CN115905655A - User portrait construction method, device and equipment and readable storage medium - Google Patents
User portrait construction method, device and equipment and readable storage medium Download PDFInfo
- Publication number
- CN115905655A CN115905655A CN202211398883.7A CN202211398883A CN115905655A CN 115905655 A CN115905655 A CN 115905655A CN 202211398883 A CN202211398883 A CN 202211398883A CN 115905655 A CN115905655 A CN 115905655A
- Authority
- CN
- China
- Prior art keywords
- data
- index
- model
- machine learning
- user
- 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.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application provides a user portrait construction method. When the method is executed, user data are obtained firstly, then the user data are analyzed respectively based on a hierarchical analysis model and a machine learning model to obtain analysis results, then grading values are obtained based on the analysis results according to preset weight coefficients, and finally grading results are obtained based on the grading values according to preset rules to achieve user portrait construction. Therefore, by using the method of fusing the analytic hierarchy process model and the machine learning model, the user portrait construction is enabled to quantitatively fuse expert experience, a large amount of data is utilized, and the effects of reasonably quantizing user data and constructing an objective user portrait are achieved. Therefore, the reasonability and the reliability of user portrait construction can be improved, and the efficiency of user portrait construction is improved.
Description
Technical Field
The present application relates to the field of data processing technologies, and in particular, to a user portrait construction method, apparatus, device, and readable storage medium.
Background
With the rapid development of the internet, the rapid iteration of technologies such as artificial intelligence algorithm, big data, cloud computing and the like, and the optimization and improvement of services by using information technology become more and more important. The expense reimbursement management is gradually converted from a manual auditing mode of a traditional mode to a financial sharing service mode, and in order to improve the employee reimbursement efficiency, a user portrait is constructed by adopting an employee reimbursement rating credit management mechanism at present.
At present, user portrait construction is generally customized by using rules, and the employee is subjected to reimbursement credit rating according to historical information by collecting relevant records of employee reimbursement activities. If the employee has records of reimbursement violations and the like, the employee is gradually degraded, a black and white list of the employee is set, and differentiated treatment is carried out on the aspects of business flow, audit range, frequency and the like, so that the employee is encouraged to carry out compliance reimbursement in reimbursement business activities, and the guarantee efficiency is improved.
In the implementation scheme of the user portrait construction: on the one hand, the business personnel of the financial sharing center score according to the historical information of the staff. However, personal experiences of different business personnel are different, subjective judgment is needed, and objective basis is lacked. On the other hand, because many reimbursement departments are involved, a great deal of manpower and material resources are needed to be spent in bill evaluation, so that the enterprise management cost is high, the reimbursement period is long, and the user experience is influenced.
Disclosure of Invention
In view of this, the present application provides a user portrait construction method, apparatus, device and readable storage medium, which are intended to improve the rationality and reliability of user portrait construction, improve the efficiency of user portrait construction, and provide a reasonable basis for user portrait construction.
In a first aspect, the present application provides a user portrait construction method, including:
acquiring user data;
analyzing the user data respectively based on a hierarchical analysis model and a machine learning model to obtain analysis results;
obtaining a score value based on the analysis result according to a preset distribution coefficient;
and obtaining a rating result based on the rating value according to a preset rule.
Optionally, the obtaining user data includes:
acquiring original data;
and preprocessing the original data to obtain processed standardized data.
Optionally, the analyzing the user data based on the hierarchical analysis model and the machine learning model to obtain an analysis result includes:
establishing an index system based on the data screening; the index system comprises a plurality of different indexes;
constructing each index judgment matrix according to expert scores;
carrying out consistency check on the matrix by utilizing an analytic hierarchy process and calculating each index weight coefficient;
and calculating the scores of all dimensions based on the index weight coefficients to obtain a hierarchical analysis model analysis result.
Optionally, the calculating the score of each dimension based on the index weight coefficients to obtain a hierarchical analysis model analysis result includes:
carrying out percentile standardization processing on each index to obtain a standardized index value;
obtaining an initial score through the weighted average of the standardized index value and the index weight;
performing min-max normalization based on the initial fraction to obtain a normalized numerical value;
and calculating the scores of all dimensions based on the normalized numerical values to obtain a hierarchical analysis model analysis result.
Optionally, the analyzing the data fields based on the hierarchical analysis model and the machine learning model to obtain analysis results respectively includes:
screening the in-mold features based on the data through the stability, correlation, collinearity and information content of variables of the features;
predicting data probabilities using a machine learning model based on the in-mold features;
and obtaining a machine learning model analysis result according to the probability transformation.
Optionally, before the screening of the modeled features based on the data, the stability, correlation, collinearity, and information content of the variables of the passing features includes:
adjusting parameters by Bayesian optimization;
the Bayesian optimization self-defined objective function is as follows:
T arg et＝offks+|offks-devks|*λ
wherein, offks is a parameter of the discrimination degree of good and bad samples in the training set, devks is a parameter of the discrimination degree of good and bad samples in the testing set, and λ is a weight coefficient.
Optionally, analyzing the user data based on the hierarchical analysis model to obtain an analysis result includes a hierarchical analysis model index, analyzing the user data based on the machine learning model to obtain an analysis result includes a machine learning model index, and obtaining a score value based on the analysis result according to a preset distribution coefficient includes:
obtaining a score value according to the level analysis model index and the machine learning model index
The score value S is calculated by the following formula:
s＝s 1 ×λ+s 2 ×(1-λ)
wherein, s is 1 For the hierarchical analysis model index, s 2 And the lambda is the distribution coefficient of the hierarchical analysis model in the overall score value.
In a second aspect, the present application provides a user representation construction apparatus, including:
the acquisition module is used for acquiring user data;
the analysis module is used for analyzing the user data respectively based on a hierarchical analysis model and a machine learning model to obtain analysis results;
a first determination module for obtaining a score value based on the analysis result according to a preset distribution coefficient;
and the second determining module is used for obtaining a rating result based on the rating value according to a preset rule.
In a third aspect, the present application provides an electronic device comprising a memory for storing instructions or code and a processor for executing the instructions or code to cause the device to perform the user representation construction method of any of the preceding first aspects.
In a fourth aspect, the present application provides a computer storage medium having code stored therein, wherein when the code is executed, an apparatus for executing the code implements the user representation construction method of any one of the first aspect.
The application provides a user portrait construction method. When the method is executed, user data are obtained firstly, then the user data are analyzed respectively based on a hierarchical analysis model and a machine learning model to obtain analysis results, then grading values are obtained based on the analysis results according to preset weight coefficients, and finally grading results are obtained based on the grading values according to preset rules to achieve user portrait construction. Therefore, by using the method of fusing the analytic hierarchy process model and the machine learning model, the user portrait construction not only quantificationally fuses expert experience, but also utilizes a large amount of data, and achieves the effects of reasonably quantifying user data and constructing objective user portrait. Therefore, the reasonability and the reliability of user portrait construction can be improved, and the efficiency of user portrait construction is improved.
Drawings
To illustrate the technical solutions in the present embodiment or the prior art more clearly, the drawings needed to be used in the description of the embodiment or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a flow chart of a method of a user representation construction method according to an embodiment of the present application;
FIG. 2 is a flowchart of an alternative implementation of step 102 provided by an embodiment of the present application;
FIG. 3 is a flowchart of another alternative implementation of step 102 provided by embodiments of the present application;
FIG. 4 is a schematic structural diagram of a user representation creation apparatus according to an embodiment of the present disclosure.
Detailed Description
As previously described, user portrayal construction is typically customized using rules, such as a credit rating by collecting a record of the user's activities and rating the user's credit based on historical information. Namely, the user is classified into a black list and a white list according to the historical audit record. The audit record is mainly the authenticity and compliance record of the user historical data, wherein authenticity refers to whether the related data is reasonable legal cost, and compliance refers to whether the related processes are used correctly. If the user has records of violation and the like, the user is gradually degraded, a black-and-white list of the user is set, and the user is treated differently in the aspects of business process, audit range, frequency and the like, so that the user is encouraged to reasonably perform operation in business activities, and the guarantee efficiency is improved.
Research shows that for user portrait construction, by using a method of fusing an analytic hierarchy process model and a machine learning model, the user portrait construction quantificationally fuses expert experience, and a large amount of data is utilized, so that a reasonable basis can be provided for the user portrait construction.
In view of the above, the present application provides a user portrait construction method. When the method is executed, user data are obtained firstly, then the user data are analyzed respectively based on a hierarchical analysis model and a machine learning model to obtain analysis results, then grading values are obtained based on the analysis results according to preset weight coefficients, and finally grading results are obtained based on the grading values according to preset rules to achieve user portrait construction. Therefore, by using the method of fusing the analytic hierarchy process model and the machine learning model, the user portrait construction is enabled to quantitatively fuse expert experience, a large amount of data is utilized, and the effects of reasonably quantizing user data and constructing an objective user portrait are achieved. Therefore, the reasonability and the reliability of user portrait construction can be improved, and the efficiency of user portrait construction is improved.
In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Referring to fig. 1, fig. 1 is a flowchart of a method of an image constructing method according to an embodiment of the present application, and with reference to fig. 1, the abnormal attribution analyzing method according to the embodiment of the present application may include:
s101: user data is acquired.
In this embodiment, the historical data of the user is obtained, and taking a financial reimbursement scene as an example, the data of the user, such as the passing behavior, the policy returning behavior, the image returning behavior, the repayment timeliness, the process familiarity, the cancel process, the quality of the bill of delivery, and the like, can be obtained. Specifically, in the process of acquiring data, the data may be acquired in different time windows according to requirements, for example, 1 month, 3 months, 6 months, 12 months or data from history, and the user data information under a specific condition may be acquired based on different limiting conditions, such as department, age, post, and the like, and in this application, the specific acquisition manner is not limited.
Optionally, raw data is obtained, which includes historical data of the user, which is stored in a data repository and can be retrieved and recalled at any time. Specifically, taking financial reimbursement user portrait construction as an example, user data related to user portrait construction is acquired. Optionally, the data includes five-dimensional features including financial auditing, personnel relationships, post-hoc spot checks, risk anomalies, and management applications. Each dimension has a plurality of influence factors, including a passing behavior, a receipt-returning behavior, an image-returning behavior, a repayment timeliness, a flow familiarity, a flow cancellation, a traffic order quality, a behavior exception, a risk characteristic, a post-quality inspection, a post-mutual inspection and the like. For example: the financial auditing dimension specifically comprises the influence factors such as passing behavior, bill returning behavior, image returning behavior, repayment timeliness, flow familiarity, flow canceling, bill delivery quality and the like; each influence factor has corresponding basic and derivative fields, for example, the action of returning the bill specifically may include the maximum amount of the reimbursement bill, the one-time passing rate of the bill, the number of the reason and the like of returning the bill.
Optionally, the raw data is preprocessed to obtain processed standardized data, that is, the raw data is extracted from the data center, the raw data is collated into standardized data through calculation, screening and cleaning, a data field capable of being imported into the model is formed, and the collated standardized data is stored in the data center. Specifically, the acquired user data is preprocessed, for example, a unified data table is established through data inspection such as data alignment and fusion ratio peer-to-peer methods, standardized data is established for data under a certain influence factor, data among different data sources are fused, data with a complementary relationship are aligned and fused, and data with a redundant relationship are deduplicated; and further, carrying out multi-dimensional multi-data sorting to obtain processed standardized data.
S102: and analyzing the user data respectively based on a hierarchical analysis model and a machine learning model to obtain analysis results.
In this embodiment, a hierarchical analysis model and a machine learning model are called to analyze the data acquired in the above steps, respectively.
The Analytic Hierarchy Process (AHP) is a systematic and hierarchical Analytic method combining qualitative analysis and quantitative analysis, and refers to a decision-making method for performing qualitative and quantitative analysis based on the decomposition of elements always related to decision-making into levels such as targets, criteria, schemes, and the like. Specifically, the analytic hierarchy process is a systematic method which takes a complex multi-target decision problem as a system, decomposes a target into a plurality of targets or criteria, further decomposes the targets into a plurality of layers of multi-index (or criteria, constraint), and calculates the single-layer ordering (weight) and the total ordering of the layers by a qualitative index fuzzy quantization method to be taken as the target (multi-index) and multi-scheme optimization decision. That is, the analytic hierarchy process decomposes the decision problem into different hierarchical structures according to the sequence of the total target, sub targets of each layer, evaluation criteria to the specific backup delivery scheme, then uses the method of solving and judging the characteristic vector of the matrix to obtain the priority weight of each element of each layer to a certain element of the previous layer, and finally uses the method of weighted sum to hierarchically merge the final weight of each backup scheme to the total target, and the maximum final weight is the optimal scheme. Taking a financial reimbursement scenario as an example, the analytic hierarchy process decomposes a required decision event into a target layer (e.g., reimburser), a criterion layer (i.e., factors influencing the decision, such as characteristics of reimbursement, risk, quality of delivery, etc.), and a scheme layer (i.e., a scheme, such as how to identify a high-quality reimburser, find credit blacklist personnel, etc.), and combines scores of digital financial auditing experts on each influencing factor, and after performing weighted average according to the result of consistency check, the final weight of each factor is obtained.
A machine learning model refers to a model in which an algorithm is trained to classify or predict by using a statistical method. Machine learning can be specifically classified into supervised machine learning, unsupervised machine learning, and semi-supervised learning. The application mainly uses supervised learning, and it can be understood that the supervised machine learning uses a tagged data set training algorithm to accurately classify data or predict results. Such as logistic regression, XGBOOST (eXtreme Gradient Boosting), lightGBM (Light Gradient Boosting Machine), etc., preferably the XGBOOST model is used in the present application.
And analyzing the data respectively through a hierarchical analysis model and a machine learning model to obtain an analysis result.
S103: and obtaining a score value based on the analysis result according to a preset distribution coefficient.
In this embodiment, since the hierarchical analysis model and the machine learning model respectively analyze data to obtain an analysis result, and the analysis result is quantitatively calculated according to a preset distribution coefficient to obtain a score value, the distribution coefficient may be determined according to a requirement, for example, the distribution coefficient may be a distribution coefficient of the hierarchical analysis model index occupying the overall credit score, for example, 50%.
In a specific embodiment, the analysis result obtained by analyzing the user data based on the hierarchical analysis model includes a hierarchical analysis model index, and the analysis result obtained by analyzing the user data based on the machine learning model includes a machine learning model index, it is understood that the hierarchical analysis model index and the machine learning model index may be a specific score, such as 300 to 900 scores, and a specific obtaining manner of the index may be obtained by specifically adjusting the model. And then, calculating according to the hierarchical analysis model index and the machine learning model index to obtain a score value.
The score value S is calculated by the following formula:
s＝s 1 ×λ+s 2 ×(1-λ)
wherein, said s 1 For the hierarchical analysis model index, s 2 And the lambda is the distribution coefficient of the hierarchical analysis model in the overall score value.
It can be understood that business personnel can modify the distribution coefficients according to actual needs, so that the result is more suitable for business requirements. In the early stage of system operation, the distribution coefficient value can be adjusted to be larger by a proper amount, and the business experience is taken as the leading factor, such as 50% or 60%; after the system runs for a certain time and accumulates enough data and experiences, the distribution coefficient value is gradually reduced, for example, 30%, so that the dominant proportion of the machine learning model is increased.
S104: and obtaining a rating result based on the rating value according to a preset rule.
In this embodiment, according to the above-mentioned comprehensive score value, the users can be classified into different grades according to the score value, for example: s, A, B, C grade. It will be appreciated that the rules for converting a particular score into a rating may be selected at the discretion of the business person, such as rating the top 5% of the total score as S. The user portrait construction can help business personnel to manage users in a grading mode, different authorities of users in different grades can be appointed, and for example, financial reimbursement scenes are taken as examples, reimbursers in S grades are taken as trial-free candidates, reimbursers in C grades are supervised more strictly, and the like, so that bill circulation efficiency is improved, cost is reduced, efficiency is improved, and user experience is improved.
In the embodiment, by using the method of fusing the analytic hierarchy process model and the machine learning model, the user portrait construction not only quantificationally fuses expert experience, but also utilizes a large amount of data, and the effects of reasonably quantifying user data and constructing an objective user portrait are achieved. Therefore, the reasonability and the reliability of user portrait construction can be improved, and the efficiency of user portrait construction is improved.
Referring to fig. 2, fig. 2 is a flowchart of a possible manner of analyzing the user data based on a hierarchical analysis model and a machine learning model to obtain an analysis result according to an embodiment of the present disclosure, and with reference to fig. 2, a user portrait construction method according to an embodiment of the present disclosure may include:
s201: establishing an index system based on the data screening; the index system includes a plurality of different indices.
In the embodiment, an index system for constructing the user portrait is constructed based on the acquired user data, and the data for constructing the index mainly comes from historical data; for example, taking a financial reimbursement scene as an example, data such as passage behavior, bill refunding behavior, image refunding behavior, repayment timeliness, process familiarity, process cancellation, quality of delivery, behavior abnormality, risk characteristics, post-quality inspection, and post-mutual inspection are determined as each index of the index system.
S202: and constructing the index judgment matrix according to the expert scores.
In the embodiment, questionnaire results of the importance of multiple experts on the index are obtained, so that a judgment matrix is constructed by combining the expert experience.
S203: and carrying out consistency check on the matrix by utilizing an analytic hierarchy process and calculating each index weight coefficient.
In the present embodiment, the matrix is subjected to consistency check, and data that cannot pass the consistency check is eliminated. Then, the weight coefficient of each index is calculated by matrix operation, and the specific calculation method is not limited in the application.
S204: and calculating the scores of all dimensions based on the index weight coefficients to obtain a hierarchical analysis model analysis result.
In this embodiment, the score of each dimension is calculated based on each index weight coefficient, and a hierarchical analysis model analysis result is obtained.
In a specific embodiment, carrying out percentile standardization on each index to obtain a standardized index value; specifically, the index values are sorted in ascending order, and the normalization processing is performed on the basis of a standard value, for example, the standard value is 100, and the index value of the first 20% normalization processing is 20, for the factor values with the ranking percentage higher than 20% or the ranking percentage lower than 30%. And then obtaining an initial score through the weighted average of the standardized index values of the indexes and the index weights, namely summing after the products of the standardized index values and the index weights to obtain the initial score. And then carrying out min-max normalization on the initial fraction to obtain a normalized value, wherein the min-max normalization is also called a range method, and is a linear transformation on the original data to map the original data to the range of [0-1 ]. By using the method, the numerical value is controlled to be 0-1 so as to reasonably calculate the grade of each dimension based on the normalized numerical value to obtain the analysis result of the hierarchical analysis model, and specifically, the value is controlled to be 300-900 to obtain the index value of the hierarchical analysis model, namely the hierarchical analysis model.
Referring to fig. 3, fig. 3 is a flowchart of another possible manner of analyzing the user data based on a hierarchical analysis model and a machine learning model to obtain an analysis result according to an embodiment of the present disclosure, and with reference to fig. 3, a user portrait construction method according to an embodiment of the present disclosure may include:
s301: screening the modelled features based on the data by their stability, correlation, collinearity, and information content of the variables.
In this embodiment, XGBOOST modeling is preferably used in the present application, and in the modeling process, not only feature screening is performed according to the importance of features, but also a method of combining multiple feature screening is integrated. And performing primary feature screening through the stability, correlation, collinearity and variable information quantity of the features, and removing partial redundant features.
Optionally, in the modeling process, the business importance of the model entering feature and the stability of the model entering feature are combined, dynamic balance adjustment is performed, namely, the model entering feature is manually adjusted according to the output result of the model, the number of the model entering features is reduced as much as possible, data cost is saved, psi (model distribution stability) of the model is in a controllable range, and the model stability is ensured while the sample has the optimal discrimination.
Optionally, during parameter adjustment of the algorithm, a bayesian optimization method is used for parameter adjustment. Bayesian parameter adjustment adopts a Gaussian process, historical parameter information is considered, prior information is continuously updated, and training speed is improved. Taking a financial reimbursement scene as an example, due to the fact that the sample size is large and the number of features is large (more than three hundred features), the model training speed is slow when other parameter adjusting methods such as a grid search method are used, and therefore in order to improve the model training speed, the optimization method of Bayes parameter adjustment is adopted in the application. The bayesian optimization can define an objective function, and the following objective optimization functions are defined in the embodiment:
T arg et＝offks+|offks-devks|*λ
wherein, offks is a parameter of the discrimination degree of good and bad samples in the training set, devks is a parameter of the discrimination degree of good and bad samples in the testing set, and λ is a weight coefficient.
The target optimization function can strictly control the effect difference of the model on a training set and a testing set, so that the model achieves optimal generalization.
S302: and predicting the data probability by utilizing a machine learning model based on the model-entering characteristics.
In the embodiment, according to the model-entering characteristics, the machine learning model predicts the probability of possible abnormality of the next data by using the model-entering data. Taking a financial reimbursement scene as an example, the data information of the reimburser is brought into the model, so that the probability of abnormity of the next reimbursement can be obtained, and the probability is converted into the machine learning model score by using the linear part conversion score. Optionally, because each sub-data module has different reimbursement processes, personnel reimbursement familiarity, data storage and the like, a respective modeling mode is adopted for each sub-data module, wherein the retained important variables are different from each other, and the maximization of the model effect is ensured.
S303: and obtaining a machine learning model analysis result according to the probability transformation.
In this embodiment, the machine learning model analysis result is obtained through probability transformation, specifically, the probability predicted by the machine model according to the model-entering features is transformed into a specific score, which may be, for example, a score value of 0 to 100, or the probability score may be transformed into a score value of 300 to 900 as required, and the score value is used as the machine learning model analysis result.
Optionally, in addition to obtaining the rating result, in a specific application scenario, taking a financial reimbursement user portrait as an example, the application may also display credit-related information of the reimburser in a five-dimensional radar graph.
For a specific reimburser, the data completed by the reimburser's behavior of returning orders, behavior of returning images, post-quality inspection, post-mutual inspection and the like are subjected to statistical analysis by the method, and a five-dimensional radar chart is automatically output. Each vertex of the radar map is a feature of five dimensions, namely financial audit, personnel relationship, post-casual inspection, risk abnormity and management application, wherein the financial audit specifically comprises a passing behavior, a receipt behavior, an image rejection behavior, a repayment timeliness, a flow familiarity, a flow cancellation and a transaction quality; the human relationship includes: staff and personnel characteristics; the post-hoc spot check comprises the following steps: post-incident quality inspection and post-incident mutual inspection; the risk anomalies include: risk features and abnormal behavior; the management and operation comprises the following steps: the credit base table is supplemented.
The specific score for each dimension is calculated as follows:
wherein m is a dimension serial number, i is an influence factor serial number below each dimension, j is a field serial number below each influence factor, and k m Number of influencing factors of the mth dimension, k n For the number of fields, w, contained under the ith influencing factor of the mth dimension i To influence the factor weights, consistent with the results in the analytic hierarchy process. Specific score s for each dimension Dimension m By the respective influence factor score s in that dimension Influencing factor i Multiplied by a weight w i Calculating to obtain; fraction s of influence factor Influencing factor i Ranking percentile p of overall reimbursers by calculating factor of field value under influence factor Field j And (6) obtaining. If the field is a positive impact on the dimension score, such as process familiarity, pass behavior, etc., the percentiles are found in order; if the field is a negative impact on the dimension score, such as a drop-out behavior, etc., then the percentile is given by the reverse order percentage. For example, the number of times that an invoice is returned for a seller within one year is 6, and after sorting the field data of all the sellers, the number of returned bills of the seller exceeds 68%, and the factor percentile of the negatively affected field is 32%.
In the present embodiment, the score s of each dimension is calculated Dimension m Marking in the radar chart, the reimbursement history information of the reimburser can be clearly and objectively visualized.
The foregoing provides some specific implementation manners of the user portrait construction method for the embodiment of the present application, and based on this, the present application also provides a corresponding apparatus. The device provided by the embodiment of the present application will be described in terms of functional modularity.
Referring to FIG. 4, a schematic diagram of a user representation construction apparatus 400 is shown, the apparatus 400 includes an obtaining module 401, an analyzing module 402, a first determining module 403, and a second determining module 404.
An obtaining module 401, configured to obtain user data.
An analysis module 402, configured to analyze the user data based on a hierarchical analysis model and a machine learning model respectively to obtain analysis results.
A first determining module 403, configured to obtain a score value based on the analysis result according to a preset weighting factor.
And a second determining module 404, configured to obtain a rating result based on the rating value according to a preset rule.
The obtaining module 401 includes: an acquisition unit 4011 and a processing unit 4012.
An acquisition unit 4011 configured to acquire raw data;
and the processing unit 4012 is configured to perform preprocessing on the raw data to obtain processed normalized data.
The analysis module comprises: a creating unit 4021, a constructing unit 4022, a check calculating unit 4023, and a first determining unit 4024.
The establishing unit 4021 is configured to establish an index system based on the data screening; the index system includes a plurality of different indices.
The constructing unit 4022 is configured to construct the index determination matrices according to the expert scores.
And the inspection calculating unit 4023 is configured to perform consistency inspection on the matrix by using an analytic hierarchy process and calculate each index weight coefficient.
The first determining unit 4024 is configured to calculate a score of each dimension based on each index weight coefficient, so as to obtain a hierarchical analysis model analysis result.
The analysis module includes: a screening unit 4021a, a prediction unit 4022a, and a conversion unit 4023a.
A screening unit 4021a for screening the in-mode features based on the data by the stability, correlation, collinearity of the features and the information amount of the variables;
a prediction unit 4022a, configured to predict a data probability by using a machine learning model based on the mode-entry features;
and the conversion unit 4023a is used for obtaining a machine learning model analysis result according to the probability conversion.
The embodiment of the application also provides corresponding equipment and a computer storage medium, which are used for realizing the scheme provided by the embodiment of the application.
Wherein the device comprises a memory for storing instructions or codes and a processor for executing the instructions or codes to make the device execute the user portrait construction method according to any embodiment of the present application.
The computer storage medium has code stored therein, and when the code is executed, a device running the code implements the user representation construction method according to any embodiment of the present application.
In the embodiments of the present application, the names "first" and "second" (if present) in the names "first" and "second" are used for name identification, and do not represent the first and second in sequence.
As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that all or part of the steps in the method of the above embodiments may be implemented by software plus a general hardware platform. Based on such understanding, the technical solution of the present application may be embodied in the form of a software product, which may be stored in a storage medium, such as a read-only memory (ROM)/RAM, a magnetic disk, an optical disk, or the like, and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network communication device such as a router) to execute the method according to the embodiments or some parts of the embodiments of the present application.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus embodiment, since it is substantially similar to the method embodiment, it is relatively simple to describe, and reference may be made to some descriptions of the method embodiment for relevant points. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
The above description is only an exemplary embodiment of the present application, and is not intended to limit the scope of the present application.
Claims (10)
1. A user portrait construction method, comprising:
acquiring user data;
analyzing the user data respectively based on a hierarchical analysis model and a machine learning model to obtain analysis results;
obtaining a score value based on the analysis result according to a preset distribution coefficient;
and obtaining a rating result based on the rating value according to a preset rule.
2. The method of claim 1, wherein the obtaining data comprises:
acquiring original data;
and preprocessing the original data to obtain processed standardized data.
3. The method of claim 1, wherein analyzing the user data based on the hierarchical analysis model and the machine learning model respectively to obtain an analysis result comprises:
establishing an index system based on the data screening; the index system comprises a plurality of different indexes;
constructing each index judgment matrix according to expert scores;
carrying out consistency check on the matrix by utilizing an analytic hierarchy process and calculating each index weight coefficient;
and calculating the scores of all dimensions based on the index weight coefficients to obtain a hierarchical analysis model analysis result.
4. The method of claim 3, wherein the calculating the score of each dimension based on the index weight coefficients to obtain the analysis result of the hierarchical analysis model comprises:
carrying out percentile standardization processing on each index to obtain a standardized index value;
obtaining an initial score through the weighted average of the standardized index value and the index weight;
performing min-max normalization based on the initial fraction to obtain a normalized numerical value;
and calculating the scores of all dimensions based on the normalized numerical values to obtain a hierarchical analysis model analysis result.
5. The method of claim 1, wherein analyzing the user data based on the hierarchical analysis model and the machine learning model respectively to obtain an analysis result comprises:
screening the in-mold features based on the data through the stability, correlation, collinearity and information content of variables of the features;
predicting data probabilities using a machine learning model based on the in-mold features;
and obtaining a machine learning model analysis result according to the probability transformation.
6. The method of claim 5, wherein the passing feature stability, correlation, collinearity, and variable information content comprises, prior to screening the modeled features based on the data:
adjusting parameters by Bayesian optimization;
the Bayesian optimization self-defined objective function is as follows:
T arg et＝offks+|offks-devks|*λ
wherein, offks is a parameter of the discrimination degree of good and bad samples in the training set, devks is a parameter of the discrimination degree of good and bad samples in the testing set, and λ is a weight coefficient.
7. The method of claim 1, wherein analyzing the user data based on the hierarchical analysis model to obtain an analysis result comprises a hierarchical analysis model index, analyzing the user data based on the machine learning model to obtain an analysis result comprises a machine learning model index, and obtaining a score value based on the analysis result according to a preset distribution coefficient comprises:
obtaining a score value according to the level analysis model index and the machine learning model index
The score value S is calculated by the following formula:
s＝s 1 ×λ+s 2 ×(1-λ)
wherein, said s 1 For the hierarchical analysis model index, s 2 And the lambda is the distribution coefficient of the hierarchical analysis model in the overall score value.
8. A user representation construction apparatus, comprising:
the acquisition module is used for acquiring user data;
the analysis module is used for analyzing the user data respectively based on a hierarchical analysis model and a machine learning model to obtain analysis results;
a first determination module for obtaining a score value based on the analysis result according to a preset distribution coefficient;
and the second determining module is used for obtaining a rating result based on the rating value according to a preset rule.
9. An electronic device, wherein the device comprises a memory for storing instructions or code and a processor for executing the instructions or code to cause the device to perform the user representation construction method of any of claims 1 to 7.
10. A computer storage medium having code stored therein, the code, when executed, causing an apparatus running the code to implement a user representation construction method as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211398883.7A CN115905655A (en) | 2022-11-09 | 2022-11-09 | User portrait construction method, device and equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211398883.7A CN115905655A (en) | 2022-11-09 | 2022-11-09 | User portrait construction method, device and equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115905655A true CN115905655A (en) | 2023-04-04 |
Family
ID=86494917
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211398883.7A Pending CN115905655A (en) | 2022-11-09 | 2022-11-09 | User portrait construction method, device and equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115905655A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116956097A (en) * | 2023-09-18 | 2023-10-27 | 湖南华菱电子商务有限公司 | Expert portrait analysis method and system based on K-means |
-
2022
- 2022-11-09 CN CN202211398883.7A patent/CN115905655A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116956097A (en) * | 2023-09-18 | 2023-10-27 | 湖南华菱电子商务有限公司 | Expert portrait analysis method and system based on K-means |
| CN116956097B (en) * | 2023-09-18 | 2023-12-12 | 湖南华菱电子商务有限公司 | Expert portrait analysis method and system based on K-means |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110009479B (en) | Credit evaluation method and device, storage medium and computer equipment | |
| CN112668822B (en) | Scientific and technological achievement transformation platform sharing system, method, storage medium and mobile phone APP | |
| CN113537807B (en) | Intelligent wind control method and equipment for enterprises | |
| CN110866832A (en) | Risk control method, system, storage medium and computing device | |
| CN111192133A (en) | Method and device for generating risk model after user loan and electronic equipment | |
| CN115905655A (en) | User portrait construction method, device and equipment and readable storage medium | |
| CN111626855A (en) | Bond credit interest difference prediction method and system | |
| CN115760332A (en) | Risk prediction method, system, medium and device based on enterprise data analysis | |
| CN114493686A (en) | Operation content generation and pushing method and device | |
| CN116911994B (en) | External trade risk early warning system | |
| CN113283673A (en) | Model performance attenuation evaluation method, model training method and device | |
| CN116091206B (en) | Credit evaluation method, credit evaluation device, electronic equipment and storage medium | |
| CN111738610A (en) | Public opinion data-based enterprise loss risk early warning system and method | |
| CN117132383A (en) | Credit data processing method, device, equipment and readable storage medium | |
| CN116468273A (en) | Customer risk identification method and device | |
| CN115860924A (en) | Supply chain financial credit risk early warning method and related equipment | |
| Garcia de Alford et al. | Reducing age bias in machine learning: An algorithmic approach | |
| KR102499182B1 (en) | Loan regular auditing system using artificia intellicence | |
| CN117455681A (en) | Service risk prediction method and device | |
| CN113592292A (en) | Power grid primary equipment technical improvement overhaul project admission evaluation method | |
| Lakshmi et al. | Machine learning approach for taxation analysis using classification techniques | |
| Terzi et al. | Comparison of financial distress prediction models: Evidence from turkey | |
| CN117726434A (en) | Credit scoring card model training method, application method and related products | |
| Pusapati | Loan Under writing prediction using Deep learning techniques | |
| Wang | Data-driven Investment Decisions in P2P Lending: Strategies of Integrating Credit Scoring and Profit Scoring |
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 |