CN116187989A

CN116187989A - Bill auditing method, device, electronic equipment and storage medium

Info

Publication number: CN116187989A
Application number: CN202211662743.6A
Authority: CN
Inventors: 王瑜; 李红云; 孙洋洋; 熊建胜; 周莹; 沙升升
Original assignee: China United Network Communications Group Co Ltd
Current assignee: China United Network Communications Group Co Ltd
Priority date: 2022-12-23
Filing date: 2022-12-23
Publication date: 2023-05-30

Abstract

The application discloses a bill auditing method, a bill auditing device, electronic equipment and a storage medium, relates to the technical field of artificial intelligence, and solves the problem of low accuracy of the conventional bill auditing. The method comprises the following steps: acquiring historical bill information of a target object, wherein the historical bill information is bill information of the target object in a preset period; determining predicted bill information of the target object based on the target prediction model library and the historical bill information, wherein the predicted bill information is bill information in a preset period after the historical bill information, and the target prediction model library comprises a plurality of target prediction models; acquiring real bill information of a target object, wherein the time period corresponding to the real bill information is the same as the time period corresponding to the predicted bill information; based on the predicted bill information and the real bill information, a comparison result is obtained, and the comparison result is used for indicating whether abnormal bill information exists in the real bill information. According to the embodiment of the application, the accuracy of bill auditing can be improved.

Description

Bill auditing method, device, electronic equipment and storage medium

Technical Field

The present invention relates to the field of artificial intelligence technologies, and in particular, to a bill auditing method, apparatus, electronic device, and storage medium.

Background

With the discovery of internet technology, more and more services realize corresponding functions by means of the internet. For example, the approval work of the bill is realized through the internet technology. The current bill approval work is to directly mark bills exceeding a certain proportion as abnormal data through the bill daily average ring ratio of two adjacent account periods and mark and lock the bills in red.

However, by adopting the method for bill examination and approval, not only the condition of bill marking errors can occur, but also the proportion is usually set to be too large in order to reduce the errors, so that bill information which does not exceed the preset proportion but is still obvious in difference can not be detected, the accuracy of bill examination and approval is lower, and the working efficiency of the newspaper staff is further reduced.

Disclosure of Invention

The invention provides a bill auditing method, a bill auditing device, electronic equipment and a storage medium, which are used for improving the accuracy of bill auditing.

In order to achieve the above purpose, the invention adopts the following technical scheme:

in a first aspect, the present invention provides a bill auditing method, including:

acquiring historical bill information of a target object, wherein the historical bill information is bill information of the target object in a preset period;

Determining predicted bill information of the target object based on the target prediction model library and the historical bill information, wherein the predicted bill information is bill information in a preset period after the historical bill information, and the target prediction model library comprises a plurality of target prediction models;

acquiring real bill information of a target object, wherein the time period corresponding to the real bill information is the same as the time period corresponding to the predicted bill information;

based on the predicted bill information and the real bill information, a comparison result is obtained, and the comparison result is used for indicating whether abnormal bill information exists in the real bill information.

In this embodiment, since the target prediction model in the target prediction model library is a trained model and the historical billing information is the billing information before the predicted billing information, the higher the standard degree of the predicted billing information obtained according to the target prediction model library and the historical billing information is, so that according to the predicted billing information and the real billing information, whether the real billing information has abnormal billing information is determined, the accuracy of bill audit can be improved, the occurrence of the situations of wrong detection and false detection of bills can be reduced, convenience is provided for the approval work of the bill issuers, and the work efficiency of the bill issuers is improved.

In one possible embodiment, the plurality of target prediction models are trained by:

acquiring a bill information set, wherein the bill information set comprises a plurality of sample bill information, and each sample bill information comprises a plurality of sample information parameters;

preprocessing sample bill information to obtain sample bill data;

based on a preset feature classification library, carrying out feature classification on each sample information parameter to obtain a type classification result and a trend classification result of each sample information parameter, wherein the preset feature classification library comprises type features and trend features, the type classification result comprises type feature data, and the trend classification result comprises at least one trend feature data;

based on the type classification result and the trend classification result of each sample information parameter, respectively training a plurality of prediction models in the prediction model library to obtain a plurality of target prediction models.

In this embodiment, each sample information parameter is divided into a type classification result and a trend classification result, and according to the type classification result and the trend classification result, a plurality of prediction models in a prediction model library are respectively trained to obtain a plurality of target prediction models, so that the prediction models in the prediction model library can be respectively trained according to the characteristics of the type characteristic data and the trend characteristic data, so that in the subsequent bill information prediction process, the accuracy of bill information prediction can be improved, convenience is provided for approval work of the account-reporting personnel, and the work efficiency of the account-reporting personnel is improved.

In one possible implementation manner, the preprocessing the sample bill information to obtain sample bill data includes:

determining a null value rate and an outlier rate of the sample bill information;

under the condition that the null value rate is not higher than a first preset proportion and the abnormal value rate is not higher than a second preset proportion, carrying out information filling processing on the sample bill information to obtain target sample bill information;

and carrying out information processing on the target sample bill information to obtain sample bill data.

In this embodiment, if the null value rate is not higher than the first preset proportion and the outlier value rate is not higher than the second preset proportion, the sample bill information is subjected to information filling processing so as to ensure the integrity of the sample bill information, and the target sample bill information is subjected to information processing so as to obtain sample bill data, so that convenience is provided for subsequent model training work, and the accuracy of bill information prediction is improved.

In one possible embodiment, the method further comprises:

and deleting the sample bill information under the condition that the null value rate is higher than the first preset proportion and/or the abnormal value rate is higher than the second preset proportion.

In this embodiment, if the null rate is higher than the first preset proportion, it is indicated that the bill data in the sample bill information is too much lack and has no referential, and at the same time, the training result of the prediction model may be disturbed; if the abnormal value rate is higher than the second preset proportion, the abnormal bill data in the sample bill information is too much, the abnormal bill data is not referenced, and the training result of the prediction model is possibly disturbed, so that the sample bill information needs to be deleted, the training of the prediction model, which is influenced by the lack of excessive bill data or excessive abnormal bill data, is reduced, and the accuracy of bill information prediction is further improved.

In a possible implementation manner, the feature classification is performed on each sample information parameter based on a preset feature classification library to obtain a type classification result and a trend classification result of each sample information parameter, which includes:

respectively matching each sample information parameter with type features in a preset feature classification library to obtain a type classification result of each sample information parameter;

and determining a trend classification result of each sample information parameter based on the same sample information parameter in the plurality of sample bill data and trend characteristics in a preset characteristic classification library.

In this embodiment, the type classification result is obtained by matching the sample information parameter with the type feature in the preset feature classification library, so that accuracy of determining the type classification result can be improved, and the trend classification result is determined by comparing the same sample information parameter in the plurality of sample bill data with the trend feature in the preset feature classification library, so that accuracy of determining the trend classification result can be improved, and training precision of the prediction model is improved, so that the prediction result obtained by the prediction model is closer to the real result.

In one possible implementation manner, the determining the trend classification result of each sample information parameter based on the same sample information parameter in the plurality of sample bill data and the trend feature in the preset feature classification library includes:

aiming at each trend feature, calculating the same sample information parameter in the plurality of sample bill data according to different index rules to obtain a plurality of index values corresponding to the different index rules;

respectively carrying out error matching on the plurality of index values to obtain a plurality of error values;

determining a target number lower than a preset value in the error values, and determining whether the target number exceeds the preset number;

and determining a trend classification result based on the trend characteristics under the condition that the target number exceeds the preset number.

In this embodiment, for each trend feature, the same sample information parameter in the plurality of sample bill data is calculated according to different index rules to obtain a plurality of index values, error matching is performed on the plurality of index values respectively to obtain a plurality of error values, whether the target number lower than the preset value in the plurality of error values exceeds the preset number is determined, if the target number exceeds the preset number, a trend classification result can be determined according to the trend feature, so that the accuracy of determining the trend classification result can be improved, and the training precision of the prediction model is improved, so that the prediction result obtained by the prediction model is closer to the real result.

In one possible implementation manner, the training the multiple prediction models in the prediction model library based on the type classification result and the trend classification result of each sample information parameter to obtain multiple target prediction models includes:

for each feature data in the sample information parameters, determining a first prediction model matched with the feature data from a prediction model library, wherein the feature data is type feature data or trend feature data;

inputting the characteristic data into a first prediction model to obtain prediction data, wherein the prediction data is adjacent characteristic data positioned in a preset period after the characteristic data;

acquiring real data in the same time period as the predicted data;

and carrying out parameter adjustment on the first prediction model based on the real data and the prediction data to obtain a target prediction model.

In this embodiment, a first prediction model matched with feature data is determined from a prediction model library, the feature data is input into the first prediction model to obtain prediction data, and finally, parameter adjustment is performed on the first prediction model according to the prediction data and real data to obtain a target prediction model, so that each feature data corresponds to one prediction model, the occurrence of inaccurate prediction results caused by mismatching of the prediction model and the feature data is reduced, and the accuracy of determination of the prediction data is improved.

In one possible implementation manner, the inputting the feature data into the first prediction model to obtain the prediction data includes:

determining a model type of the first prediction model;

and when the first prediction model is a tree model, inputting the characteristic data into the first prediction model to obtain prediction data.

In this embodiment, since the first prediction model is a tree model, and the tree model can automatically find a higher-order relationship between features, the feature data can be directly input into the first prediction model without processing the feature data, and thus, the training process can be simplified without affecting the training result of the first prediction model.

In one possible embodiment, the method further comprises:

under the condition that the first prediction model is a neural network, carrying out standardization processing on the characteristic data;

and inputting the processed characteristic data into a first prediction model to obtain prediction data.

In this embodiment, since the first prediction model is a neural network and the sensitivity of the neural network to the higher-order relationships between features is low, the higher-order relationships between features cannot be automatically found and processed, and it is necessary to perform normalization processing on the feature data before training the first prediction model, so as to improve the robustness of the target prediction model.

In one possible implementation, the historical billing information of the target object includes a plurality of billing information parameters, each billing information parameter includes a plurality of feature information, and determining the predicted billing information of the target object based on the target prediction model library and the historical billing information includes:

selecting a first target prediction model matched with the characteristic information from a target prediction model library aiming at each characteristic information;

inputting the characteristic information into a first target prediction model to obtain prediction bill data corresponding to the characteristic information;

and obtaining the predicted bill information of the target object based on the predicted bill data corresponding to each piece of characteristic information.

In this embodiment, each bill information parameter is input into the first target prediction model matched with the first target prediction model to obtain predicted bill data corresponding to the bill information parameter, and predicted bill information of the target object is obtained according to the predicted bill data corresponding to each bill information parameter, so that each bill information parameter is input into the prediction model matched with the first target prediction model to predict, accuracy of bill data prediction can be improved, a basis is provided for subsequent bill information comparison work, and whether abnormal bill information exists in real bill information can be accurately determined.

In one possible implementation manner, the obtaining the comparison result based on the predicted bill information and the real bill information includes:

comparing the predicted bill information with the real bill information to obtain a difference result, wherein the difference result is used for indicating whether the difference exists between the predicted bill information and the real bill information;

determining an absolute value of a difference ratio corresponding to a bill information parameter in which the predicted bill information and the real bill information are different under the condition that the difference result indicates that the difference exists between the predicted bill information and the real bill information;

determining whether the absolute value of the difference proportion is larger than a third preset proportion;

and under the condition that the absolute value of the difference proportion is larger than a third preset proportion, determining that abnormal bill information exists in the real bill information.

In this embodiment, if there is a difference between the predicted bill information and the real bill information, it needs to be determined whether the absolute value of the difference ratio is greater than a third preset ratio, and if the absolute value of the difference ratio is greater than the third preset ratio, it may be determined that there is abnormal bill information in the real bill information, so that it may be reduced that the situation that the difference between the predicted bill information and the real bill information is smaller and is mistaken that there is abnormal bill information in the real bill information is occurred, and accuracy of bill abnormality detection is improved.

In one possible embodiment, the method further comprises:

and under the condition that the comparison result indicates that the abnormal bill information exists in the real bill information, highlighting the abnormal bill information in the real bill information.

In this embodiment, if there is abnormal bill information in the real bill information, the abnormal bill information needs to be highlighted, so that the abnormal bill information can be more intuitively reflected, so that the account applicant can quickly find the abnormal bill information, thereby facilitating checking the abnormal bill information in time, reducing the condition of bill checking errors, and improving the working efficiency of the account applicant.

In a second aspect, the present invention provides a bill auditing apparatus comprising:

the first acquisition module is used for acquiring historical bill information of the target object, wherein the historical bill information is bill information of the target object in a preset period;

the information determining module is used for determining predicted bill information of the target object based on the target prediction model library and the historical bill information, wherein the predicted bill information is bill information in a preset period after the historical bill information, and the target prediction model library comprises a plurality of target prediction models;

The second acquisition module is used for acquiring the real bill information of the target object, and the time period corresponding to the real bill information is the same as the time period corresponding to the predicted bill information;

the result determining module is used for obtaining a comparison result based on the predicted bill information and the real bill information, and the comparison result is used for indicating whether abnormal bill information exists in the real bill information.

In one possible embodiment, the apparatus further comprises:

the model training module is used for acquiring a bill information set, wherein the bill information set comprises a plurality of sample bill information, and each sample bill information comprises a plurality of sample information parameters;

preprocessing sample bill information to obtain sample bill data;

In one possible implementation, the model training module is specifically configured to:

acquiring real data in the same time period as the predicted data;

determining a model type of the first prediction model;

In a possible implementation manner, the historical bill information of the target object includes a plurality of bill information parameters, each bill information parameter includes a plurality of feature information, and the information determining module is specifically configured to:

In one possible implementation manner, the result determining module is specifically configured to:

In one possible embodiment, the apparatus further comprises:

the information display module is used for highlighting the abnormal bill information in the real bill information under the condition that the comparison result indicates that the abnormal bill information exists in the real bill information.

In a third aspect, embodiments of the present application further provide an electronic device, including: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication via the bus when the electronic device is running, the machine-readable instructions when executed by the processor performing the bill auditing method described in any one of the possible embodiments above.

In a fourth aspect, embodiments of the present application further provide a computer readable storage medium having a computer program stored thereon, which when executed by a processor performs a bill auditing method as described in any one of the possible embodiments above.

For a detailed description of the second to fourth aspects and various implementations thereof in this application, reference may be made to the detailed description of the first aspect and various implementations thereof; moreover, the advantages of the second aspect to the fourth aspect and the various implementations thereof may be referred to for analysis of the advantages of the first aspect and the various implementations thereof, and will not be described here again.

These and other aspects of the present application will be more readily apparent from the following description.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic structural diagram of a deep neural network according to an embodiment of the present application;

FIG. 2 is a flow chart of a bill auditing method according to an embodiment of the present application;

FIG. 3 is a flowchart of a method for determining forecast billing information for a target object according to an embodiment of the present application;

FIG. 4 is a flowchart of a method for determining an alignment result according to an embodiment of the present disclosure;

FIG. 5 is a flowchart of a training method for multiple target prediction models according to an embodiment of the present disclosure;

FIG. 6 is a flowchart of a method for preprocessing sample billing information according to an embodiment of the present application;

FIG. 7 is a flowchart of a method for classifying features of a sample information parameter according to an embodiment of the present disclosure;

FIG. 8 is a flowchart of a method for determining trend classification results according to an embodiment of the present application;

FIG. 9 is a flowchart of a method for training a predictive model in a predictive model library according to an embodiment of the present application;

fig. 10 is a schematic structural diagram of a bill auditing device according to an embodiment of the present application;

FIG. 11 is a schematic diagram of another bill auditing device according to an embodiment of the present application;

fig. 12 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The terms "first," "second," and the like, are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more such feature. In the description of the present invention, unless otherwise indicated, the meaning of "a plurality" is two or more.

The current bill approval work is usually to directly mark bills exceeding a certain proportion as abnormal data through the bill daily average ring ratio of two adjacent account periods and mark and lock the bills in red. However, by adopting the method for bill examination and approval, not only the condition that the bill meets the standard and is wrong can occur, but also the proportion is usually set to be too large in order to reduce the error, so that bill information which does not exceed the proportion but is still obvious in difference can not be detected, the accuracy of bill examination and approval is lower, and the working efficiency of the newspaper staff is reduced.

In view of the above problems, the present application provides a bill auditing method, including: acquiring historical bill information of a target object, wherein the historical bill information is bill information of the target object in a preset period; determining predicted bill information of the target object based on the target prediction model library and the historical bill information, wherein the predicted bill information is bill information in a preset period after the historical bill information, and the target prediction model library comprises a plurality of target prediction models; acquiring real bill information of a target object, wherein the time period corresponding to the real bill information is the same as the time period corresponding to the predicted bill information; based on the predicted bill information and the real bill information, a comparison result is obtained, and the comparison result is used for indicating whether abnormal bill information exists in the real bill information.

The execution main body of the bill auditing method provided in the embodiment of the present application is generally an electronic device or a server or other processing devices with a certain computing capability, where the electronic device may be a mobile device, a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. The server can be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and a cloud server for providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data, artificial intelligent platforms and the like.

In some possible implementations, the form auditing method may be implemented by way of a processor invoking computer readable instructions stored in memory.

Some algorithm models referred to in this application are explained below.

1. Deep neural network (Deep Neural Networks DNN)

DNN is also known as Multi-Layer perceptron (MLP). The DNN is formed by fully connecting an input layer, a hidden layer and an output layer, wherein the layers are fully connected, i.e. any neuron of the ith layer is necessarily connected with any neuron of the (i+1) th layer. The small local model is a miniature perceptron, namely a linear relation plus an activation function, the activation function is utilized for de-linearization, cross entropy and the like are used as loss functions, and a random gradient descent algorithm, a batch gradient descent algorithm and the like are utilized for carrying out learning training and adjusting and updating weights among neurons. The deep neural network comprises DNN-4, DNN-5, DNN-6, etc.

Specifically, referring to fig. 1, a schematic structural diagram of a deep neural network according to an embodiment of the present application is provided, where the schematic structural diagram includes an input layer (input layer), a hidden layer1 (hidden layer 1), a hidden layer2 (hidden layer 2), a hidden layer3 (hidden layer 3), and an output layer (output layer). In this embodiment, the deep neural network is trained using a forward propagation algorithm.

Correspondingly, the forward propagation algorithm performs a series of linear operation and activation operation by using a plurality of weight coefficient matrixes omega, offset vectors b and input value vectors x, and performs backward calculation layer by layer from an input layer until the operation reaches an output layer to obtain an output result, wherein a specific calculation formula is shown as the following (1):

wherein, sigma (·) is an activation function, m represents the number of neurons of the first-1 hidden layer,

output value of ith neuron for the 1 st-1 st hidden layer,/for the first hidden layer>

Representing weights from the f-th neuron of the first-1 th hidden layer to the j-th neuron of the first hidden layer,/for the first hidden layer>

Representing from the first- ₁ The j-th neuron offset value from one hidden layer to one hidden layer. Typical activation functions are piecewise linear functions (Rectified Linear Unit, reLU), sigmoid, hyperbolic tangent (tanh), normalized exponential (Softmax) functions.

2. Tree model

The tree model includes algorithms such as a distributed gradient lifting framework (Light Gradient Boosting Machine, lightGBM), random Forest (RF), integrated lifting algorithm framework (XGBoost), and the like. It should be noted that, the algorithms included in the tree model all belong to an ensemble learning (ensemble learni ng), and the ensemble learning is to complete learning tasks by constructing and combining a plurality of machine learners, so as to improve the accuracy of classification. The theoretical basis is derived from several important concepts in the decision tree: entropy, conditions, information gain ratio, and coefficient of kunity, etc.

Specifically, the calculation formula of the entropy is as follows (2):

wherein, the random variable X is the sample category, H (X) is entropy, n is the category total of the sample, P (X=x) _i )＝p _i ，i＝1，2，…，n，p _i Probability for each category.

Specifically, the calculation formula of the conditional entropy is as follows (3):

H(X|Y)＝∑P(X|Y) _log P(X|Y) (3)

where H (X|Y) is conditional entropy, P (X|Y) is probability of X occurrence under the condition of occurrence of class Y, and log () is a logarithmic function.

Specifically, the calculation formula of the information gain is as follows (4):

g(D，A)＝H(A)-H(D|A) (4)

where g (D, a) is the information gain, a is a feature, and D is the training dataset.

Specifically, the calculation formula of the information gain ratio is as follows (5):

Wherein g _r (D, A) is the information gain ratio.

Specifically, the calculation formula of the coefficient of ken is as follows (6):

where Gini (D) is the coefficient of kunning, k=1, 2, …, K.

The bill auditing method provided in the embodiment of the present application is described below.

Referring to fig. 2, a flowchart of a bill auditing method according to an embodiment of the present application is shown, where the method includes steps S101 to S104:

s101, acquiring historical bill information of a target object, wherein the historical bill information is bill information of the target object in a preset period.

Specifically, historical billing information for the target object may be obtained from the billing management system. Wherein, the bill management system contains all bill information of the target object. The target object refers to any resource that can be used. In this embodiment, the target object may be an energy source or an iron tower, where the energy source is a resource capable of providing energy, such as thermal energy, electric energy, and the like. The iron tower is applied to the fields of communication and the like. In other embodiments, the target object may be any item capable of being exchanged, such as a book, a beverage, and the like, which is not particularly limited.

Correspondingly, the historical bill information refers to bill information of the target object in a preset period. The preset period may be set according to practical situations, for example, may be 1 year, 1 month, 1 quarter, and the like, and is not specifically limited. For example, if the preset period is 1 year, the historical billing information is billing information within 1 year of the target object.

It will be appreciated that the historical billing information includes a plurality of billing information parameters, each billing information parameter including a plurality of characteristic information including trend characteristic information and type characteristic information. The trend characteristic information is determined according to a change trend corresponding to the same bill information parameter in the plurality of historical bill information, and the trend characteristic information can be periodical change, linear increase, exponential increase, gaussian curve, irregular change and the like, and the periodical change can also comprise annual periodical trend, seasonal periodical trend, monthly periodical trend, zhou Du periodical trend, daily periodical trend, hourly periodical trend and the like. The type characteristic information is determined according to the type corresponding to the bill information parameter, and the type characteristic information can be account period, bill identification, province, ground city, meter code, meter type, account reporting time, settlement state, data source type, meter state, allocation proportion, settlement mode, outage time and the like. For example, the classification is performed according to the city of the place, and the classification can be classified into a city, B city, C city, etc.; the power-off time is classified into 0-3 points, 4-7 points, 8-11 points, 12-15 points, 16-19 points, 20-23 points, etc. By way of example, if the historical billing information includes 0-3 points, the type characteristic information may be determined as outage time, the trend characteristic information may be a periodic trend in the year, a linear increase, and the like.

In one possible implementation, if the historical billing information is energy consumption billing information, the billing information parameters may be billing period, billing identifier, province, city, meter code, meter type, electricity fee, cost, average daily electricity, average daily cost, average daily electricity loop ratio, average daily fee loop ratio, billing time, settlement status, data source type, etc. It will be appreciated that since the power is recorded by the meter, the billing information parameters also include energy consumption meter resource information, such as meter code, meter status, split ratio, settlement mode, power down time, shared information, etc.

In one possible implementation, if the historical billing information is iron tower billing information, the billing information parameters may be billing period, billing identification, province, city, business bill number, site name, site code, iron tower category, product type, tower rental fee, machine room rental fee, maintenance fee, site fee, oil engine power generation fee, electricity introduction fee, mating rental fee, other fee, billing status, etc. It will be appreciated that since the iron tower may be affected by wind pressure, etc., the type of billing corresponding to the bill may be affected, and thus the billing information parameters include wind pressure, iron tower sharing information, iron tower type, number of product units, actual highest antenna hanging, whether the remote radio unit (Remote Radio Unit, RRU) is on the tower, total number of iron tower sharing clients, whether the baseband processing unit (Building Base band Unit, BBU) is placed in the iron tower room, room configuration, room sharing condition, maintenance fee mode, maintenance fee sharing information, maintenance fee reference price, site fee mode, site fee sharing information, site fee sharing manufacturer, power introduction fee mode, power introduction fee sharing information, power introduction fee, oil engine power generation mode, oil engine power generation service fee, power service fee, matched product, matched sharing information, whether remote and pulled party information, microwave information, wireless access point (WLAN Access Point, WLAN AP) information, extra battery guarantee, non-standard construction cost, etc.

In this embodiment, the transmitting end is an iron tower side, and the receiving end is a communication side, but the transmitting end and the receiving end are the same in transmitting information.

S102, determining predicted bill information of the target object based on a target prediction model library and historical bill information, wherein the predicted bill information is bill information in a preset period after the historical bill information, and the target prediction model library comprises a plurality of target prediction models.

Specifically, after the historical bill information is obtained, the predicted bill information of the target object may be determined according to the target prediction model library and the historical bill information. The target prediction model library includes a plurality of target prediction models, which may be tree models, deep neural networks, or other algorithms, for example, a logistic regression algorithm (Logistic Regression, LR), a time series algorithm (Auto Regressive Moving Average, ARMA), a K-Mean algorithm (K-Mean), an isolated Forest algorithm (isolation Forest), and the like, which are not particularly limited.

It should be noted that, the forecast billing information refers to billing information located within a preset period after the history billing information. For example, if the historical billing information is billing information for 1 month, the billing information is predicted to be billing information for 2 months. Wherein the predicted bill information corresponds to standard information of the bill, that is, the real bill information should be the same as the predicted bill information under normal circumstances.

Specifically, referring to fig. 3, a flowchart of a method for determining predicted bill information of a target object according to an embodiment of the present application includes the following steps S1021 to S1023:

s1021, selecting a first target prediction model matched with the characteristic information from a target prediction model library according to each characteristic information.

Specifically, for each feature information in the historical billing information, a first target prediction model matching the feature information is selected from a target prediction model library. The matching relationship between the feature information and the target prediction model is matched according to the evaluation index of the algorithm model, and the algorithm model with the best output evaluation index is determined as the first target prediction model matched with the feature information.

S1022, inputting the characteristic information into the first target prediction model to obtain the prediction bill data corresponding to the characteristic information.

Specifically, after the first target prediction model is determined, the feature information may be input into the first target prediction model to perform prediction, so as to obtain predicted bill data corresponding to the feature information.

S1023, obtaining the predicted bill information of the target object based on the predicted bill data corresponding to each piece of characteristic information.

Specifically, after obtaining the predicted bill data corresponding to the feature information, obtaining the predicted bill information of the target object according to the predicted bill data corresponding to each feature information. Therefore, each bill information parameter is input into the matched prediction model to be predicted, the accuracy of bill data prediction can be improved, a basis is provided for subsequent bill information comparison work, and whether abnormal bill information exists in real bill information can be accurately determined.

S103, acquiring real bill information of the target object, wherein the time period corresponding to the real bill information is the same as the time period corresponding to the predicted bill information.

Specifically, after obtaining the predicted bill information of the target object, the real bill information of the target object may be obtained from the bill management system. The time period corresponding to the real bill information is the same as the time period corresponding to the predicted bill information. For example, if the predicted billing information is 1 month of predicted information, the real billing information is also 1 month of real information.

In one possible implementation manner, the bill information corresponding to the bill identification which is successfully associated can be determined as the real bill information by predicting the bill identification corresponding to the bill information and associating the bill identification corresponding to the bill information in the bill management system.

And S104, based on the predicted bill information and the real bill information, obtaining a comparison result, wherein the comparison result is used for indicating whether abnormal bill information exists in the real bill information.

Specifically, after the predicted bill information and the real bill information are obtained, a comparison result can be obtained according to the predicted bill information and the real bill information. The comparison result is used for indicating whether abnormal bill information exists in the real bill information.

In a possible implementation manner, referring to fig. 4, a flowchart of a method for determining an alignment result provided in the embodiment of the present application includes the following steps S1041 to S1045:

s1041, comparing the predicted bill information with the real bill information to obtain a difference result, wherein the difference result is used for indicating whether the predicted bill information and the real bill information have differences or not.

Specifically, after obtaining the predicted bill information and the real bill information, the predicted bill information may be compared with the real bill information to obtain a difference result. Wherein the difference result is used for indicating whether a difference exists between the predicted bill information and the real bill information.

S1042, determining an absolute value of a difference value corresponding to a billing information parameter in which the predicted billing information and the real billing information differ, if the difference result indicates that there is a difference between the predicted billing information and the real billing information.

In one possible embodiment, if the difference result indicates that there is a difference between the predicted bill information and the real bill information, it is necessary to determine an absolute value of a difference ratio corresponding to a bill information parameter in which the predicted bill information and the real bill information are different. Wherein the difference ratio refers to the ratio between the difference value between the predicted bill data and the real bill data of the bill information parameter having the difference and the predicted bill data.

For example, if the predicted bill data is 100 and the real bill data is 150, it may be determined that the difference value between the predicted bill data and the real difference data is-50, the difference ratio is-50%, and the absolute value of the difference ratio is 50%; if the predicted bill data is 100 and the real bill data is 40, it may be determined that the difference value between the predicted bill data and the real difference data is 60, the difference ratio is 60%, and the absolute value of the difference ratio is still 60%. It will be appreciated that since the predicted bill data may be higher or lower than the real bill data, i.e. the difference value between the predicted bill data and the real difference data may be positive or negative, it is necessary to determine the absolute value of the difference ratio to further determine whether the difference ratio exceeds the third preset ratio.

S1043, determining whether the absolute value of the difference proportion is larger than a third preset proportion; if yes, go to step S1044; if not, step S1045 is performed.

Specifically, after determining the absolute value of the difference proportion, determining whether the absolute value is greater than a third preset proportion, if the absolute value is greater than the third preset proportion, determining that abnormal bill information exists in the real bill information; if the absolute value is not greater than the third preset proportion, it can be determined that abnormal bill information does not exist in the real bill information. The third preset proportion can be set in actual situations. In this embodiment, the third preset proportion is 50%. In other embodiments, the third preset ratio may be 40%, 20%, or the like.

S1044, determining that abnormal bill information exists in the real bill information.

Specifically, after determining that the absolute value of the difference ratio is greater than the third preset ratio, it may be determined that abnormal bill information exists in the real bill information. Therefore, the occurrence of the situation that abnormal bill information exists in the real bill information due to the fact that the difference between the predicted bill information and the real bill information is smaller can be reduced, and the accuracy of bill abnormal detection is improved.

In one possible implementation, if the comparison result indicates that there is abnormal bill information in the real bill information, the abnormal bill information in the real bill information is highlighted. The highlighting may be performed by thickening, highlighting, coloring the remaining fonts, and the like. For example, if the real billing information includes information of S province, a city, 1 year, etc., and the S province can be determined to be abnormal billing information according to the comparison result, the S province can be subjected to red marking processing. So, can more audio-visual embody unusual bill information for the account reigning personnel can be quick discover unusual bill information, in order to in time recheck unusual bill information, reduce the wrong condition of bill audit, promoted account reigning personnel's work efficiency.

In this embodiment, the rechecking personnel include grid personnel, equipment responsible personnel, related opposite-end system maintenance personnel, and the like. Wherein, grid personnel refer to personnel who are inspecting in a limited range. The equipment responsible personnel refer to personnel managing equipment such as meters, towers and the like. The related maintenance personnel of the opposite-end system refer to personnel for maintaining the systems such as the iron tower side, the communication side and the like.

In another possible implementation, the audit report information is also displayed if the comparison indicates that there is abnormal billing information in the real billing information. The audit report information comprises account period, audit time, bill unique identification, meter code, current daily average cost value, daily average cost marker post value, daily average cost forecast value and audit result description. The daily average cost target value refers to the daily average cost under normal conditions. The audit description is used to indicate a specific percentage value that exceeds a third predetermined proportion. Therefore, the auditing result can be displayed more intuitively.

S1045, determining that abnormal bill information does not exist in the real bill information.

Specifically, after determining that the absolute value of the difference ratio is not greater than the third preset ratio, it may be determined that no abnormal bill information exists in the real bill information.

In a possible implementation manner, before the historical bill information of the target object is obtained, a plurality of target prediction models need to be trained respectively, specifically, referring to fig. 5, a flowchart of a method for training a plurality of target prediction models provided in an embodiment of the present application is shown, and the method includes the following steps S201 to S204:

s201, a billing information set is acquired, where the billing information set includes a plurality of sample billing information, and each sample billing information includes a plurality of sample information parameters.

Specifically, the bill information set includes a plurality of sample bill information, and by training different sample bill information, a target prediction model with stronger generalization performance can be learned. The sample bill information comprises a plurality of sample information parameters.

Alternatively, the sample billing information may obtain the historical data directly from the database. In this embodiment, the database is a relational database service (Relational Database Service, RDS). In other embodiments, a relational database management system (MySQL) may be used, which is not particularly limited.

Optionally, the sample bill information may obtain peripheral system data from a resource system, a network management system, and the like, and transmit the peripheral system data to a database through an interface; and/or the sample billing information may directly obtain the historical data through the billing management system.

Optionally, the sample bill information may collect the historical data through a meter, an iron tower, and other devices, specifically, the device is verified first, under the condition that the device state is normal, it is determined whether the bill index participating in the operation needs to be optimized, if the bill index participating in the operation does not need to be optimized, the historical data is directly collected, and if the bill index participating in the operation needs to be optimized, the historical index needs to be extracted, and the historical data corresponding to the historical index needs to be adjusted or deleted. The bill indexes participating in the operation comprise resource index detection subclasses, starting conditions of cross detection, influence conditions of seasons and months, corresponding adjustment factors, judging conditions of resource state transition and the like.

Specifically, the resource index detection subclass is an operation factor for determining whether to influence the corresponding type of billing fees according to whether the aforementioned resource indexes such as wind pressure participate in operation comparison. The starting condition of the cross detection is that the pointers are used for renting the electricity fee without towers, renting the electricity fee without towers and the like, whether the states of equipment such as a meter, an iron tower and the like are normal needs to be determined, and if the states of the equipment are normal, the sample bill information can be described as an abnormal bill. The influence of seasons and months refers to determining whether environmental factors such as seasons and months affect the cost, for example, the use time of an air conditioner in summer is longer than that of an air conditioner in spring and autumn, so that the electricity fee expenditure in summer is greater than that in spring and autumn. The corresponding adjustment factors are used for adjusting multiplying power for training big data according to different seasons so as to be closer to real data. The determination of the resource state transition refers to determining whether the sample bill information is an abnormal bill according to the renting condition of the iron tower and the use condition of the meter, for example, if the iron tower is rented back and the meter is deactivated, the related abnormal bill scene can be eliminated.

It should be noted that, the bill index participating in the operation may be changed and optimized according to the related parameters after multiple models, so as to make the processing logic and the prediction data more reasonable.

In a possible implementation manner, the historical data can be directly determined as sample bill information, or the historical data can be deleted and combined according to a preset collection rule and collection range, and the processed data is converted and mapped to obtain the sample bill information, so that the predicted data obtained through training of the target prediction model is more accurate. The preset collection rules and collection ranges are preset according to the account period and the warehouse-in frequency table of the bill.

S202, preprocessing sample bill information to obtain sample bill data.

Specifically, after the sample billing information is obtained, the sample billing information needs to be preprocessed to obtain sample billing data.

In one possible implementation, referring to fig. 6, a flowchart of a method for preprocessing sample bill information provided in the embodiment of the present application includes the following steps S2021 to S2025:

s2021, a null value rate and an outlier rate of the sample bill information are determined.

Specifically, after the sample billing information is acquired, the null value rate and the outlier rate of the sample billing information need to be determined. Wherein the null rate refers to a ratio between the number of sample information parameters having no numerical value in the sample billing information and the total number of sample information parameters. For example, if the total number of sample information parameters is 10 and the number of sample information parameters without values is 4, it may be determined that the null rate of the sample billing information is 40%.

Accordingly, the outlier ratio refers to a ratio between the number of sample information parameters of outliers in the sample billing information and the total number of sample information parameters. For example, if the total number of sample information parameters is 10 and the number of sample information parameters of an outlier value is 5, it may be determined that the outlier rate of the sample billing information is 50%.

S2022, determining whether the null rate is smaller than a first preset proportion and determining whether the abnormal value rate is smaller than a second preset proportion; if yes, go to step S2023; if not, step S2024 is performed.

For example, if the null rate is higher than the first preset proportion and/or the outlier rate is higher than the second preset proportion, the sample bill information may be directly deleted; if the null value rate is not higher than the first preset proportion and the abnormal value rate is not higher than the second preset proportion, the sample bill information needs to be filled with information. The first preset proportion and the second preset proportion are preset according to actual conditions, and the first preset proportion and the second preset proportion can be the same or different and are not limited in detail. In this embodiment, the first preset proportion and the second preset proportion are both 50%.

And S2023, deleting the sample bill information.

Specifically, if the null rate is higher than the first preset proportion, it is indicated that the bill data in the sample bill information is too much lack and has no referential, and at the same time, the training result of the prediction model may be disturbed; if the abnormal value rate is higher than the second preset proportion, the abnormal bill data in the sample bill information is too much, the abnormal bill data is also not referenced, and the training result of the prediction model is possibly disturbed, so that the sample bill information with the null value rate higher than the first preset proportion and/or the sample bill information with the abnormal value rate higher than the second preset proportion need to be deleted, the training of the prediction model, which is influenced by the lack of excessive bill data or the excessive abnormal bill data, is reduced, and the accuracy of bill information prediction is further improved.

S2024, performing information filling processing on the sample bill information to obtain target sample bill information.

Specifically, after it is determined that the null value rate is not higher than a first preset proportion and the outlier rate is not higher than a second preset proportion, information filling processing is required to be performed on the sample bill information so as to obtain target sample bill information. For example, default values for mean, mode, median, etc. may be populated into the sample billing information. In this way, the integrity of the sample billing information may be ensured.

And S2025, performing information processing on the target sample bill information to obtain sample bill data.

Specifically, after the target sample bill information is obtained, information processing needs to be performed on the target sample bill information. The information processing refers to a process of digitizing character type information, for example, processing character type data of a checkout time into time differences, and performing One-Hot (One-Hot) encoding on data of character types such as a settlement state. Therefore, convenience can be provided for subsequent model training work, and the accuracy of bill information prediction is improved.

S203, carrying out feature classification on each sample information parameter based on a preset feature classification library to obtain a type classification result and a trend classification result of each sample information parameter, wherein the preset feature classification library comprises type features and trend features, the type classification result comprises type feature data, and the trend classification result comprises at least one trend feature data.

Specifically, after the sample bill data is obtained, feature classification is required to be performed on each sample information parameter in the bill information set according to a preset feature classification library. The preset feature classification library comprises type features and trend features. The type feature refers to a feature that can be directly determined by the type corresponding to the sample information parameter. Trend characteristics refer to characteristics determined by the trend of variation of the same sample information parameter between different sample billing data.

Illustratively, the type of feature includes a feature billing period, bill identification, province, municipality, meter code, meter type, time of statement, settlement status, data source type, meter status, apportionment ratio, settlement mode, power down time, outage time, and the like. The trend features include periodic variation trend, linear increase, exponential increase, gaussian curve, irregular variation, etc. Among these, the periodic variation trend includes an annual periodic trend, a seasonal periodic trend, a monthly periodic trend, a Zhou Du periodic trend, a daily periodic trend, an hourly periodic trend, and the like.

It will be appreciated that since the type feature is directly determined according to the type to which the sample information parameter corresponds, the type classification result includes only one type feature data. However, the trend feature is determined according to the variation trend of the same sample information parameter between different sample bill data, so the type classification result may include at least one trend feature data, for example, the daily average electric quantity feature satisfies both significant seasonal periodic variation and certain periodic variation, that is, the daily average electric quantity of monday through friday is relatively stable, but the daily average electric quantity of friday is significantly increased, and the like.

In a possible implementation manner, referring to fig. 7, a flowchart of a method for classifying characteristics of a sample information parameter according to an embodiment of the present application is shown, and includes the following steps S2031 to S2032:

s2031, respectively matching each sample information parameter with the type features in the preset feature classification library to obtain a type classification result of each sample information parameter.

Specifically, after the sample bill data is obtained, each sample information parameter in the sample bill data needs to be matched with the type feature in the preset feature classification library so as to obtain the type classification result of each sample information parameter. For example, if the sample information parameter is a city, it may be determined that the type classification result is a city. In this way, the accuracy of the type classification result determination can be improved.

In a possible implementation manner, feature matching can be performed in a preset feature classification library according to the english name corresponding to each sample information parameter, so that the matching degree of the type features can be improved, and the type classification result is more accurate.

S2032, determining a trend classification result of each sample information parameter based on the same sample information parameter in the plurality of sample bill data and the trend feature in the preset feature classification library.

Specifically, after obtaining the same sample information parameter in the plurality of sample bill data, the trend classification result may be determined according to the same sample information parameter in the plurality of sample bill data and the trend feature in the preset feature classification library. Therefore, the accuracy of determining the trend classification result can be improved, so that the training accuracy of the prediction model is improved, and the prediction result obtained through the prediction model is closer to the real result.

In one possible implementation manner, a change trend of the sample information parameter may be determined according to a value corresponding to the same sample information parameter in the plurality of sample bill data, and the change trend may be matched with a trend feature in a preset feature classification library, so as to obtain a trend classification result. For example, the values corresponding to the same sample information parameter in the plurality of sample bill data are linearly increased, and the preset feature classification library includes a trend feature of linear increase, so that the linear increase can be determined as a trend classification result.

In another possible implementation manner, for the periodical change trend, calculation needs to be performed on the data in the period according to different index rules to determine a trend classification result, specifically, as shown in fig. 8, a flowchart of a method for determining a trend classification result provided in the embodiment of the present application includes the following steps S20321 to S20324:

S20321, for each trend feature, calculating the same sample information parameter in the plurality of sample bill data according to different index rules, to obtain a plurality of index values corresponding to the different index rules.

Specifically, after the sample bill data is obtained, the same sample information parameters in the sample bill data in the period need to be calculated according to different index rules respectively aiming at different trend characteristics. Wherein the index rule comprises a maximum value, a minimum value, a mean value, a variance, a mode, a multi-level quantile and the like. It will be appreciated that each rule indicator corresponds to an indicator value.

S20322, performing error matching on the plurality of index values, respectively, to obtain a plurality of error values.

Specifically, after obtaining a plurality of index values, error matching needs to be performed on the plurality of index values respectively to obtain a plurality of error values.

In one possible implementation manner, a standard value corresponding to the trend feature, the sample information parameter and the index rule may be obtained from a preset feature classification library, and the difference between the standard value and the index value is determined as an error value. It will be appreciated that the standard value is predetermined.

S20323, determining a target number of the plurality of error values that is lower than a preset value, and determining whether the target number exceeds the preset number.

Specifically, after determining the plurality of error values, it is necessary to further determine a target number of the plurality of error values that is lower than a preset value, and after determining the target number, it is necessary to determine whether the target number exceeds the preset number. The preset values and the preset numbers are preset according to actual conditions. It will be appreciated that the preset values of the error values corresponding to the different index rules are different.

S20324, in the case where the target number exceeds the preset number, the trend classification result is determined based on the trend feature.

Specifically, after determining the target number, it needs to determine whether the target number exceeds a preset number, and if the target number exceeds the preset number, the trend feature may be determined as a trend classification result; if the target number does not exceed the preset number, the trend feature cannot be determined as a trend classification result. For example, by using 6 index rules for illustration, calculation is performed according to the 6 index rules, so that 6 index values can be obtained, and further, 6 error values can be obtained. If the target number lower than the preset number in the 6 error values is 5 and the preset number is 4, the trend feature can be determined as a trend classification result. Therefore, the accuracy of determining the trend classification result can be improved, and the training accuracy of the prediction model is further improved.

S204, respectively training a plurality of prediction models in a prediction model library based on the type classification result and the trend classification result of each sample information parameter to obtain a plurality of target prediction models.

Specifically, after the type classification result and the trend classification result are obtained, a plurality of prediction models in the prediction model library can be respectively trained for each feature data in the type classification result and the trend classification result, so as to obtain a plurality of target prediction models. Wherein, a plurality of prediction models in the prediction model library are pre-constructed prediction models. The prediction model library comprises a deep neural network, a tree model and other algorithms.

Correspondingly, a plurality of prediction models in the prediction model library can be divided into supervised algorithm models, unsupervised algorithm models such as k-means models, isolated forest models and the like according to the existence of labels, the types of the labels are divided into classification algorithm models and regression algorithm models, and the classification algorithm models are divided into two classification models and multiple classification models. Therefore, the prediction model library is not only built based on labels, but also takes classification of features as a basis for building.

In a possible implementation manner, referring to fig. 9, a flowchart of a method for training a prediction model in a prediction model library according to an embodiment of the present application is shown, including the following S2041 to S2044:

S2041, for each feature data in the sample information parameter, determining a first prediction model matching the feature data from the prediction model library, the feature data being type feature data or trend feature data.

Specifically, after the type classification result and the trend classification result of each sample information parameter are obtained, a first prediction model matched with the feature data may be determined from the prediction model library. The feature data is type feature data or trend feature data.

In one possible implementation, before determining the first predictive model that matches the feature data, a plurality of predictive models that are more sensitive to the feature data and that are more efficient in performance need to be obtained from a library of predictive models. And then, respectively carrying out model self-training and test verification on the plurality of prediction models according to the characteristic data so as to determine evaluation indexes such as AUC (automatic Power control), accuracy, precision, recall rate, F1 value, model stability and the like among the models, and analyzing according to index values corresponding to the evaluation indexes so as to obtain a first prediction model with optimal fitting degree. The AUC is the area under the ROC curve (receiver operating characteristic curve, subject working characteristic curve) and is used for measuring the quality of the learner.

S2042, inputting the feature data into the first prediction model to obtain prediction data, wherein the prediction data is adjacent feature data positioned in a preset period after the feature data.

Specifically, after determining the first prediction model, the feature data may be input to the first prediction model for training to obtain the prediction data. The prediction data is adjacent characteristic data located in a preset period after the characteristic data. For example, if the feature data is 1 month data, the predicted data is 2 months data.

In one possible implementation, the model type of the first predictive model needs to be determined before training the first predictive model; if the first prediction model is a tree model, the feature data can be directly input into the first prediction model for training, so that the training process can be simplified under the condition that the training result of the first prediction model is not influenced; if the first prediction model is a neural network, the feature data needs to be standardized, and the processed feature data is input into the first prediction model for training, so that the robustness of the target prediction model can be improved.

It can be understood that, as the weak classifier, the decision tree can make the tree models such as Lightgbm, xgboost have better interpretation and robustness, and can automatically find the higher-order relation between the features, and special normalization and other treatments are not needed for the data, but the models such as DNN, LR and the like are insensitive to the higher-order relation between the features and cannot be automatically found and treated, so that the minimum-maximum normalization (0-1 normalization), zero-average normalization (z-score normalization) and other treatments are needed for the feature dataset of the input prediction model, so that the training precision of the prediction model is improved.

S2043, the real data within the same period of time as the predicted data is acquired.

Specifically, after obtaining the predicted data, the real data within the same period of time as the predicted data may be obtained from the bill information set.

And S2044, carrying out parameter adjustment on the first prediction model based on the real data and the prediction data to obtain a target prediction model.

Specifically, after the real data is obtained, parameter adjustment can be performed on the first prediction model to be trained according to the real data and the prediction data until a trained target prediction model is obtained. Therefore, each piece of characteristic data corresponds to one prediction model, the occurrence of inaccurate prediction results caused by mismatching of the prediction model and the characteristic data is reduced, and the accuracy of determination of the prediction data is improved.

It will be appreciated by those skilled in the art that in the above-described method of the specific embodiments, the written order of steps is not meant to imply a strict order of execution but rather should be construed according to the function and possibly inherent logic of the steps.

Based on the same inventive concept, the present embodiment also provides a bill auditing device corresponding to the bill auditing method, and since the principle of solving the problem by the device in the embodiment of the present application is similar to that of the bill auditing method in the embodiment of the present application, the implementation of the device can refer to the implementation of the method, and the repetition is omitted.

Referring to fig. 10, a schematic structural diagram of a bill auditing device according to an embodiment of the present application is shown, where the bill auditing device 1000 includes:

the first obtaining module 1001 is configured to obtain historical billing information of the target object, where the historical billing information is billing information of the target object in a preset period;

the information determining module 1002 is configured to determine predicted bill information of the target object based on a target prediction model library and historical bill information, where the predicted bill information is bill information within a preset period after the historical bill information, and the target prediction model library includes a plurality of target prediction models;

a second obtaining module 1003, configured to obtain real bill information of the target object, where a period of time corresponding to the real bill information is the same as a period of time corresponding to the predicted bill information;

the result determining module 1004 is configured to obtain a comparison result based on the predicted bill information and the real bill information, where the comparison result is used to indicate whether abnormal bill information exists in the real bill information.

Referring to fig. 11, in one possible embodiment, the bill auditing device 1000 further includes:

a model training module 1005 for obtaining a billing information set, the billing information set comprising a plurality of sample billing information, each sample billing information comprising a plurality of sample information parameters;

Preprocessing sample bill information to obtain sample bill data;

In one possible implementation, the model training module 1005 is specifically configured to:

acquiring real data in the same time period as the predicted data;

determining a model type of the first prediction model;

In a possible implementation manner, the historical billing information of the target object includes a plurality of billing information parameters, each billing information parameter includes a plurality of feature information, and the information determining module 1002 is specifically configured to:

In one possible implementation, the result determining module 1004 is specifically configured to:

In one possible embodiment, the bill auditing apparatus 1000 further includes:

the information display module 1006 is configured to highlight the abnormal bill information in the real bill information when the comparison result indicates that the abnormal bill information exists in the real bill information.

The process flow of each module in the apparatus and the interaction flow between the modules may be described with reference to the related descriptions in the above method embodiments, which are not described in detail herein.

Based on the same technical concept, the embodiment of the application also provides electronic equipment. Referring to fig. 12, a schematic structural diagram of an electronic device 1200 according to an embodiment of the present application includes a processor 1201, a memory 1202 and a bus 1203. The memory 1202 is used for storing execution instructions, including a memory 12021 and an external memory 12022; the memory 12021 is also referred to as an internal memory, and is used for temporarily storing operation data in the processor 1201 and data exchanged with the external memory 12022 such as a hard disk, and the processor 1201 exchanges data with the external memory 12022 via the memory 12021.

In the embodiment of the present application, the memory 1202 is specifically configured to store application program codes for executing the solution of the present application, and the processor 1201 controls the execution. That is, when the electronic device 1200 is running, communication between the processor 1201 and the memory 1202 is through the bus 1203, causing the processor 1201 to execute the application code stored in the memory 1202, thereby performing the methods disclosed in any of the foregoing embodiments.

The Memory 1202 may be, but is not limited to, random access Memory (Random Acces sMemory, RAM), read Only Memory (ROM), programmable Read Only Memory (Programmable Read-Only Memory, PROM), erasable Read Only Memory (Erasa ble Programmable Read-Only Memory, EPROM), electrically erasable Read Only Memory (El ectric Erasable Programmable Read-Only Memory, EEPROM), etc.

The processor 1201 may be an integrated circuit chip having signal processing capabilities. The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; but also digital signal processors (Digi tal Signal Processing, DSP), application specific integrated circuits (Application SpecificInt egrated Circuit, ASIC), field programmable gate arrays (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The disclosed methods, steps, and logic blocks in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

It is to be understood that the structure illustrated in the embodiments of the present application does not constitute a specific limitation on the electronic device 1200. In other embodiments of the present application, electronic device 1200 may include more or fewer components than shown, or may combine certain components, or split certain components, or a different arrangement of components. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

The present application also provides a computer readable storage medium, on which a computer program is stored, which when executed by a processor performs the bill auditing method described in the above method embodiments. Wherein the storage medium may be a volatile or nonvolatile computer readable storage medium.

The embodiments of the present application further provide a computer program product, where the computer program product carries program code, where instructions included in the program code may be used to perform the bill auditing method described in the foregoing method embodiments, and specifically reference may be made to the foregoing method embodiments, which are not described herein.

Wherein the above-mentioned computer program product may be realized in particular by means of hardware, software or a combination thereof. In an alternative embodiment, the computer program product is embodied as a computer storage medium, and in another alternative embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), or the like.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system and apparatus may refer to corresponding procedures in the foregoing method embodiments, which are not described herein again. In the several embodiments provided in the present disclosure, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present disclosure may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer readable storage medium executable by a processor. Based on such understanding, the technical solution of the present disclosure may be embodied in essence or a part contributing to the prior art or a part of the technical solution, or in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the embodiments of the present disclosure. And the aforementioned storage medium includes: various media capable of storing program codes, such as a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk.

Finally, it should be noted that: the foregoing examples are merely specific embodiments of the present disclosure, and are not intended to limit the scope of the disclosure, but the present disclosure is not limited thereto, and those skilled in the art will appreciate that while the foregoing examples are described in detail, it is not limited to the disclosure: any person skilled in the art, within the technical scope of the disclosure of the present disclosure, may modify or easily conceive changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features thereof; such modifications, changes or substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the disclosure, and are intended to be included within the scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.

Claims

1. A method for auditing bills, comprising:

determining predicted bill information of the target object based on a target predicted model library and the historical bill information, wherein the predicted bill information is bill information in the preset period after the historical bill information, and the target predicted model library comprises a plurality of target predicted models;

acquiring real bill information of the target object, wherein the time period corresponding to the real bill information is the same as the time period corresponding to the predicted bill information;

and obtaining a comparison result based on the predicted bill information and the real bill information, wherein the comparison result is used for indicating whether abnormal bill information exists in the real bill information.

2. The bill auditing method according to claim 1, wherein the plurality of target predictive models are trained by:

obtaining a billing information set, the billing information set comprising a plurality of sample billing information, each sample billing information comprising a plurality of sample information parameters;

Preprocessing the sample bill information to obtain sample bill data;

aiming at each sample information parameter, carrying out feature classification on each sample information parameter based on a preset feature classification library to obtain a type classification result and a trend classification result of each sample information parameter, wherein the preset feature classification library comprises type features and trend features, the type classification result comprises type feature data, and the trend classification result comprises at least one trend feature data;

and training a plurality of prediction models in a prediction model library respectively based on the type classification result and the trend classification result of each sample information parameter to obtain a plurality of target prediction models.

3. The bill auditing method according to claim 2, wherein the preprocessing the sample bill information to obtain sample bill data includes:

And carrying out information processing on the target sample bill information to obtain the sample bill data.

4. A bill auditing method according to claim 3, in which the method further comprises:

5. The bill auditing method according to claim 2, wherein the feature classification is performed on each sample information parameter based on a preset feature classification library to obtain a type classification result and a trend classification result of each sample information parameter, including:

respectively matching each sample information parameter with the type features in the preset feature classification library to obtain a type classification result of each sample information parameter;

and determining a trend classification result of each sample information parameter based on the same sample information parameter in the plurality of sample bill data and the trend characteristics in the preset characteristic classification library.

6. The bill auditing method according to claim 5, wherein the determining the trend classification result for each sample information parameter based on the same sample information parameter in the plurality of sample bill data and trend features in the preset feature classification library includes:

respectively carrying out error matching on the index values to obtain a plurality of error values;

and determining the trend classification result based on the trend characteristic under the condition that the target number exceeds the preset number.

7. The bill auditing method according to claim 2, wherein the training a plurality of prediction models in a prediction model library based on the type classification result and the trend classification result of each sample information parameter to obtain the plurality of target prediction models includes:

determining a first prediction model matched with the characteristic data from the prediction model library for each characteristic data in the sample information parameters, wherein the characteristic data is the type characteristic data or the trend characteristic data;

inputting the characteristic data into the first prediction model to obtain prediction data, wherein the prediction data is adjacent characteristic data positioned in the preset period after the characteristic data;

Acquiring real data in the same time period as the predicted data;

and carrying out parameter adjustment on the first prediction model based on the real data and the prediction data to obtain the target prediction model.

8. The bill auditing method according to claim 7, wherein the inputting the feature data into the first predictive model results in predictive data, comprising:

determining a model type of the first predictive model;

and when the first prediction model is a tree model, inputting the characteristic data into the first prediction model to obtain the prediction data.

9. The bill auditing method according to claim 8, further comprising:

and inputting the processed characteristic data into the first prediction model to obtain the prediction data.

10. The bill auditing method of claim 1, wherein the historical bill information for the target object includes a plurality of bill information parameters, each bill information parameter including a plurality of feature information, the determining the predicted bill information for the target object based on a target prediction model library and the historical bill information comprising:

Selecting a first target prediction model matched with the characteristic information from the target prediction model library aiming at each characteristic information;

inputting the characteristic information into the first target prediction model to obtain predicted bill data corresponding to the characteristic information;

11. The bill auditing method according to claim 1, wherein the obtaining a comparison result based on the predicted bill information and the real bill information includes:

comparing the predicted bill information with the real bill information to obtain a difference result, wherein the difference result is used for indicating whether a difference exists between the predicted bill information and the real bill information;

determining an absolute value of a difference ratio corresponding to a bill information parameter in which the predicted bill information and the real bill information are different in case that the difference result indicates that the predicted bill information and the real bill information are different;

And under the condition that the absolute value of the difference proportion is larger than the third preset proportion, determining that the abnormal bill information exists in the real bill information.

12. The bill auditing method according to claim 1, further comprising:

13. A bill auditing device, comprising:

the first acquisition module is used for acquiring historical bill information of a target object, wherein the historical bill information is bill information of the target object in a preset period;

the information determining module is used for determining predicted bill information of the target object based on a target predicted model library and the historical bill information, wherein the predicted bill information is bill information in the preset period after the historical bill information, and the target predicted model library comprises a plurality of target predicted models;

14. An electronic device, comprising: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication over the bus when the electronic device is running, the machine-readable instructions when executed by the processor performing the bill auditing method according to any of claims 1 to 12.

15. A computer readable storage medium, characterized in that the computer readable storage medium has stored thereon a computer program which, when executed by a processor, performs a bill auditing method according to any of claims 1 to 12.