CN111340375A - Electricity charge recycling risk prediction method and device, electronic equipment and storage medium - Google Patents
Electricity charge recycling risk prediction method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN111340375A CN111340375A CN202010128050.3A CN202010128050A CN111340375A CN 111340375 A CN111340375 A CN 111340375A CN 202010128050 A CN202010128050 A CN 202010128050A CN 111340375 A CN111340375 A CN 111340375A
- Authority
- CN
- China
- Prior art keywords
- data
- user
- payment
- continuous
- characteristic parameter
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The application relates to a method and a device for predicting electric charge recycling risk, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring continuous payment record data of a user to be predicted within a preset time period; obtaining a plurality of pieces of original data which are continuous in time based on the continuous payment record data, wherein each piece of original data comprises data related to payment under various dimensions; obtaining multiple kinds of characteristic parameter data under different dimensions based on the multiple pieces of original data; and determining the electric charge recycling risk of the user to be predicted based on the pre-trained anti-noise neural network model and the multiple characteristic parameter data. In the embodiment of the application, the anti-noise neural network model is adopted to predict the electric charge recovery risk of the user to be predicted, so that the anti-noise neural network model has better anti-interference performance on noise data, and can not be interfered by errors in the electric power data labeling process any more, thereby improving the prediction accuracy.
Description
Technical Field
The application belongs to the technical field of artificial intelligence, and particularly relates to a method and a device for predicting electric charge recycling risk, electronic equipment and a storage medium.
Background
The recovery of the electric charge is always a difficult problem for power supply enterprises, and because the current operation means is to transmit electricity and then pay the fee, various hidden dangers that the charging period consumes time and labor and the recovery cannot be performed exist. The result of the electric charge recovery is closely related to the operation result of the power supply enterprise, and the electric charge recovery is the key content of electric power marketing all the time. The power supply enterprise has a huge number of customers, but the credit worthiness of each user is greatly different, and the credit worthiness of the users seriously affects the payment condition of the electric charge.
At present, various administrative management methods and technical means are also provided for the power supply enterprises aiming at the problems, and corresponding recovery strategies are provided. The current collection urging means is mainly manual collection urging, and manual strategies such as telephone reminding, home entry note attaching, on-the-spot charging and the like are adopted. Meanwhile, with the rapid development of information technology, domestic power supply enterprises also try to use the method of mining and analyzing mass data of power utilization users, stripping relevant characteristics of the power utilization users, establishing a user portrait and a classification system for the users, and trying to achieve risk early warning so as to judge whether the users have the possibility of defaulting power charges. If the logistic regression model is applied, modeling is carried out on the influence degree of the key electricity utilization factors of the high-voltage users on the electricity charge recovery; establishing a risk identification model of customer arrearage by utilizing a decision tree algorithm; and predicting the user electricity arrearage risk by improving an LR-Bagging algorithm based on theoretical analysis.
The method of adopting the manual pre-collection not only wastes a large amount of time and money, but also has potential safety hazard due to uncontrollable factors such as human factors, the manager is difficult to monitor and manage, the collection-collection return rate is low, and advance prevention and risk early warning cannot be realized. The existing prediction model also has disadvantages, because models such as a decision tree, logistic regression and the like belong to supervised models, training data needs to be labeled, the quality of the label influences the prediction accuracy, once errors exist in the data labeling process, the effect of the whole model is greatly reduced, and the performance of the model is damaged; meanwhile, because the formation of the power data is complex, power users have different types, power price execution standards and payment modes, so that the data dimension is large and nonlinear, and the nonlinear relation is difficult to fit by a common machine learning algorithm.
Disclosure of Invention
In view of this, embodiments of the present application provide an electric charge recycling risk prediction method, an apparatus, an electronic device, and a storage medium, so as to solve the problem that the existing prediction model is greatly affected by a tag, which results in weak anti-noise capability.
The embodiment of the application is realized as follows:
in a first aspect, an embodiment of the present application provides an electric charge recycling risk prediction method, including: acquiring continuous payment record data of a user to be predicted within a preset time period; obtaining a plurality of pieces of original data which are continuous in time based on the continuous payment record data, wherein each piece of original data comprises data related to payment under various dimensions; obtaining multiple kinds of characteristic parameter data under different dimensions based on the multiple pieces of original data; and determining the electric charge recovery risk of the user to be predicted based on a pre-trained anti-noise neural network model and the multiple characteristic parameter data, wherein a loss function of the anti-noise neural network model is mutual information entropy DMI of a determinant of a joint distribution matrix based on two discrete random variables. In the embodiment of the application, mutual information entropy DMI of a determinant based on a joint distribution matrix of two discrete random variables is used as a loss function of an anti-noise neural network model to predict the electricity charge recovery risk of a user to be predicted, and DMI is an absolute value of the determinant (det) of the joint distribution matrix of the two random discrete variables with the same value range. Thus DMI has the following properties: DMI is nonnegative, symmetrical, and meets the requirement of information monotonicity, and simultaneously meets the requirement of relative invariance, so that the DMI has better anti-interference performance on noise data, and can not be interfered by errors in the process of marking electric power data, thereby improving the accuracy of prediction.
With reference to one possible implementation manner of the embodiment of the first aspect, the loss function is: l isDMI(Qh(X),Y′)＝-log(DMI(h(x),Y′))＝-log(|det(Qh(X),Y′) L) wherein，Qh(X),Y′Is a joint distribution of a discrete random variable h (X) and a discrete random variable Y', Qh(X),Y′Is in the form of a matrix of c × c, and the randomness of h (X) is derived from h and a random variable X.
With reference to one possible implementation manner of the embodiment of the first aspect, after obtaining multiple kinds of feature parameter data in different dimensions, the method further includes: respectively carrying out characteristic normalization processing on each characteristic parameter data in the multiple characteristic parameter data; correspondingly, determining the electric charge recycling risk of the user to be predicted based on the anti-noise neural network model trained in advance and the various characteristic parameter data, and the method comprises the following steps: and determining the electric charge recycling risk of the user to be predicted based on the anti-noise neural network model trained in advance and the various characteristic parameter data subjected to normalization processing. In the embodiment of the application, each characteristic parameter data in the various characteristic parameter data is subjected to characteristic normalization processing, and then the electricity charge recovery risk of the user to be predicted is predicted based on the various characteristic parameter data after normalization processing, so that prediction errors caused by different values of the characteristic parameter data in different dimensions in the various characteristic parameter data are reduced.
With reference to a possible implementation manner of the embodiment of the first aspect, the acquiring continuous payment record data of the user to be predicted in a preset time period includes: acquiring continuous payment record data of the user to be predicted within two years; correspondingly, based on the continuous payment record data, obtaining a plurality of pieces of original data which are continuous in time, wherein the original data comprise: based on the continuous payment record data, twelve continuous original data in time are obtained at intervals of two months, wherein each piece of original data comprises: the method comprises the following steps of (1) user number, account opening date, charge default fee, charge fee, power consumption, charge date and charge mode; correspondingly, based on the plurality of pieces of original data, obtaining a plurality of kinds of feature parameter data under different dimensions, including: based on the plurality of pieces of original data, user numbers, account opening years, comprehensive scores of payment behaviors, predicted payment time periods, power consumption electric quantity fluctuation conditions, the most recent electric quantity cyclic ratio, payment mode change times, arrearage times, payment times before 5 days, payment times before 20 days, payment times before 25 days, total electric charges paid in a period, total electric quantities in a period, total default funds in a period, average electric charges in a period, average electric quantities in a period and total payment record times are obtained. In the embodiment of the application, twelve pieces of original data including a user number, an account opening date, a default fee, an electric charge receivable, an electric quantity, a payment date and a payment mode which are continuous in time are obtained by acquiring continuous payment record data of a user to be predicted within two years and taking two months as intervals, and various characteristic parameter data under different dimensions are obtained so as to cover the characteristic parameter data of various dimensions, so that the accuracy and the reliability of a prediction result are ensured.
With reference to a possible implementation manner of the embodiment of the first aspect, before determining the risk of recovering the electricity fee of the user to be predicted based on the anti-noise neural network model trained in advance and the multiple kinds of characteristic parameter data, the method further includes: acquiring training sample data with labels, wherein the training sample data comprises characteristic parameter data of a plurality of power users; and training the initial anti-noise neural network model by using the sample data to obtain the trained anti-noise neural network model, wherein a loss function of the initial anti-noise neural network model is mutual information entropy DMI based on a determinant of a joint distribution matrix of two discrete random variables. In the embodiment of the application, the model for predicting the user electric charge recovery risk is trained in advance, so that the model can be conveniently used for predicting the user electric charge recovery risk in the subsequent process, meanwhile, in the training process, mutual information entropy DMI of a determinant based on a combined distribution matrix of two discrete random variables is used as a loss function of an anti-noise neural network model, and the electric charge recovery risk of a user to be predicted is predicted according to the loss function, and because DMI is an absolute value of the determinant (det) of the combined distribution matrix of the two random discrete variables with the same value range. Thus DMI has the following properties: DMI is nonnegative, symmetrical, and meets the requirement of information monotonicity, and simultaneously meets the requirement of relative invariance, so that the DMI has better anti-interference performance on noise data, and can not be interfered by errors in the process of marking electric power data, thereby improving the accuracy of prediction.
With reference to a possible implementation manner of the embodiment of the first aspect, the characteristic parameter data of the electricity consumer is obtained through the following steps: acquiring continuous payment record data of the electricity user in a preset time period; obtaining a plurality of pieces of original data which are continuous in time based on the continuous payment record data, wherein each piece of original data comprises data related to payment under various dimensions; and obtaining various characteristic parameter data of the electricity user under different dimensions and a label for representing whether the electricity user is a defaulting user or not based on the original data. In the embodiment of the application, when the characteristic parameter data of the electricity user is acquired, twelve pieces of original data including a user number, an account opening date, a default fee, an electricity consumption amount, a payment date and a payment mode are acquired continuously in time by acquiring continuous payment record data of the electricity user within two years and taking two months as an interval, and accordingly, various characteristic parameter data under different dimensions and a label for representing whether the electricity user is a defaulting user are acquired, so that the electricity user covers the characteristic parameter data of various dimensions, and the accuracy and the reliability of a model acquired by training in prediction are guaranteed.
In a second aspect, an embodiment of the present application further provides an electric charge recycling risk prediction apparatus, including: the device comprises an acquisition module, a first acquisition module, a second acquisition module and a determination module; the system comprises an acquisition module, a storage module and a processing module, wherein the acquisition module is used for acquiring continuous payment record data of a user to be predicted within a preset time period; the first obtaining module is used for obtaining a plurality of continuous original data in time based on the continuous payment record data, and each piece of original data comprises data related to payment under various dimensions; the second obtaining module is used for obtaining multiple kinds of characteristic parameter data under different dimensions based on the multiple pieces of original data; and the determining module is used for determining the electric charge recovery risk of the user to be predicted based on a pre-trained anti-noise neural network model and the multiple characteristic parameter data, wherein a loss function of the anti-noise neural network model is mutual information entropy DMI of a determinant of a joint distribution matrix based on two discrete random variables.
With reference to one possible implementation manner of the embodiment of the second aspect, the loss function is: l isDMI(Qh(X),Y′)＝-log(DMI(h(x),Y′))＝-log(|det(Qh(X),Y′) In which Q)h(X),Y′Is a joint distribution of a discrete random variable h (X) and a discrete random variable Y', Qh(X),Y′Is in the form of a matrix of c × c, and the randomness of h (X) is derived from h and a random variable X.
In a third aspect, an embodiment of the present application further provides an electronic device, including: a memory and a processor, the processor coupled to the memory; the memory is used for storing programs; the processor is configured to invoke a program stored in the memory to perform the method according to the first aspect embodiment and/or any possible implementation manner of the first aspect embodiment.
In a fourth aspect, embodiments of the present application further provide a storage medium, on which a computer program is stored, where the computer program is executed by a processor to perform the method provided in the foregoing first aspect and/or any one of the possible implementation manners of the first aspect.
Additional features and advantages of the application will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the embodiments of the application. The objectives and other advantages of the application may be realized and attained by the structure particularly pointed out in the written description and drawings.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the embodiments will be briefly described 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 without creative efforts. The foregoing and other objects, features and advantages of the application will be apparent from the accompanying drawings. Like reference numerals refer to like parts throughout the drawings. The drawings are not intended to be to scale as practical, emphasis instead being placed upon illustrating the subject matter of the present application.
Fig. 1 shows a flow chart of an electric charge recycling risk prediction method provided in an embodiment of the present application.
Fig. 2 shows a block diagram of an electric charge recycling risk prediction apparatus according to an embodiment of the present application.
Fig. 3 shows a schematic structural diagram of an electronic device provided in an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures. Meanwhile, relational terms such as "first," "second," and the like may be used solely in the description herein to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, 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 an … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
Further, the term "and/or" in the present application is only one kind of association relationship describing the associated object, and means that three kinds of relationships may exist, for example, a and/or B may mean: a exists alone, A and B exist simultaneously, and B exists alone.
In view of the fact that the existing risk prediction model is greatly influenced by the label, the quality of the label influences the prediction accuracy, once an error exists in the data marking process, the effect of the whole model is greatly reduced, and the anti-noise capability is weak. The embodiment of the application provides an anti-noise neural network model, so that the anti-noise neural network model has better anti-interference performance on noise data, and can not be interfered by errors in the process of marking electric power data, thereby improving the accuracy of prediction.
Next, a training process of the anti-noise neural network model provided by the application is described, wherein during training, training sample data with labels is obtained firstly, wherein the training sample data comprises characteristic parameter data of each of a plurality of power users; and then training the initial anti-noise neural network model by using the obtained sample data to obtain the trained anti-noise neural network model, wherein a loss function of the initial anti-noise neural network model is a Mutual Information entropy (DMI) of a determinant of a joint distribution matrix based on two discrete random variables. Optionally, the loss function is:
LDMI(Qh(X),Y′)＝-log(DMI(h(x),Y′))＝-log(|det(Qh(X),Y′) In which Q)h(X),Y′Is a joint distribution of a discrete random variable h (X) and a discrete random variable Y', Qh(X),Y′Is in the form of a matrix of c × c, and the randomness of h (X) is derived from h and a random variable X.
The process of acquiring characteristic parameter data of an electricity consumer may be: acquiring continuous payment record data of the electricity consumer within a preset time period, for example, acquiring continuous payment record data of the electricity consumer within two years (it can be understood that the preset time period may also be other values than two years, for example, a value of one year, a half year, and the like); then, based on the continuous payment record data, obtaining a plurality of pieces of original data which are continuous in time, wherein each piece of original data comprises data related to payment under various dimensions, for example, after obtaining continuous payment record data of an electricity user within two years, twelve pieces of original data which are continuous in time are obtained at intervals of two months based on the continuous payment record data, and each piece of original data comprises: the payment method comprises the following steps that data related to payment under various dimensions, such as a user number, an account opening date, a default fee to be paid, an electric charge to be paid, an electric quantity to be used, a payment date, a payment mode and the like, are obtained; finally, based on the plurality of pieces of raw data, obtaining a plurality of characteristic parameter data of the electricity consumer in different dimensions and a label for representing whether the electricity consumer is a defaulting consumer, for example, obtaining a plurality of characteristic parameter data of the electricity consumer in different dimensions after obtaining the twelve pieces of raw data, for example, obtaining a user number, an account opening year, a comprehensive rating of a payment behavior, a predicted payment time period, a fluctuation condition of the electricity consumption, a most recent electricity quantity ring ratio, a payment mode change frequency, a defaulting frequency, a payment frequency before 5 days, a payment frequency before 20 days, a payment frequency before 25 days, a total electricity fee paid in a cycle, a total electricity quantity in a cycle, a total default fee in a cycle, an average electricity fee in a cycle, and a total payment record frequency; and taking the payment condition of the last piece of original data as a marking basis for judging whether the electricity user is a label of a defaulting user, if the default amount is more than 0, the electricity user is the defaulting user, the label is 1, if the default amount is less than 0, the electricity user is a normal user, and the label is 0.
It is understood that the label may be reversed, and if the default amount is greater than 0, the label is an owing user, the label is 0, and if the default amount is less than 0, the label is a normal user, and the label is 1. In addition, the data used as the marking basis may not be based on the payment condition of the last piece of raw data, or may be based on the payment conditions of other pieces of raw data. In addition, as an implementation manner, the payment conditions of the plurality of pieces of raw data may be used as a basis, for example, the payment conditions of three consecutive pieces of raw data are used as a basis, and of course, the payment conditions of twelve pieces of raw data may also be used as a basis, where a default amount of more than six pieces of raw data is greater than 0, a label is 1, a default amount of less than 0 is a normal user, and a label is 0.
Wherein, after acquiring the multiple characteristic parameter data of each electricity user in different dimensions, the multiple characteristic parameter data (such as user number, account opening time limit, comprehensive rating of payment behavior, estimated payment time period, fluctuation condition of electricity consumption, recent electricity quantity ring ratio, payment mode change times, arrearage times, payment times before 5 days, payment times before 20 days, payment times before 25 days, total payment charges in a period, total electricity quantity in a period, total default money in a period, average electricity charges in a period, average electricity quantity in a period, total record times of payment) are considered to represent different meanings and have different values, therefore, the method can also carry out characteristic normalization processing on each characteristic parameter data in the plurality of characteristic parameter data, for example, a Standard Scaler, a MinMaxScaler, a QuantileTransformer and the like provided by scinit-leran are used for feature normalization; correspondingly, during training, the initial anti-noise neural network model is trained based on the various feature parameter data after normalization processing.
During training, the obtained training sample data with labels, including the characteristic parameter data of each of a plurality of electricity users, is divided into a training set and a test set according to a certain proportion, for example, a train _ test _ split method provided by scimit-lean is utilized according to a proportion of 5:1, and then the training set is divided into the training set and a verification set according to a certain proportion, for example, according to a proportion of 5: 1. The method comprises the steps of initializing the structure of a neural network by utilizing a PyTorch framework, inputting and outputting the number of nodes of a hidden layer, positioning Adam by a network optimizer, and setting the number of iterations to be 1000, wherein the network optimization algorithm can be replaced by modes such as SGD (generalized minimum) mode, Momentum mode, Nesterov mode and Adadelta besides the Adam mode. And taking the training set as the input of the initial anti-noise neural network model, and obtaining the network accuracy, the positive recall rate and the negative recall rate by using the verification set as the reference standard of model parameter adjustment every time iteration is completed. The iteration is continued until the number of iterations reaches a preset value, such as 1000. And finally, storing the model by using PyTorch, loading a test set, and carrying out algorithm prediction to obtain a prediction result, accuracy, a positive recall rate and a negative recall rate. It should be noted that the specific training process is well known to those skilled in the art, and is not described herein.
The initialization neural network structure adopted by the embodiment of the application can be a BP neural network structure, and the algorithm can directly reduce the error correction operation amount step by step to be only in direct proportion to the number of the neurons. Meanwhile, the BP neural network is a computing system simulating the development of human brain neural tissue, has a network system formed by a large number of processing units and wide interconnection, has the basic characteristics of biological brain-like, and has the advantages of large-scale parallel, distributed processing, self-learning and the like. In addition, the electric power data related in the application has the characteristics of more dimensionality, higher complexity and nonlinearity, and the BP neural network has stronger nonlinear mapping capability, and meanwhile, the generalization capability and the fault tolerance capability are also stronger than those of the traditional machine learning algorithm.
For the loss function employed, the following are defined:
combining equations (1), (2) and property theorem, we can obtain:
this algebraic structure of the DMI makes it possible to measure the DMI of the classifier output W2 and the channel input W1, the channel output W3, respectively, with the noisy channel (T) fixed.
In view of this, the present application proposes a completely new loss function LDMI that is easy to optimize, based on DMI:
LDMI(Qh(X),Y′)＝-log(DMI(h(x),Y′))＝-log(|det(Qh(X),Y′)|)(4)
wherein Q ish(X),Y′Is a joint distribution of a discrete random variable h (X) and a discrete random variable Y', Qh(X),Y′The randomness of h (X) is in a matrix form of c × c, and the randomness of h (X) comes from h and a random variable X, reasoning and experimental results prove that the rationality of LDMI and the antagonism and robustness to noise can be proved, so that when LDMI is used as a loss function, the effect of a classification predictor trained on a data set with a noise label and a data set without the noise label is the same, better interference resistance is realized on noise data, the method has a remarkable advantage on application to power charge recovery prediction, and error interference in a power data labeling process can be avoided.
After the anti-noise neural network model is trained by the method, the electric charge recovery risk of the user to be predicted can be predicted so as to solve the problem of difficult electric charge recovery. The method for predicting the risk of recovering electricity charges provided by the embodiment of the present application will be described with reference to fig. 1.
Step S101: and acquiring continuous payment record data of the user to be predicted within a preset time period.
When the electric charge recycling risk of a certain electricity user needs to be predicted, the continuous payment record data of the user to be predicted in a preset time period is obtained, for example, the continuous payment record data of the user to be predicted in two years is obtained. It is understood that the preset time period may be other than two years, for example, a year, a half year, a year and a half, etc.
Step S102: and obtaining a plurality of pieces of original data which are continuous in time based on the continuous payment record data, wherein each piece of original data comprises data related to payment under various dimensions.
After acquiring continuous payment record data of a user to be predicted within a preset time period, acquiring multiple pieces of temporally continuous raw data based on the continuous payment record data, wherein each piece of raw data comprises data related to payment under multiple different dimensions, for example, twelve pieces of temporally continuous raw data can be acquired based on continuous payment record data within two years at two-month intervals, and each piece of raw data comprises: the method comprises the steps of user number, account opening date, payment default fee, payment electric charge, electric quantity, payment date and payment mode.
Step S103: and obtaining a plurality of characteristic parameter data under different dimensions based on the plurality of pieces of original data.
After a plurality of pieces of original data which are continuous in time are obtained, based on the original data and according to the category of statistics, on the basis, new features can be obtained, and therefore a plurality of feature parameter data under different dimensions are obtained. For example, based on twelve pieces of raw data which are continuous in time, a user number, an account opening age, a comprehensive rating of a payment behavior, a predicted payment time period, a fluctuation condition of electricity consumption, a latest electricity quantity ring ratio, a payment mode change frequency, an arrearage frequency, a payment frequency before 5 days, a payment frequency before 20 days, a payment frequency before 25 days, a total electricity charge paid in a period, a total electricity quantity in a period, a total default fund in a period, an average electricity charge in a period, an average electricity quantity in a period, and a total record frequency of payment are obtained.
Step S104: and determining the electric charge recycling risk of the user to be predicted based on the pre-trained anti-noise neural network model and the multiple characteristic parameter data.
After various feature parameter data under different dimensions are obtained based on the continuous payment record data of the user to be predicted in a preset time period, the electric charge recycling risk of the user to be predicted is determined based on a pre-trained anti-noise neural network model and the various feature parameter data. And the loss function of the anti-noise neural network model is mutual information entropy DMI based on a determinant of a joint distribution matrix of two discrete random variables. Wherein the loss function is:
LDMI(Qh(X),Y′)＝-log(DMI(h(x),Y′))＝-log(|det(Qh(X),Y′) |) itIn, Qh(X),Y′Is a joint distribution of a discrete random variable h (X) and a discrete random variable Y', Qh(X),Y′Is in the form of a matrix of c × c, and the randomness of h (X) is derived from h and a random variable X.
In consideration of different meanings and different values of the multiple characteristic parameter data, as an implementation manner, the present application may further perform characteristic normalization processing on each characteristic parameter data in the multiple characteristic parameter data, for example, by using a method such as Standard Scaler, minmaxscale, quantum transformer provided by sciit-leran; correspondingly, when the electric charge recycling risk is predicted, the electric charge recycling risk of the user to be predicted is determined based on the anti-noise neural network model trained in advance and the various characteristic parameter data after normalization processing.
When the electric charge recycling risk prediction is carried out, the adopted prediction model is an anti-noise neural network model which is trained in advance by adopting the training method, namely, training sample data with labels is obtained during training, and the training sample data comprises characteristic parameter data of each of a plurality of power users; and training the initial anti-noise neural network model by using the sample data to obtain the trained anti-noise neural network model, wherein a loss function of the initial anti-noise neural network model is mutual information entropy DMI based on a determinant of a joint distribution matrix of two discrete random variables.
According to the method and the device, when the electric charge recycling risk of the user to be predicted is predicted, the anti-noise neural network model is adopted, mutual information entropy DMI of a determinant of a joint distribution matrix based on two discrete random variables is adopted as a loss function, and DMI is an absolute value of the determinant (det) of the joint distribution matrix of the two random discrete variables with the same value range. Thus DMI has the following properties: DMI is nonnegative, symmetrical, and meets the requirement of information monotonicity, and simultaneously meets the requirement of relative invariance, so that the DMI has better anti-interference performance on noise data, and can not be interfered by errors in the process of marking electric power data, thereby improving the accuracy of prediction.
The embodiment of the present application further provides an electric charge recycling risk prediction apparatus 100, as shown in fig. 2. The electric charge collection risk prediction device 100 includes: an acquisition module 110, a first acquisition module 120, a second acquisition module 130, and a determination module 140.
The obtaining module 110 is configured to obtain continuous payment record data of a user to be predicted within a preset time period. Optionally, the obtaining module 110 is configured to obtain continuous payment record data of the user to be forecasted within two years.
A first obtaining module 120, configured to obtain multiple pieces of raw data that are continuous in time based on the continuous payment record data, where each piece of raw data includes data related to payment in multiple different dimensions. Optionally, the first obtaining module 120 is configured to obtain twelve pieces of raw data that are continuous in time at intervals of two months based on the continuous payment record data, where each piece of raw data includes: the method comprises the steps of user number, account opening date, payment default fee, payment electric charge, electric quantity, payment date and payment mode.
A second obtaining module 130, configured to obtain multiple kinds of feature parameter data under different dimensions based on the multiple pieces of raw data. Optionally, the second obtaining module 130 is configured to obtain, based on the multiple pieces of raw data, a user number, an account opening time, a comprehensive rating of a payment behavior, a predicted payment time period, a fluctuation condition of power consumption amount, a latest power amount ring ratio, a payment mode change time, an arrearage time, a payment time before 5 days, a payment time before 20 days, a payment time before 25 days, a total power charge paid in a cycle, a total power amount in a cycle, a total default fee in a cycle, an average power charge in a cycle, an average power amount in a cycle, and a total record time of the payment.
The determining module 140 is configured to determine the electric charge recycling risk of the user to be predicted based on a pre-trained anti-noise neural network model and the multiple kinds of characteristic parameter data, where a loss function of the anti-noise neural network model is mutual information entropy DMI of a determinant of a joint distribution matrix based on two discrete random variables. Wherein the loss function is:
LDMI(Qh(X),Y′)＝-log(DMI(h(x),Y′))＝-log(|det(Qh(X),Y′) In which Q)h(X),Y′Is a joint distribution of a discrete random variable h (X) and a discrete random variable Y', Qh(X),Y′Is in the form of a matrix of c × c, and the randomness of h (X) is derived from h and a random variable X.
Optionally, the electric charge collection risk prediction apparatus 100 further includes: and a normalization module. The normalization module is configured to, after the second obtaining module 130 obtains multiple kinds of feature parameter data in different dimensions based on the multiple pieces of original data, perform feature normalization processing on each kind of feature parameter data in the multiple kinds of feature parameter data. Correspondingly, the determining module 140 is configured to determine the electric charge recycling risk of the user to be predicted based on the anti-noise neural network model trained in advance and the multiple feature parameter data after the normalization processing.
Optionally, the electric charge collection risk prediction apparatus 100 further includes: and a training module. Correspondingly, the obtaining module 110 is further configured to obtain training sample data with labels, where the training sample data includes characteristic parameter data of each of a plurality of power users. The method comprises the following steps of obtaining characteristic parameter data of a power utilization user: acquiring continuous payment record data of the electricity user in a preset time period; obtaining a plurality of pieces of original data which are continuous in time based on the continuous payment record data, wherein each piece of original data comprises data related to payment under various dimensions; and obtaining various characteristic parameter data of the electricity user under different dimensions and a label for representing whether the electricity user is a defaulting user or not based on the original data.
The training module is used for acquiring training sample data with labels, and the training sample data comprises characteristic parameter data of each of a plurality of power users.
The electric charge recycling risk prediction apparatus 100 provided in the embodiment of the present application has the same implementation principle and technical effect as those of the foregoing method embodiments, and for brief description, reference may be made to the corresponding contents in the foregoing method embodiments for the parts of the apparatus embodiments that are not mentioned.
As shown in fig. 3, fig. 3 is a block diagram illustrating a structure of an electronic device 200 according to an embodiment of the present disclosure. The electronic device 200 includes: a transceiver 210, a memory 220, a communication bus 230, and a processor 240.
The elements of the transceiver 210, the communication memory 220, and the processor 240 are electrically connected to each other directly or indirectly to achieve data transmission or interaction. For example, the components may be electrically coupled to each other via one or more communication buses 230 or signal lines. The transceiver 210 is used for transceiving data. The general memory 220 is used to store a computer program, such as the software functional module shown in fig. 2, that is, the electric charge recycling risk prediction apparatus 100. The electric charge recycling risk prediction apparatus 100 includes at least one software function module that may be stored in the memory 220 in the form of software or firmware (firmware) or solidified in an Operating System (OS) of the electronic device 200. The processor 240 is configured to execute an executable module stored in the memory 220, such as a software function module or a computer program included in the electric charge charging recovery risk prediction apparatus 100. For example, when the processor 240 executes the electric charge recycling risk prediction apparatus 100 shown in fig. 2, the processor 240 is configured to obtain continuous payment record data of the user to be predicted within a preset time period; obtaining a plurality of pieces of original data which are continuous in time based on the continuous payment record data, wherein each piece of original data comprises data related to payment under various dimensions; obtaining multiple kinds of characteristic parameter data under different dimensions based on the multiple pieces of original data; and determining the electric charge recovery risk of the user to be predicted based on a pre-trained anti-noise neural network model and the multiple characteristic parameter data, wherein a loss function of the anti-noise neural network model is mutual information entropy DMI of a determinant of a joint distribution matrix based on two discrete random variables.
The Memory 220 may be, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Read-Only Memory (EPROM), an electrically Erasable Read-Only Memory (EEPROM), and the like.
The processor 240 may be an integrated circuit chip having signal processing capabilities. The processor may be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components. The various methods, steps, and logic blocks disclosed in the embodiments of the present application may be implemented or performed. A general purpose processor may be a microprocessor or the processor 240 may be any conventional processor or the like.
The electronic device 200 includes, but is not limited to, a computer, a personal computer, and the like.
The embodiment of the present application further provides a non-volatile computer-readable storage medium (hereinafter, referred to as a storage medium), where a computer program is stored on the storage medium, and when the computer program is run by the electronic device 200 as described above, the method for predicting the risk of electric charge recycling described above is executed.
It should be noted that, in the present specification, the embodiments are all described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments may be referred to each other.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative, and for example, the flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
In addition, functional modules in the embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.
The functions, if implemented in the form of software functional modules and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a notebook computer, a server, or an electronic device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. A method for predicting risk of recovering electricity charges is characterized by comprising the following steps:
acquiring continuous payment record data of a user to be predicted within a preset time period;
obtaining a plurality of pieces of original data which are continuous in time based on the continuous payment record data, wherein each piece of original data comprises data related to payment under various dimensions;
obtaining multiple kinds of characteristic parameter data under different dimensions based on the multiple pieces of original data;
and determining the electric charge recovery risk of the user to be predicted based on a pre-trained anti-noise neural network model and the multiple characteristic parameter data, wherein a loss function of the anti-noise neural network model is mutual information entropy DMI of a determinant of a joint distribution matrix based on two discrete random variables.
2. The method of claim 1, wherein the loss function is: l isDMI(Qh(X),Y′)＝-log(DMI(h(x),Y′))＝-log(|det(Qh(X),Y′) In which Q)h(X),Y′Is a joint distribution of a discrete random variable h (X) and a discrete random variable Y', Qh(X),Y′Is in the form of a matrix of c × c, and the randomness of h (X) is derived from h and a random variable X.
3. The method of claim 1, wherein after obtaining a plurality of feature parameter data in different dimensions, the method further comprises:
respectively carrying out characteristic normalization processing on each characteristic parameter data in the multiple characteristic parameter data;
correspondingly, determining the electric charge recycling risk of the user to be predicted based on the anti-noise neural network model trained in advance and the various characteristic parameter data, and the method comprises the following steps:
and determining the electric charge recycling risk of the user to be predicted based on the anti-noise neural network model trained in advance and the various characteristic parameter data subjected to normalization processing.
4. The method according to claim 1, wherein the step of obtaining continuous payment record data of the user to be predicted in a preset time period comprises the following steps:
acquiring continuous payment record data of the user to be predicted within two years;
correspondingly, based on the continuous payment record data, obtaining a plurality of pieces of original data which are continuous in time, wherein the original data comprise:
based on the continuous payment record data, twelve continuous original data in time are obtained at intervals of two months, wherein each piece of original data comprises: the method comprises the following steps of (1) user number, account opening date, charge default fee, charge fee, power consumption, charge date and charge mode;
correspondingly, based on the plurality of pieces of original data, obtaining a plurality of kinds of feature parameter data under different dimensions, including:
based on the plurality of pieces of original data, user numbers, account opening years, comprehensive scores of payment behaviors, predicted payment time periods, power consumption electric quantity fluctuation conditions, the most recent electric quantity cyclic ratio, payment mode change times, arrearage times, payment times before 5 days, payment times before 20 days, payment times before 25 days, total electric charges paid in a period, total electric quantities in a period, total default funds in a period, average electric charges in a period, average electric quantities in a period and total payment record times are obtained.
5. The method according to any one of claims 1 to 4, wherein before determining the risk of recovering the electricity charge of the user to be predicted based on the anti-noise neural network model trained in advance and the plurality of characteristic parameter data, the method further comprises:
acquiring training sample data with labels, wherein the training sample data comprises characteristic parameter data of a plurality of power users;
and training the initial anti-noise neural network model by using the sample data to obtain the trained anti-noise neural network model, wherein a loss function of the initial anti-noise neural network model is mutual information entropy DMI based on a determinant of a joint distribution matrix of two discrete random variables.
6. The method according to claim 5, characterized in that the characteristic parameter data of the electricity user is obtained by the following steps:
acquiring continuous payment record data of the electricity user in a preset time period;
obtaining a plurality of pieces of original data which are continuous in time based on the continuous payment record data, wherein each piece of original data comprises data related to payment under various dimensions;
and obtaining various characteristic parameter data of the electricity user under different dimensions and a label for representing whether the electricity user is a defaulting user or not based on the original data.
7. An electric charge recovery risk prediction device characterized by comprising:
the system comprises an acquisition module, a storage module and a processing module, wherein the acquisition module is used for acquiring continuous payment record data of a user to be predicted within a preset time period;
the first obtaining module is used for obtaining a plurality of continuous original data in time based on the continuous payment record data, and each piece of original data comprises data related to payment under various dimensions;
the second obtaining module is used for obtaining multiple kinds of characteristic parameter data under different dimensions based on the multiple pieces of original data;
and the determining module is used for determining the electric charge recovery risk of the user to be predicted based on a pre-trained anti-noise neural network model and the multiple characteristic parameter data, wherein a loss function of the anti-noise neural network model is mutual information entropy DMI of a determinant of a joint distribution matrix based on two discrete random variables.
8. The apparatus of claim 7,the loss function is: l isDMI(Qh(X),Y′)＝-log(DMI(h(x),Y′))＝-log(|det(Qh(X),Y′) In which Q)h(X),Y′Is a joint distribution of a discrete random variable h (X) and a discrete random variable Y', Qh(X),Y′Is in the form of a matrix of c × c, and the randomness of h (X) is derived from h and a random variable X.
9. An electronic device, comprising:
a memory and a processor, the processor coupled to the memory;
the memory is used for storing programs;
the processor to invoke a program stored in the memory to perform the method of any of claims 1-6.
10. A storage medium having stored thereon a computer program which, when executed by a processor, performs the method according to any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010128050.3A CN111340375A (en) | 2020-02-28 | 2020-02-28 | Electricity charge recycling risk prediction method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010128050.3A CN111340375A (en) | 2020-02-28 | 2020-02-28 | Electricity charge recycling risk prediction method and device, electronic equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111340375A true CN111340375A (en) | 2020-06-26 |
Family
ID=71183980
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010128050.3A Pending CN111340375A (en) | 2020-02-28 | 2020-02-28 | Electricity charge recycling risk prediction method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111340375A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111724000A (en) * | 2020-06-29 | 2020-09-29 | 南方电网科学研究院有限责任公司 | Method, device and system for predicting user electric charge recycling risk |
| CN112269815A (en) * | 2020-10-29 | 2021-01-26 | 维沃移动通信有限公司 | Structured data processing method and device and electronic equipment |
| CN112598171A (en) * | 2020-12-21 | 2021-04-02 | 国网雄安金融科技集团有限公司 | Enterprise electric charge prediction method and system under electric power big data |
| CN113256008A (en) * | 2021-05-31 | 2021-08-13 | 国家电网有限公司大数据中心 | Arrearage risk level determination method, device, equipment and storage medium |
| CN113592140A (en) * | 2021-06-22 | 2021-11-02 | 国网宁夏电力有限公司吴忠供电公司 | Electric charge payment prediction model training system and electric charge payment prediction model |
| CN116433403A (en) * | 2023-06-14 | 2023-07-14 | 国网安徽省电力有限公司营销服务中心 | Account tracking-based electric enterprise accounts receivable early warning method and system |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106251049A (en) * | 2016-07-25 | 2016-12-21 | 国网浙江省电力公司宁波供电公司 | A kind of electricity charge risk model construction method of big data |
| CN107085769A (en) * | 2017-04-28 | 2017-08-22 | 国网山东省电力公司泰安供电公司 | Electricity charge method for prewarning risk and device |
| CN107180392A (en) * | 2017-05-18 | 2017-09-19 | 北京科技大学 | A kind of electric power enterprise tariff recovery digital simulation method |
| CN107679491A (en) * | 2017-09-29 | 2018-02-09 | 华中师范大学 | A kind of 3D convolutional neural networks sign Language Recognition Methods for merging multi-modal data |
| CN108053030A (en) * | 2017-12-15 | 2018-05-18 | 清华大学 | A kind of transfer learning method and system of Opening field |
| CN108229659A (en) * | 2017-12-29 | 2018-06-29 | 陕西科技大学 | Piano singly-bound voice recognition method based on deep learning |
| CN109189933A (en) * | 2018-09-14 | 2019-01-11 | 腾讯科技（深圳）有限公司 | A kind of method and server of text information classification |
| CN110210686A (en) * | 2019-06-13 | 2019-09-06 | 郑州轻工业学院 | A kind of electricity charge risk model construction method of electric power big data |
-
2020
- 2020-02-28 CN CN202010128050.3A patent/CN111340375A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106251049A (en) * | 2016-07-25 | 2016-12-21 | 国网浙江省电力公司宁波供电公司 | A kind of electricity charge risk model construction method of big data |
| CN107085769A (en) * | 2017-04-28 | 2017-08-22 | 国网山东省电力公司泰安供电公司 | Electricity charge method for prewarning risk and device |
| CN107180392A (en) * | 2017-05-18 | 2017-09-19 | 北京科技大学 | A kind of electric power enterprise tariff recovery digital simulation method |
| CN107679491A (en) * | 2017-09-29 | 2018-02-09 | 华中师范大学 | A kind of 3D convolutional neural networks sign Language Recognition Methods for merging multi-modal data |
| CN108053030A (en) * | 2017-12-15 | 2018-05-18 | 清华大学 | A kind of transfer learning method and system of Opening field |
| CN108229659A (en) * | 2017-12-29 | 2018-06-29 | 陕西科技大学 | Piano singly-bound voice recognition method based on deep learning |
| CN109189933A (en) * | 2018-09-14 | 2019-01-11 | 腾讯科技（深圳）有限公司 | A kind of method and server of text information classification |
| CN110210686A (en) * | 2019-06-13 | 2019-09-06 | 郑州轻工业学院 | A kind of electricity charge risk model construction method of electric power big data |
Non-Patent Citations (1)
| Title |
|---|
| XUYILUN, CAOPENG: "A Novel Information-theoretic Loss Function for Training Deep Nets Robust to Label Noise", 33RD CONFERENCE ON NEURAL INFORMATION PROCESSING SYSTEMS (NEURIPS 2019) * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111724000A (en) * | 2020-06-29 | 2020-09-29 | 南方电网科学研究院有限责任公司 | Method, device and system for predicting user electric charge recycling risk |
| CN111724000B (en) * | 2020-06-29 | 2024-02-09 | 南方电网科学研究院有限责任公司 | User electricity charge recycling risk prediction method, device and system |
| CN112269815A (en) * | 2020-10-29 | 2021-01-26 | 维沃移动通信有限公司 | Structured data processing method and device and electronic equipment |
| CN112598171A (en) * | 2020-12-21 | 2021-04-02 | 国网雄安金融科技集团有限公司 | Enterprise electric charge prediction method and system under electric power big data |
| CN113256008A (en) * | 2021-05-31 | 2021-08-13 | 国家电网有限公司大数据中心 | Arrearage risk level determination method, device, equipment and storage medium |
| CN113592140A (en) * | 2021-06-22 | 2021-11-02 | 国网宁夏电力有限公司吴忠供电公司 | Electric charge payment prediction model training system and electric charge payment prediction model |
| CN116433403A (en) * | 2023-06-14 | 2023-07-14 | 国网安徽省电力有限公司营销服务中心 | Account tracking-based electric enterprise accounts receivable early warning method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111340375A (en) | Electricity charge recycling risk prediction method and device, electronic equipment and storage medium | |
| Nguyen et al. | Forecasting and Anomaly Detection approaches using LSTM and LSTM Autoencoder techniques with the applications in supply chain management | |
| US20200242483A1 (en) | Method and system of dynamic model selection for time series forecasting | |
| Sentas et al. | Software productivity and effort prediction with ordinal regression | |
| CN110400022B (en) | Cash consumption prediction method and device for self-service teller machine | |
| US20210125073A1 (en) | Method and system for individual demand forecasting | |
| Nassar et al. | Fuzzy clustering validity for contractor performance evaluation: Application to UAE contractors | |
| KR20200039852A (en) | Method for analysis of business management system providing machine learning algorithm for predictive modeling | |
| Johnson et al. | A method for predicting the probability of business network profitability | |
| CN113256324A (en) | Data asset pricing method, device, computer equipment and storage medium | |
| WO2021077226A1 (en) | Method and system for individual demand forecasting | |
| CN115375205A (en) | Method, device and equipment for determining water user portrait | |
| Luo et al. | Financial high-frequency time series forecasting based on sub-step grid search long short-term memory network | |
| Lakhno et al. | Methodology for assessing the effectiveness of measures aimed at ensuring information security of the object of informatization | |
| Nawaz et al. | A novel technique for detecting electricity theft in secure smart grids using CNN and XG-boost | |
| Zhang et al. | Predicting carbon futures prices based on a new hybrid machine learning: Comparative study of carbon prices in different periods | |
| Precioso et al. | Thresholding methods in non-intrusive load monitoring | |
| US20230401468A1 (en) | Methods and systems for generating forecasts using an ensemble online demand generation forecaster | |
| CN115358878A (en) | Financing user risk preference level analysis method and device | |
| Abdurohman et al. | Forecasting model for lighting electricity load with a limited dataset using xgboost | |
| Heymann et al. | Reviewing 40 years of artificial intelligence applied to power systems–A taxonomic perspective | |
| Wang et al. | Early-warning of generator collusion in Chinese electricity market based on Information Deep Autoencoding Gaussian Mixture Model | |
| Ding et al. | LSTM Deep Neural Network Based Power Data Credit Tagging Technology. | |
| Byrne et al. | Man bites dog: Looking for interesting inconsistencies in structured news reports | |
| Tanoni et al. | Knowledge Distillation for Scalable Nonintrusive Load Monitoring |
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 |