding to each first object according to the matching result.

4. The method of claim 1, wherein determining target class information for each first object based on the first initial class information and the second initial class information comprises:

judging whether the first initial category information and the second initial category information are the same or not;

if the first initial category information is different from the second initial category information, determining a priority between the voting classifier and the virtual classifier;

if the priority of the voting classifier is higher than that of the virtual classifier, determining the first initial class information as the target class information;

And if the priority of the voting classifier is lower than that of the virtual classifier, determining the second initial category information as the target category information.

5. The method of claim 1, wherein prior to predicting the class of the first object based on the target data information by a voting classifier in a fusion model, obtaining first initial class information corresponding to each first object, the method further comprises:

acquiring a training set, wherein the training set at least comprises a plurality of sample objects, sample data information corresponding to each sample object and real category information corresponding to each sample object;

training a plurality of machine learning models according to the training set to obtain a plurality of target machine learning models, and carrying out model fusion on the plurality of target machine learning models in a voting classifier mode to obtain the voting classifier;

determining a plurality of expert rules according to the training set, and obtaining the virtual classifier according to the expert rules;

and carrying out model fusion according to the voting classifier and the virtual classifier to obtain the fusion model.

6. The method of claim 5, wherein model fusing the plurality of target machine learning models by way of a voting classifier, the deriving the voting classifier comprising:

constructing an initial voting classifier, and inputting the multiple target machine learning models into the initial voting classifier to obtain a processed initial voting classifier;

and setting the attribute of the processed initial voting classifier to obtain the voting classifier.

7. The method of claim 5, wherein determining a plurality of expert rules from the training set comprises:

obtaining a plurality of characteristic information corresponding to each sample object according to the sample data information in the training set;

calculating the information value of each piece of characteristic information to obtain the information value corresponding to each piece of characteristic information, and calculating the evidence weight of each piece of characteristic information to obtain the evidence weight corresponding to each piece of characteristic information;

screening the plurality of characteristic information according to the information value and the evidence weight to obtain a plurality of target characteristic information;

and determining the expert rules according to the target feature information.

8. The method of claim 1, wherein after determining target class information corresponding to each first object from the first initial class information and the second initial class information, the method further comprises:

and determining a plurality of second objects from the plurality of first objects according to the target category information, wherein the second objects are objects with the probability of purchasing the target telecommunication product being larger than the preset value.

9. An apparatus for processing an object, comprising:

a first obtaining unit, configured to obtain a plurality of first objects and target data information of each first object, where the target data information includes at least historical telecommunication product purchase information of the first object and identity information of the first object;

the first prediction unit is used for predicting the class of the first object based on the target data information through a voting classifier in the fusion model to obtain first initial class information corresponding to each first object, wherein the class corresponding to the first object is a first class or a second class, the first object of the first class is an object with the probability of purchasing a target telecommunication product being greater than a preset value, and the first object of the second class is an object with the probability of purchasing the target telecommunication product being less than or equal to the preset value;

The second prediction unit is used for predicting the category of the first object based on the target data information through a virtual classifier in the fusion model to obtain second initial category information corresponding to each first object;

and the first determining unit is used for determining target category information corresponding to each first object according to the first initial category information and the second initial category information.

10. An electronic device comprising one or more processors and a memory for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the method of processing an object of any of claims 1-8.