CN117876134A - Declaration sample output method, equipment and medium - Google Patents

Abstract

The embodiment of the specification discloses a declaration sample output method, equipment and medium, comprising the following steps: extracting each business item sample from a plurality of real reporting samples; according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample; constructing a business item index model according to the calculation model of each business item sample; and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample. The declaration sample output method of the embodiment of the specification can simplify the filling process, improve the accuracy and increase the flexibility of calculation.

Description

Declaration sample output method, equipment and medium

Technical Field

The present disclosure relates to the field of computer technologies, and in particular, to a method, an apparatus, and a medium for outputting a declaration sample.

Background

Tax payment declaration is an important job for enterprises and individuals to declare tax payment to tax authorities according to national tax laws. At present, tax payment declaration data is required to be calculated through calculation formulas among the tabulated tool tables and in the tables. Specifically, the tax payer fills out relevant information according to the form specified by the tax authority, and then calculates tax declaration data through calculation formulas among and in the form.

However, there are some problems with the current tax return. First, filling out the form is cumbersome and complex, requiring the tax payer to spend a great deal of time and effort in performing the calculations and checks. This not only increases the tax of the tax payer, but also easily causes filling errors and omission, bringing unnecessary trouble to the tax payer and tax authorities. Secondly, the complexity of the computational formulas between and within the tables also presents a hurdle to the tax payer. Because the conditions of different tax types and different tax payers are different, the calculation formulas between the tables and in the tables need to be adjusted and applied according to specific conditions, which is a difficult problem for tax payers without professional knowledge.

Disclosure of Invention

One or more embodiments of the present disclosure provide a method, apparatus, and medium for outputting a declaration sample, which are used to solve the technical problems set forth in the background art.

One or more embodiments of the present disclosure adopt the following technical solutions:

one or more embodiments of the present disclosure provide a method for outputting a declaration sample, including:

extracting each business item sample from a plurality of real reporting samples;

according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample;

constructing a business item index model according to the calculation model of each business item sample;

and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.

The above-mentioned matters of the embodiments of the present disclosure have the following beneficial effects:

simplifying the filling process: by extracting each service item sample from a plurality of real reporting samples and learning a calculation model of each service item sample, the tax payer can multiplex the service items of the real reporting samples by applying the common characteristics in the learned calculation model to the specific service items of the actual samples. Thus, the tax payer can complete calculation and check more simply and rapidly when filling out the declaration form, and the complexity and complexity of filling out are reduced.

Accuracy is improved: by learning the calculation model of each business item sample and constructing the business item index model, the declaration data can be calculated more accurately. For tax payers without expert knowledge, the complexity of calculation formulas among tables and in tables can cause errors and omission, and the adoption of the method can reduce the occurrence of the problems and improve the accuracy of tax declaration.

Increasing the flexibility of the computation: the calculation modes of various tax types are obtained in advance and are applied to calculation models of different business item samples, so that the calculation modes can be flexibly adapted to different tax types and different tax payers. Therefore, tax payers do not need to independently process the adjustment and application problems of calculation formulas among tables and in tables, dependency on professional knowledge is reduced, and the flexibility of calculation is improved.

In summary, the declaration sample output method can simplify the filling process, improve the accuracy and increase the flexibility of calculation.

Further, one or more embodiments of the present disclosure provide for extracting each business item sample from a plurality of real declaration samples, including:

respectively calculating entropy values and information gains for the features in the plurality of real reporting samples, and matching an optimal feature combination;

Dividing the plurality of real declaration samples according to the optimal feature combination to form an accurate decision tree;

and extracting the business item samples according to the accurate decision tree.

It should be noted that, the above method for extracting each business item sample from multiple real declaration samples has the following specific beneficial effects:

the accuracy of sample extraction is improved: by carrying out entropy calculation and information gain analysis on the characteristics in a plurality of real declaration samples, the optimal characteristic combination can be matched, so that the accuracy of sample extraction is improved. Therefore, the extracted business item sample can reflect the characteristics and rules of the real declaration data more truly, and a more accurate data basis is provided for the subsequent calculation model construction.

Further, in one or more embodiments of the present disclosure, calculating entropy values and information gains for features in the plurality of real declaration samples, respectively, and matching an optimal feature combination includes:

for the features in the plurality of real declaration samples, respectively calculating entropy values of the features, wherein the entropy values are used for measuring the confusion degree or uncertainty of the features in the samples;

calculating information gain of each feature according to entropy values of the features for the features in the plurality of real declaration samples, wherein the information gain measures the degree of uncertainty reduction of the result under given feature conditions;

And selecting the feature with the maximum information gain as the optimal feature combination according to the information gain of each feature.

It should be noted that, the method for calculating entropy and information gain for the features in the multiple real declaration samples and matching the optimal feature combination has the following specific beneficial effects:

determining an optimal feature combination: by calculating entropy values and information gains of features in a plurality of real declaration samples, the degree of confusion of the features in the samples and the degree of influence of the features on the uncertainty of results can be evaluated. By selecting the feature with the greatest information gain, an optimal combination of features, i.e. the feature with the greatest impact on the result, can be determined.

The accuracy of feature selection is improved: by calculating the entropy value and the information gain, the influence of each feature on the result can be objectively evaluated. The feature with the maximum information gain is selected as the optimal feature combination, so that the feature with the greatest influence on the result can be more accurately selected, and an accurate basis is provided for subsequent sample segmentation and decision tree construction.

Optimizing sample segmentation and decision tree construction: by selecting the optimal feature combination, a plurality of real declaration samples can be segmented to form an accurate decision tree. The accurate decision tree can better reflect the characteristics and rules of the real declaration data, and provides a more accurate and reliable basis for the subsequent calculation model construction and business item sample extraction.

In summary, the method for calculating the entropy and the information gain of the features in the plurality of real declaration samples and matching the optimal feature combination can determine the optimal feature combination, improve the accuracy of feature selection and optimize sample segmentation and decision tree construction.

Further, in one or more embodiments of the present disclosure, the dividing the plurality of real declaration samples according to the optimal feature combination to form an accurate decision tree includes:

performing first segmentation on a plurality of real reporting samples according to the optimal feature combination to form an initial decision tree;

circularly segmenting a plurality of real declaration samples according to the optimal feature combination, and judging whether to continue segmentation according to predefined conditions, wherein the predefined conditions comprise a predefined maximum depth or a predefined sample number;

and if the predefined condition is met, forming the accurate decision tree.

It should be noted that, the forming of the accurate decision tree has the following specific beneficial effects:

the accuracy of the decision tree is improved: by using the optimal feature combination to divide a plurality of real declaration samples, the nodes and branches of the decision tree can be ensured to accurately capture the features and rules of the real declaration data. In this way, the formed accurate decision tree can more accurately predict the business item data of the current declaration sample, and reduce the possibility of filling errors and omission.

Simplifying the decision tree construction process: by specifying predefined conditions, such as a predefined maximum depth or number of samples, the complexity and size of the decision tree can be controlled. In this way, the construction process of the decision tree can be simplified, unnecessary segmentation and nodes are reduced, and the interpretability and usability of the decision tree are improved.

The segmentation efficiency is improved: the efficiency of segmentation can be improved by performing the cyclic segmentation using the optimal feature combination and judging whether to continue the segmentation according to the predefined condition. Only when the predefined condition is met, the next round of segmentation is performed, unnecessary calculation and processing are avoided, and time and resources are saved.

In summary, the plurality of real declaration samples are segmented according to the optimal feature combination to form an accurate decision tree, so that the accuracy of the decision tree can be improved, the decision tree construction process is simplified, and the segmentation efficiency is improved.

Further, according to one or more embodiments of the present disclosure, extracting the business transaction samples according to the accurate decision tree includes:

inputting a plurality of real reporting samples into an accurate decision tree to determine service matters to which each real reporting sample belongs according to the path of each real reporting sample on the accurate decision tree;

And extracting each real reporting sample according to the business item to which each real reporting sample belongs so as to extract the real reporting sample belonging to the specific business item from a plurality of real reporting sample sets to form each business item sample.

It should be noted that the following specific beneficial effects are provided with respect to the above matters:

automatic classification: the classification of a plurality of real declaration samples can automatically identify different business matters without manual intervention, so that the risk of manual errors is reduced, and the classification accuracy and efficiency are improved.

Accurately extracting a sample: by judging the path of the accurate decision tree, the specific business item of each real reporting sample can be determined, and the real reporting sample belonging to the specific business item is extracted from a plurality of real reporting sample sets. Thus, complicated steps and checking work of tax payers in filling forms can be reduced, and the accuracy and efficiency of filling are improved.

Optimizing a calculation model: by extracting samples from the accurate decision tree, a computational model of each business event sample can be more accurately determined. The tax payer can operate according to a predetermined calculation model when calculating, the trouble of complex calculation formulas among and in the tables is reduced, and the accuracy and convenience of operation are improved.

Further, in one or more embodiments of the present disclosure, the learning of each service item sample according to the calculation manner of each tax type acquired in advance, and determining the calculation model of each service item sample respectively, so as to implement multiplexing of service items of a real reporting sample by applying the commonality feature in the learned calculation model to a specific service item of an actual sample, includes:

extracting each business item sample from the plurality of real reporting samples;

according to the pre-acquired tax calculation modes, learning the business item samples, and extracting the common characteristics of the samples;

and determining a calculation model of each business item sample according to the common characteristics of the samples and the tax calculation modes so as to apply the common characteristics in the calculation model of each business item sample to specific business items of an actual sample and realize multiplexing of the business items of the actual reporting sample.

It should be noted that the following specific beneficial effects are provided with respect to the above matters:

extracting common characteristics of samples: and according to the pre-acquired tax calculation modes, learning the business item samples, and extracting the common characteristics among the samples. These commonalities may be used to construct a computational model to enable multiplexing of sample business matters.

Constructing a calculation model: according to the common characteristics of the samples and the calculation mode of each tax, a calculation model of each business item sample can be determined. Thus, when the tax payer declares, the tax payer can operate based on a predetermined calculation model, the complex calculation and checking workload when filling the form is reduced, and the declaring accuracy and efficiency are improved.

In summary, by learning each business item sample according to the calculation mode of each tax acquired in advance and determining the calculation model, the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics, and the common characteristics of the sample can be extracted and the calculation model can be constructed, so that the complexity of filling in the form is reduced and the reporting accuracy and efficiency are improved.

Further, according to one or more embodiments of the present disclosure, according to the business item index model and a pre-generated tax calculation rule, calculating a currently declared business item, and outputting a currently declared sample, including:

comparing the current reporting sample with the plurality of real reporting samples to determine similar business matters of the current reporting sample;

and calculating the similar business items according to the business item index model and a pre-generated tax payment calculation rule, and outputting a current declaration sample.

It should be noted that the following specific beneficial effects are provided with respect to the above matters:

determining similar business matters: comparing the current reporting sample with a plurality of real reporting samples can determine which real reporting samples have similar business matters. Thus, the current declaration sample can be calculated more accurately according to the similar business matters.

The calculation accuracy is improved: according to the service item index model and the tax calculation rule which is generated in advance, similar service items are calculated, and the accuracy of calculation can be improved. By using pre-determined calculation rules, the trouble of complex inter-and intra-table calculation formulas is reduced, and the possibility of filling in errors and omissions is reduced.

The calculation efficiency is improved: the similar business items are calculated according to the business item index model and the tax calculation rule which is generated in advance, so that the calculation efficiency can be improved. The time and effort of complex calculation and check when filling out the form by the tax payer are reduced, and the reporting efficiency is improved.

In summary, according to the business item index model and the pre-generated tax payment calculation rule, the currently declared business items are calculated, the current declared samples are output to determine similar business items, and the calculation accuracy and calculation efficiency are improved, so that the problems of complex and complex filling of the form, trouble of calculation formulas among the form and in the form and the like are solved.

Further, after outputting the current declaration sample according to one or more embodiments of the present disclosure, the method further includes:

analyzing a difference parameter according to the comparison result of the current reporting sample and the plurality of real reporting samples;

and optimizing and adjusting the tax payment calculation rule according to the difference parameter.

It should be noted that the following specific beneficial effects are provided with respect to the above matters:

refined tax payment calculation rules: by comparing and analyzing with a plurality of real reporting samples, the difference parameters between the current reporting sample and the real reporting sample can be determined. These difference parameters reflect the specific differences in the calculation process between the current reporting sample and the actual reporting sample. Through analysis of the comparison result, the difference can be deeply known, so that more refined optimization adjustment of tax payment calculation rules is performed.

The calculation accuracy and adaptability are improved: and optimizing and adjusting tax payment calculation rules according to the difference parameters, and performing customized calculation according to the specific conditions of the current reporting samples. By optimizing the adjusted tax calculation rule, various tax fees related to the current declaration sample can be calculated more accurately, and the accuracy and adaptability of calculation are improved.

Reduce filling errors and omissions: the tax payment calculation rule is optimized and adjusted, so that the tax payment calculation rule is more in line with the characteristics and regulations of the current declaration sample, and filling errors and omission are reduced. The calculation rules with strong adaptability can reduce the trouble of tax payers on calculation formulas among complex tables and in tables, reduce the risks of errors and omission when filling the tables, and lighten the burden of the tax payers.

One or more embodiments of the present specification provide a declaration sample output device including:

at least one processor; the method comprises the steps of,

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to:

extracting each business item sample from a plurality of real reporting samples;

according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample;

Constructing a business item index model according to the calculation model of each business item sample;

and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.

One or more embodiments of the present description provide a non-volatile computer storage medium storing computer-executable instructions that, when executed by a computer, enable:

extracting each business item sample from a plurality of real reporting samples;

according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample;

constructing a business item index model according to the calculation model of each business item sample;

and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.

The above-mentioned at least one technical scheme that this description embodiment adopted can reach following beneficial effect:

Simplifying the filling process: by extracting each service item sample from a plurality of real reporting samples and learning a calculation model of each service item sample, the tax payer can multiplex the service items of the real reporting samples by applying the common characteristics in the learned calculation model to the specific service items of the actual samples. Thus, the tax payer can complete calculation and check more simply and rapidly when filling out the declaration form, and the complexity and complexity of filling out are reduced.

Accuracy is improved: by learning the calculation model of each business item sample and constructing the business item index model, the declaration data can be calculated more accurately. For tax payers without expert knowledge, the complexity of calculation formulas among tables and in tables can cause errors and omission, and the adoption of the method can reduce the occurrence of the problems and improve the accuracy of tax declaration.

Increasing the flexibility of the computation: the calculation modes of various tax types are obtained in advance and are applied to calculation models of different business item samples, so that the calculation modes can be flexibly adapted to different tax types and different tax payers. Therefore, tax payers do not need to independently process the adjustment and application problems of calculation formulas among tables and in tables, dependency on professional knowledge is reduced, and the flexibility of calculation is improved.

In summary, the declaration sample output method can simplify the filling process, improve the accuracy and increase the flexibility of calculation.

Drawings

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

FIG. 1 is a flow diagram of a method for reporting sample output according to one or more embodiments of the present disclosure;

FIG. 2 is a schematic diagram of an intelligent tax calculation engine based on an artificial neural network according to one or more embodiments of the present disclosure;

fig. 3 is a schematic structural diagram of a declaration sample output device according to one or more embodiments of the present disclosure.

Detailed Description

The embodiment of the specification provides a declaration sample output method, a declaration sample output device and a declaration sample output medium.

In order to make the technical solutions in the present specification better understood by those skilled in the art, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only some embodiments of the present specification, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present disclosure.

Fig. 1 is a schematic flow diagram of a method for outputting a declaration sample according to one or more embodiments of the present disclosure, where the flow may be performed by a declaration sample output system. Some input parameters or intermediate results in the flow allow for manual intervention adjustments to help improve accuracy.

The method flow steps of the embodiment of the present specification are as follows:

s102, extracting each business item sample from a plurality of real declaration samples.

In the embodiment of the present disclosure, entropy values and information gains may be calculated for the features in the multiple real reporting samples, and an optimal feature combination may be matched; dividing the plurality of real declaration samples according to the optimal feature combination to form an accurate decision tree; and finally, extracting the business item samples according to the accurate decision tree.

For a plurality of real declaration samples, the entropy value and the information gain may be calculated for the features in each sample. The entropy value may be calculated by counting the occurrence frequency of each characteristic value and then calculating the entropy value using an entropy formula. The information gain can be calculated by calculating the conditional entropy of each feature on the target variable and subtracting the entropy value of the target variable. And matching the optimal feature combination according to the calculation result. The criterion for selecting the feature combination may be a feature combination with the largest information gain or the smallest entropy value.

And then dividing a plurality of real declaration samples by using the obtained optimal feature combination to form an accurate decision tree. The decision tree can be constructed by recursively partitioning, starting from the root node, classifying the samples according to the optimal feature combination, and then repeating the process for each child node until all samples are correctly classified or a stop condition is reached.

And finally, extracting each business item sample according to the accurate decision tree. And predicting a new declaration sample according to the structure and characteristic conditions of the decision tree, and extracting a corresponding business item sample according to a prediction result.

It should be noted that, the above method for extracting each business item sample from multiple real declaration samples has the following specific beneficial effects:

the accuracy of sample extraction is improved: by carrying out entropy calculation and information gain analysis on the characteristics in a plurality of real declaration samples, the optimal characteristic combination can be matched, so that the accuracy of sample extraction is improved. Therefore, the extracted business item sample can reflect the characteristics and rules of the real declaration data more truly, and a more accurate data basis is provided for the subsequent calculation model construction.

Further, in one or more embodiments of the present disclosure, entropy values and information gains are calculated for features in the plurality of real reporting samples, respectively, and when an optimal feature combination is matched, entropy values of each feature may be calculated for the features in the plurality of real reporting samples, respectively, where the entropy values are used to measure a degree of confusion or uncertainty of the feature in the sample; calculating information gain of each feature according to entropy values of the features for the features in the plurality of real declaration samples, wherein the information gain measures the degree of uncertainty reduction of the result under given feature conditions; and selecting the feature with the maximum information gain as the optimal feature combination according to the information gain of each feature.

It should be noted that, regarding the matching of the optimal feature combination, the following specific steps may be adopted:

for a plurality of real declaration samples, an entropy value is calculated for each feature. Entropy values may be calculated by counting the occurrence frequency of each characteristic value and then using an entropy formula. Entropy is used to measure the degree of confusion or uncertainty in a feature in a sample. For a plurality of real declaration samples, the information gain of each feature can be calculated by the entropy value of each feature. The information gain can be calculated by calculating the conditional entropy of each feature on the target variable and subtracting the entropy value of the target variable. The information gain measures the degree to which the uncertainty of the result is reduced for a given characteristic. And selecting the feature with the maximum information gain as the optimal feature combination according to the information gain of each feature. The feature of the maximum information gain means that under this feature condition, the classification of the target variable is most affected, i.e., the uncertainty of the result can be reduced most.

It should be noted that, the method for calculating entropy and information gain for the features in the multiple real declaration samples and matching the optimal feature combination has the following specific beneficial effects:

determining an optimal feature combination: by calculating entropy values and information gains of features in a plurality of real declaration samples, the degree of confusion of the features in the samples and the degree of influence of the features on the uncertainty of results can be evaluated. By selecting the feature with the greatest information gain, an optimal combination of features, i.e. the feature with the greatest impact on the result, can be determined.

The accuracy of feature selection is improved: by calculating the entropy value and the information gain, the influence of each feature on the result can be objectively evaluated. The feature with the maximum information gain is selected as the optimal feature combination, so that the feature with the greatest influence on the result can be more accurately selected, and an accurate basis is provided for subsequent sample segmentation and decision tree construction.

Optimizing sample segmentation and decision tree construction: by selecting the optimal feature combination, a plurality of real declaration samples can be segmented to form an accurate decision tree. The accurate decision tree can better reflect the characteristics and rules of the real declaration data, and provides a more accurate and reliable basis for the subsequent calculation model construction and business item sample extraction.

In summary, the method for calculating the entropy and the information gain of the features in the plurality of real declaration samples and matching the optimal feature combination can determine the optimal feature combination, improve the accuracy of feature selection and optimize sample segmentation and decision tree construction.

Further, in one or more embodiments of the present disclosure, when dividing the plurality of real reporting samples according to the optimal feature combination to form an accurate decision tree, the plurality of real reporting samples may be first divided according to the optimal feature combination to form an initial decision tree; circularly dividing a plurality of real declaration samples according to the optimal feature combination, and judging whether to continue dividing according to predefined conditions, wherein the predefined conditions comprise predefined maximum depth or predefined sample number; and if the predefined condition is met, forming the accurate decision tree.

It should be noted that, regarding the forming of the accurate decision tree, the following specific steps may be adopted:

and according to the optimal characteristic combination, firstly, dividing a plurality of real reporting samples for the first time to form an initial decision tree. The samples can be divided according to the optimal feature combination, and the samples with the same feature value are classified into the same sub-node.

And then carrying out cyclic segmentation according to the optimal feature combination, and judging whether to continue segmentation according to predefined conditions. The predefined condition may be a maximum depth or a number of samples. If the predefined condition is met, continuing to divide, otherwise stopping dividing.

In the cyclic segmentation process, each child node may be further segmented using a recursive algorithm. In each division, selecting the feature with the maximum information gain according to the optimal feature combination as a division basis, and dividing the sample into smaller sub-nodes.

And stopping the loop segmentation when the predefined condition is reached, and forming a final accurate decision tree. Each leaf node of the accurate decision tree represents a sample of business matters, and the samples on the leaf nodes are identified as belonging to the corresponding business matters.

It should be noted that, the forming of the accurate decision tree has the following specific beneficial effects:

the accuracy of the decision tree is improved: by using the optimal feature combination to divide a plurality of real declaration samples, the nodes and branches of the decision tree can be ensured to accurately capture the features and rules of the real declaration data. In this way, the formed accurate decision tree can more accurately predict the business item data of the current declaration sample, and reduce the possibility of filling errors and omission.

Simplifying the decision tree construction process: by specifying predefined conditions, such as a predefined maximum depth or number of samples, the complexity and size of the decision tree can be controlled. In this way, the construction process of the decision tree can be simplified, unnecessary segmentation and nodes are reduced, and the interpretability and usability of the decision tree are improved.

The segmentation efficiency is improved: the efficiency of segmentation can be improved by performing the cyclic segmentation using the optimal feature combination and judging whether to continue the segmentation according to the predefined condition. Only when the predefined condition is met, the next round of segmentation is performed, unnecessary calculation and processing are avoided, and time and resources are saved.

In summary, the plurality of real declaration samples are segmented according to the optimal feature combination to form an accurate decision tree, so that the accuracy of the decision tree can be improved, the decision tree construction process is simplified, and the segmentation efficiency is improved.

Further, in one or more embodiments of the present disclosure, when extracting the service item samples according to the accurate decision tree, a plurality of real reporting samples may be input into the accurate decision tree, so as to determine, according to a path of each real reporting sample on the accurate decision tree, a service item to which each real reporting sample belongs; and extracting each real reporting sample according to the business item to which each real reporting sample belongs so as to extract the real reporting sample belonging to the specific business item from a plurality of real reporting sample sets to form each business item sample.

It should be noted that, regarding the above-mentioned extraction of each business item sample, the following specific implementation steps may be adopted:

a plurality of true declaration samples are input into an accurate decision tree. For each real declaration sample, the root node can be traversed downwards step by step according to the judgment conditions of the accurate decision tree, and the service item to which each sample belongs is determined according to the path of each sample on the decision tree. And extracting each real reporting sample from a plurality of real reporting sample sets according to the business items to which the real reporting sample belongs. The true declaration samples belonging to the specific business item are classified into the corresponding business item sample set.

When forming each business item sample, each real reporting sample can be extracted from a plurality of real reporting sample sets according to the business item to which each real reporting sample belongs to form a sample set of each business item.

It should be noted that the following specific beneficial effects are provided with respect to the above matters:

automatic classification: the classification of a plurality of real declaration samples can automatically identify different business matters without manual intervention, so that the risk of manual errors is reduced, and the classification accuracy and efficiency are improved.

Accurately extracting a sample: by judging the path of the accurate decision tree, the specific business item of each real reporting sample can be determined, and the real reporting sample belonging to the specific business item is extracted from a plurality of real reporting sample sets. Thus, complicated steps and checking work of tax payers in filling forms can be reduced, and the accuracy and efficiency of filling are improved.

Optimizing a calculation model: by extracting samples from the accurate decision tree, a computational model of each business event sample can be more accurately determined. The tax payer can operate according to a predetermined calculation model when calculating, the trouble of complex calculation formulas among and in the tables is reduced, and the accuracy and convenience of operation are improved.

S104, learning each business item sample according to a calculation mode of each tax type acquired in advance, and respectively determining a calculation model of each business item sample so as to realize multiplexing of the business items of the real reporting sample by applying the common characteristics in the learned calculation model to the specific business items of the actual sample.

In this embodiment of the present disclosure, each service item sample may be first extracted from the plurality of real reporting samples, and the plurality of real reporting samples may be classified according to the existing service item classification standard, so as to extract a sample set pertaining to each service item; the tax calculation modes are acquired in advance, the tax samples are learned, the common characteristics of the samples are extracted, the common characteristics related to the tax calculation modes can be acquired by analyzing the characteristics of the business event samples, and the common characteristics can be shared by a plurality of business events, such as income, cost, profit and the like; and finally, determining a calculation model of each business item sample according to the common characteristics of the samples and the tax calculation modes so as to apply the common characteristics in the calculation model of each business item sample to specific business items of an actual sample to realize multiplexing of the business items of the actual reporting sample, and determining the calculation model of each business item sample according to the common characteristics of the samples and the tax calculation modes. And according to the extracted commonality characteristics, combining with each tax calculation mode, establishing a calculation model aiming at each business item. The computational model should be able to apply commonality characteristics to specific business matters of the actual samples to enable multiplexing of business matters of the actual reporting samples.

It should be noted that the following specific beneficial effects are provided with respect to the above matters:

extracting common characteristics of samples: and according to the pre-acquired tax calculation modes, learning the business item samples, and extracting the common characteristics among the samples. These commonalities may be used to construct a computational model to enable multiplexing of sample business matters.

Constructing a calculation model: according to the common characteristics of the samples and the calculation mode of each tax, a calculation model of each business item sample can be determined. Thus, when the tax payer declares, the tax payer can operate based on a predetermined calculation model, the complex calculation and checking workload when filling the form is reduced, and the declaring accuracy and efficiency are improved.

In summary, by learning each business item sample according to the calculation mode of each tax acquired in advance and determining the calculation model, the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics, and the common characteristics of the sample can be extracted and the calculation model can be constructed, so that the complexity of filling in the form is reduced and the reporting accuracy and efficiency are improved.

S106, constructing a business item index model according to the calculation model of each business item sample.

In the embodiment of the present specification, regarding the above construction of the business item index model, the following specific steps may be adopted:

and determining to construct the business item indexes by using the neural network model according to the calculation model of each business item sample. Neural network models can be used to model complex nonlinear relationships and can extract features and make predictions by learning large amounts of sample data. And collecting and arranging a training data set related to the construction of the business item indexes. The training data set should include input variables and corresponding target variables (business event indicators). And designing the structure of the neural network model. The number of layers of the neural network, the number of neurons of each layer, an activation function and the like are determined, and the number of neurons of each layer, the activation function and the like are correspondingly adjusted according to the characteristics of the service matters.

Training of the neural network is performed. The training data set is input into a neural network model, the model is trained through a back propagation algorithm, and the weight and bias of the model are optimized, so that the training data set can be better fitted.

And (5) verifying and optimizing the model. The validation dataset is used to evaluate the performance of the model and to adjust and optimize the model to improve its generalization ability and accuracy.

And predicting the business item index by using the trained and verified neural network model. New input data are input into the trained model, and corresponding business item indexes are obtained through forward propagation calculation. And for each business item, constructing a business item index model according to the index result obtained by prediction. The model may include the structure, parameter settings, and related background knowledge of the neural network model.

S108, calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.

In this embodiment of the present disclosure, the current reporting sample may be compared with the plurality of real reporting samples to determine similar service matters of the current reporting sample; and calculating the similar business items according to the business item index model and a pre-generated tax payment calculation rule, and outputting a current declaration sample.

It should be noted that, regarding the output current declaration sample, the following specific implementation steps may be adopted:

the collection of information related to the plurality of actual reporting samples and the current reporting samples may include business matters and related data. The current declaration sample is compared with a plurality of real declaration samples, and the similar business items of the current declaration sample are determined through comparison, so that a data matching algorithm or a machine learning algorithm can be used for comparison and similarity calculation.

And calculating similar business matters according to the business matters index model and a pre-generated tax calculation rule. The business transaction index model may be a classification model for classifying business transactions into different categories. The tax payment calculation rule may be a preset of tax payment modes and calculation formulas for each category.

Outputting the calculation result of the current declaration sample: and generating a corresponding tax return or report according to the calculation result, and providing reference and submission for a declaration person.

Verifying and auditing the calculation result: by comparing with other real declaration samples, the accuracy and rationality of the calculation result are confirmed, and further adjustment and correction can be performed if necessary.

Compiling and submitting declaration materials: and generating corresponding reporting materials including tax return, attached sheets, related documents and the like according to the calculation result, and submitting the materials to tax authorities according to related regulations and programs.

Monitoring and tracking declaration results: and monitoring the reporting material, and timely knowing the reporting result and the related feedback information. If necessary, the system can be adjusted and corrected in time to ensure the accuracy and compliance of reporting.

Periodic data analysis and model updates are performed: and according to the actual reporting condition and the feedback information, the business item index model and the tax payment calculation rule are updated and optimized regularly, so that the calculation accuracy and the practicability are improved.

It should be noted that the following specific beneficial effects are provided with respect to the above matters:

determining similar business matters: comparing the current reporting sample with a plurality of real reporting samples can determine which real reporting samples have similar business matters. Thus, the current declaration sample can be calculated more accurately according to the similar business matters.

The calculation accuracy is improved: according to the service item index model and the tax calculation rule which is generated in advance, similar service items are calculated, and the accuracy of calculation can be improved. By using pre-determined calculation rules, the trouble of complex inter-and intra-table calculation formulas is reduced, and the possibility of filling in errors and omissions is reduced.

The calculation efficiency is improved: the similar business items are calculated according to the business item index model and the tax calculation rule which is generated in advance, so that the calculation efficiency can be improved. The time and effort of complex calculation and check when filling out the form by the tax payer are reduced, and the reporting efficiency is improved.

In summary, according to the business item index model and the pre-generated tax payment calculation rule, the currently declared business items are calculated, the current declared samples are output to determine similar business items, and the calculation accuracy and calculation efficiency are improved, so that the problems of complex and complex filling of the form, trouble of calculation formulas among the form and in the form and the like are solved.

Further, after outputting the current reporting sample in one or more embodiments of the present disclosure, the difference parameter may be analyzed according to a comparison result between the current reporting sample and the plurality of real reporting samples; and then optimizing and adjusting the tax payment calculation rule according to the difference parameter.

It should be noted that the following specific beneficial effects are provided with respect to the above matters:

refined tax payment calculation rules: by comparing and analyzing with a plurality of real reporting samples, the difference parameters between the current reporting sample and the real reporting sample can be determined. These difference parameters reflect the specific differences in the calculation process between the current reporting sample and the actual reporting sample. Through analysis of the comparison result, the difference can be deeply known, so that more refined optimization adjustment of tax payment calculation rules is performed.

The calculation accuracy and adaptability are improved: and optimizing and adjusting tax payment calculation rules according to the difference parameters, and performing customized calculation according to the specific conditions of the current reporting samples. By optimizing the adjusted tax calculation rule, various tax fees related to the current declaration sample can be calculated more accurately, and the accuracy and adaptability of calculation are improved.

Reduce filling errors and omissions: the tax payment calculation rule is optimized and adjusted, so that the tax payment calculation rule is more in line with the characteristics and regulations of the current declaration sample, and filling errors and omission are reduced. The calculation rules with strong adaptability can reduce the trouble of tax payers on calculation formulas among complex tables and in tables, reduce the risks of errors and omission when filling the tables, and lighten the burden of the tax payers.

The above-mentioned matters of the embodiments of the present disclosure have the following beneficial effects:

simplifying the filling process: by extracting each service item sample from a plurality of real reporting samples and learning a calculation model of each service item sample, the tax payer can multiplex the service items of the real reporting samples by applying the common characteristics in the learned calculation model to the specific service items of the actual samples. Thus, the tax payer can complete calculation and check more simply and rapidly when filling out the declaration form, and the complexity and complexity of filling out are reduced.

Accuracy is improved: by learning the calculation model of each business item sample and constructing the business item index model, the declaration data can be calculated more accurately. For tax payers without expert knowledge, the complexity of calculation formulas among tables and in tables can cause errors and omission, and the adoption of the method can reduce the occurrence of the problems and improve the accuracy of tax declaration.

Increasing the flexibility of the computation: the calculation modes of various tax types are obtained in advance and are applied to calculation models of different business item samples, so that the calculation modes can be flexibly adapted to different tax types and different tax payers. Therefore, tax payers do not need to independently process the adjustment and application problems of calculation formulas among tables and in tables, dependency on professional knowledge is reduced, and the flexibility of calculation is improved.

In summary, the declaration sample output method can simplify the filling process, improve the accuracy and increase the flexibility of calculation.

It should be noted that, fig. 2 is a schematic diagram of an intelligent tax calculation engine based on an artificial neural network according to an embodiment of the present disclosure, and includes 7 steps: initializing a calculation engine, extracting a data set sample, extracting a business item model, constructing an index model, executing calculation by a rule engine, adjusting a sample rule, and forming a test set.

(1) Initializing a compute engine

And initializing a computing engine according to the technical characteristics of tax types, period and the like. And loading different tax types calculation modules.

(2) Extracting data set samples

And calculating entropy and information gain according to the features x, y, z...n in the initial parameters in the true declaration samples, so as to match the optimal feature combination as a splitting standard and continuously split the sample set. And circularly executing segmentation before the stopping condition is met to form an accurate decision tree. Thereby matching out a corresponding plurality of samples.

For example, in a sample set, a plurality of characteristics such as organization, period, tax type, reporting mode and the like are arranged in the experience reporting data, and then the tax type characteristic is increased to reduce the disorder degree of the sample, so that the reporting data with high matching degree is found from the experience reporting data through the characteristic combination of period and tax type.

(3) Extracting business item model

Extracting each business item in the experience declaration data matched in the step (2), and carrying out multi-sample analysis aiming at the business item attribute to gradually determine a calculation model of each business item. The common features learned into the model are applied in a computation mode that matches the business transaction of the actual sample to multiplex the data items of the historical sample.

For example: the tax declaration forms of value added tax are filled in, the tax payer unifies social credit codes in each period, the declaration forms of each tax type are consistent, the calculation is not needed again, and the numerical value of the historical declaration period can be automatically extracted according to different organizations. And for the incoming invoice data, the calculation method of the historical reporting forms is combined after various conditions such as organization, period, tax types and the like are combined, and automatic calculation is carried out on the incoming actual incoming invoice data in the current period. However, the attribution classification of different incoming invoices in the current declaration data can be determined according to the calculation rules of the historical declaration data. And a more accurate classification calculation rule is formed for more accurate inbound invoice classification calculation of the later invoices.

(4) Construction of an index model

And (3) learning the calculation model of each business item, supplementing, correcting and optimizing the index model library model, and improving the generalization capability of the model.

(5) Rules engine performs computation

And carrying out the current declaration calculation according to the established business item index model through tax payment calculation rules. And outputting a sample of the corresponding declaration data.

(6) Adjusting sample rules

Aiming at sample data of a training set and a verification set, the true reporting sample constructed at the time is combined, and the index level is improved through reporting data of the directivity and the hit rate of the reporting sample, so that the calculation complexity is reduced. And automatically optimizing and adjusting the calculation rule of tax declaration.

(7) Forming a test set

And the calculation engine calculates a test set back feeding tax payment index calculation sample formed after the calculation is completed. And a richer and more accurate tax payment index data set is formed.

Fig. 3 is a schematic structural diagram of a declaration sample output device according to one or more embodiments of the present disclosure, including:

at least one processor; the method comprises the steps of,

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to:

Extracting each business item sample from a plurality of real reporting samples;

according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample;

constructing a business item index model according to the calculation model of each business item sample;

and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.

One or more embodiments of the present description provide a non-volatile computer storage medium storing computer-executable instructions that, when executed by a computer, enable:

extracting each business item sample from a plurality of real reporting samples;

according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample;

Constructing a business item index model according to the calculation model of each business item sample;

and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.

In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for apparatus, non-volatile computer storage medium embodiments, the description is relatively simple, as it is substantially similar to method embodiments, with reference to the section of the method embodiments being relevant.

The foregoing describes specific embodiments of the present disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

The foregoing is merely one or more embodiments of the present description and is not intended to limit the present description. Various modifications and alterations to one or more embodiments of this description will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement, or the like, which is within the spirit and principles of one or more embodiments of the present description, is intended to be included within the scope of the claims of the present description.

Claims (10)

1. A method of reporting sample output, the method comprising:

extracting each business item sample from a plurality of real reporting samples;

according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample;

constructing a business item index model according to the calculation model of each business item sample;

and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.

2. The method of claim 1, wherein extracting business event samples from a plurality of true declaration samples comprises:

respectively calculating entropy values and information gains for the features in the plurality of real reporting samples, and matching an optimal feature combination;

dividing the plurality of real declaration samples according to the optimal feature combination to form an accurate decision tree;

and extracting the business item samples according to the accurate decision tree.

3. The method of claim 2, wherein the computing entropy and information gain for the features in the plurality of real declaration samples, respectively, and matching the optimal feature combinations comprises:

for the features in the plurality of real declaration samples, respectively calculating entropy values of the features, wherein the entropy values are used for measuring the confusion degree or uncertainty of the features in the samples;

calculating information gain of each feature according to entropy values of the features for the features in the plurality of real declaration samples, wherein the information gain measures the degree of uncertainty reduction of the result under given feature conditions;

and selecting the feature with the maximum information gain as the optimal feature combination according to the information gain of each feature.

4. The method of claim 2, wherein the partitioning the plurality of true declaration samples according to the optimal feature combinations to form an accurate decision tree comprises:

performing first segmentation on a plurality of real reporting samples according to the optimal feature combination to form an initial decision tree;

circularly segmenting a plurality of real declaration samples according to the optimal feature combination, and judging whether to continue segmentation according to predefined conditions, wherein the predefined conditions comprise a predefined maximum depth or a predefined sample number;

and if the predefined condition is met, forming the accurate decision tree.

5. The method of claim 2, wherein extracting the business transaction samples from the accurate decision tree comprises:

inputting a plurality of real reporting samples into an accurate decision tree to determine service matters to which each real reporting sample belongs according to the path of each real reporting sample on the accurate decision tree;

and extracting each real reporting sample according to the business item to which each real reporting sample belongs so as to extract the real reporting sample belonging to the specific business item from a plurality of real reporting sample sets to form each business item sample.

6. The method according to claim 1, wherein the learning the service item samples according to the calculation mode of each tax type acquired in advance, and determining the calculation model of each service item sample respectively, so as to implement multiplexing of service items of a real reporting sample by applying the commonality feature in the learned calculation model to a specific service item of an actual sample, includes:

extracting each business item sample from the plurality of real reporting samples;

according to the pre-acquired tax calculation modes, learning the business item samples, and extracting the common characteristics of the samples;

and determining a calculation model of each business item sample according to the common characteristics of the samples and the tax calculation modes so as to apply the common characteristics in the calculation model of each business item sample to specific business items of an actual sample and realize multiplexing of the business items of the actual reporting sample.

7. The method of claim 1, wherein the calculating the currently declared business item according to the business item indicator model and a pre-generated tax calculation rule, and outputting a currently declared sample, comprises:

Comparing the current reporting sample with the plurality of real reporting samples to determine similar business matters of the current reporting sample;

and calculating the similar business items according to the business item index model and a pre-generated tax payment calculation rule, and outputting a current declaration sample.

8. The method of claim 1, wherein after outputting the current declaration sample, the method further comprises:

analyzing a difference parameter according to the comparison result of the current reporting sample and the plurality of real reporting samples;

and optimizing and adjusting the tax payment calculation rule according to the difference parameter.

9. A declaration sample output device, comprising:

at least one processor; the method comprises the steps of,

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to:

extracting each business item sample from a plurality of real reporting samples;

according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample;

Constructing a business item index model according to the calculation model of each business item sample;

and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.

10. A non-transitory computer storage medium storing computer executable instructions that when executed by a computer enable:

extracting each business item sample from a plurality of real reporting samples;

according to the calculation mode of each tax acquired in advance, each business item sample is learned, and the calculation model of each business item sample is respectively determined, so that the multiplexing of the business items of the real reporting sample is realized by applying the common characteristics in the learned calculation model to the specific business items of the actual sample;

constructing a business item index model according to the calculation model of each business item sample;

and calculating the currently declared business item according to the business item index model and a pre-generated tax payment calculation rule, and outputting a currently declared sample.