WO2020140639A1

WO2020140639A1 - Machine learning-based report generating method, apparatus, and computer device

Info

Publication number: WO2020140639A1
Application number: PCT/CN2019/119480
Authority: WO
Inventors: 徐凯
Original assignee: 平安科技（深圳）有限公司
Priority date: 2019-01-02
Filing date: 2019-11-19
Publication date: 2020-07-09
Also published as: CN109800333A

Abstract

A machine learning-based report generating method and an apparatus, a computer device, and a storage medium, the method comprising: acquiring feature information of a current user; inputting the feature information into a preset machine learning-based report category prediction model and performing calculation on same, the report category prediction model having been trained on training data constituted by user feature information and report categories corresponding to the user feature information; outputting a predicted report category the current user will use; on the basis of the report category that will be used, retrieving a preset initial report from a database; and on the basis of information input by the current user, adjusting the initial report to obtain a final report. The report that the user requires is thereby correctly generated, reducing time and effort spent by the user in adjusting the report.

Description

Report generation method, device and computer equipment based on machine learning

This application requires the priority of the Chinese patent application submitted to the China Patent Office on January 2, 2019, with the application number 2019100029515 and the invention titled "Machine learning-based report generation method, device and computer equipment", the entire content of which is cited by reference Incorporated in this application.

Technical field

This application relates to the field of computers, and in particular to a method, device, computer equipment, and storage medium for generating reports based on machine learning.

Background technique

In different industries, different types of reports are often required. Even in the same industry, different types of reports may be required. The existing technology cannot effectively predict what kind of report the user specifically needs. Therefore, it is often necessary for the user to manually adjust the report type and report parameters, thereby requiring tedious manual operations, reducing the user's work efficiency, and the user experience must be extremely poor. In addition, users in the prior art need to manually adjust the report to obtain the final desired report, that is, there is no report template, or no suitable report template, and then when a new report is needed, manual adjustment is required. Adjust the report by a wide margin. Therefore, the report generation solution of the prior art cannot prevent users from spending extra time and effort to adjust the report.

technical problem

The main purpose of the present application is to provide a report generation method, device, computer equipment and storage medium based on machine learning, aiming to use templates to accurately generate the reports required by users and reduce the time and effort for users to adjust the reports.

Technical solution

In order to achieve the above purpose, this application proposes a method for generating a report based on machine learning, including the following steps:

Acquiring characteristic information of the current user, the characteristic information of the current user includes at least the occupation information of the current user;

The feature information is input into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model is trained by training data composed of user feature information and a report type corresponding to the user feature information Made

Output the predicted report type that the current user will use;

According to the type of report to be used, a preset preliminary report is retrieved from the database, wherein the type of the preliminary report is the same as the type of report to be used;

According to the information input by the current user, the preliminary report is adjusted to obtain a final report.

Further, the step of retrieving a preset preliminary report from the database according to the type of report to be used, wherein the step of the type of the preliminary report being the same as the type of report to be used includes:

According to the type of report to be used, retrieve preset multiple chart templates and multiple text part templates from a preset database;

Combining the chart template and the text part template selected by the current user into the preliminary report;

Recall the preliminary report.

Further, the step of adjusting the preliminary report according to the information input by the current user to obtain the final report includes:

Adjust the graph and text parts in the preliminary report according to the graph adjustment information, graph data content information and text part adjustment information input by the current user;

The text content input by the current user is filled into the text portion of the preliminary report to obtain the final report.

Further, the method for obtaining the report type prediction model includes:

Obtaining a training set including a specified amount of sample data, where the sample data includes user feature information and a report type corresponding to the user feature information;

Input the sample data of the training set into the neural network model for training, in which the stochastic gradient descent method is used in the training process, and the parameters of each layer of the neural network model are updated by the reverse conduction rule to obtain the preliminary training model;

The preliminary training model is recorded as the report type prediction model.

Further, the step of recording the preliminary training model as the report type prediction model includes:

Obtaining a verification set including a specified amount of sample data, wherein the sample data of the verification set includes user feature information and a report type corresponding to the user feature information;

Verify the preliminary training model using the sample data of the verification set;

If the verification is passed, the preliminary training model is recorded as the report type prediction model.

Further, the method for obtaining the report type prediction model includes:

Obtain a specified amount of sample data, where the sample data includes user feature information and a report type corresponding to the user feature information;

Input the sample data of the training set into the CHAID decision tree model for training to obtain a preliminary CHAID decision tree;

The preliminary CHAID decision tree is recorded as the report type prediction model.

Further, the step of inputting the sample data of the training set into the CHAID decision tree model for training to obtain a preliminary CHAID decision tree includes:

Set the modeling standard parameters of the CHAID decision tree model, the modeling standard parameters include the maximum number of decision trees, the subdividable significance level of the parent node, the minimum number of samples contained in the parent node and the minimum samples contained in the child node number;

The sample data of the training set is input into the CHAID decision tree model established by the chi-square automatic interactive detection method for training to obtain a preliminary CHAID decision tree.

This application provides a report generation device based on machine learning, including:

A characteristic information obtaining unit, configured to obtain characteristic information of the current user, the characteristic information of the current user includes at least the occupation information of the current user;

Report type prediction model operation unit, used for inputting the feature information into a preset report type prediction model based on machine learning, wherein the report type prediction model is determined by user feature information and corresponding to the user feature information The training data composed of the report type is trained;

A report type prediction unit, used to output the predicted report type that the current user will use;

A preliminary report retrieval unit, configured to retrieve a preset preliminary report from the database according to the type of report to be used, wherein the type of the preliminary report is the same as the type of report to be used;

The final report obtaining unit is used to adjust the preliminary report according to the information input by the current user, so as to obtain the final report.

The present application provides a computer device, including a memory and a processor. The memory stores computer-readable instructions. When the processor executes the computer-readable instructions, any of the steps of the method described above is implemented.

The present application provides a computer-readable storage medium on which computer-readable instructions are stored, and when the computer-readable instructions are executed by a processor, implement the steps of any one of the above methods.

Beneficial effect

The machine learning-based report generation method, device, computer equipment and storage medium of the present application, through the machine learning-based report type prediction model, to predict the report type that the current user will use, and then adjust from the database according to the report type Take the preset preliminary report, adjust the preliminary report to get the final report, so as to avoid the user's tedious operation, improve the user experience, and improve the efficiency of report completion.

BRIEF DESCRIPTION

1 is a schematic flowchart of a method for generating a report based on machine learning according to an embodiment of the application;

2 is a schematic block diagram of a structure of a report generation device based on machine learning according to an embodiment of the present application;

FIG. 3 is a schematic block diagram of a computer device according to an embodiment of the present application.

The implementation, functional characteristics and advantages of the present application will be further described in conjunction with the embodiments and with reference to the drawings.

Best Mode of the Invention

In order to make the purpose, technical solutions and advantages of the present application more clear, the following describes the present application in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

Referring to FIG. 1, an embodiment of the present application provides a report generation method based on machine learning, including the following steps:

S1. Obtain the current user's characteristic information, where the current user's characteristic information includes at least the current user's occupation information;

S2. The feature information is input into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model is trained by user feature information and a report type corresponding to the user feature information. Data training;

S3. Output the predicted report type that the current user will use;

S4. According to the type of report to be used, retrieve a preset preliminary report from the database, wherein the type of the preliminary report is the same as the type of report to be used;

S5. Adjust the preliminary report according to the information input by the current user, so as to obtain a final report.

As described in step S1 above, the current user's characteristic information is obtained, and the current user's characteristic information includes at least the current user's occupation information. Wherein, the characteristic information of the current user refers to information that can reflect the characteristics of the current user, such as the occupation of the current user, the type of report used by the current user (within a specified time), the age of the current user, and the gender of the current user. Among them, the occupation information of the current user has a greater relationship with the type of report that may be used, so the characteristic information of the current user includes at least the occupation information of the current user, so as to improve the accuracy of the prediction of the report type. For example, stock commentators in the financial industry have a higher probability of using stock statements. The process of acquiring the characteristic information of the current user includes: extracting the characteristic information of the current user from the registered account information of the current user.

As described in step S2 above, the feature information is input into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model passes user feature information and a report type corresponding to the user feature information Trained. The report type prediction model based on machine learning, through continuous self-learning, improves the accuracy of prediction and avoids predicting the wrong report type. The report type prediction model can be generated based on any machine learning model, such as a neural network model and a classification tree model, and then trained through training data. Wherein, the report type prediction model is trained by training data composed of user characteristic information and the report type corresponding to the user characteristic information. The user characteristic information may include any information that can reflect the characteristics of the user, such as the occupation of the user, the type of report recently used by the user (within a specified time), the age of the user, and the gender of the user. Among them, the user refers to a user who uses a report. Wherein, the report type prediction model is used to predict the report type to be used by the user based on the user characteristic information.

As described in step S3 above, the predicted report type to be used by the current user is output. The report type can be classified into any classification method. For example, it can be divided into the following chart types: pie chart report (including pie chart in the report), curve chart report, and histogram report; : Financial statements, financial statements, statistical analysis reports. Through the report type prediction model, the predicted report type to be used by the current user can be output.

As described in step S4 above, according to the type of report to be used, a preset preliminary report is retrieved from the database, wherein the type of the preliminary report is the same as the type of report to be used. There are pre-stored preliminary reports of different report types in the preset database, and there can be multiple preliminary reports of the same report type for users to choose. For example, the preliminary report of the pie chart report can include three types of reports: the pie chart is at the top of the report, the middle of the report, and the end of the report; the financial preliminary report can include the report with the K chart, the report with the histogram, and the report with the curve chart. Reports, etc. These preliminary reports can be directly selected by the user, thereby eliminating the user's trivial process of creating reports from scratch, and improving the speed of report formation. Wherein, the preset preliminary report may be a completed complete report, or the current user may select the exact chart template and text part template from multiple alternative chart templates and text part templates. The combined report.

As described in step S5 above, according to the information input by the current user, the preliminary report is adjusted to obtain a final report. From the foregoing, a preliminary report is provided, and then the final report can be obtained based on the information input by the current user. Wherein, the information input by the current user includes at least one of information for adjusting the parameters of the preliminary report and text content information for describing the preliminary report. Wherein, the final report includes at least a chart part, and further, may include a text part.

In an embodiment, the preset preliminary report is retrieved from the database according to the type of report to be used, wherein the step S4 of the type of the preliminary report is the same as the type of report to be used, including:

S401. According to the type of report to be used, retrieve a plurality of preset chart templates and a plurality of text part templates from a preset database;

S402: Combine the graph template selected by the current user and the text part template into the preliminary report;

S403. Recall the preliminary report.

As mentioned above, the preset preliminary report is retrieved from the database. Wherein, the preliminary report is generated by a combination of chart templates (including chart style templates and chart data templates) and text part templates. Because different reports require different specific requirements, the format, layout, charts, etc. of the report are different. By decomposing the report into charts and text parts, the chart templates and text part templates are designed in advance and stored in the database. When a specific report is needed, you only need to choose from the existing chart templates and text part templates. Combined together, the preliminary report can be formed.

In one embodiment, the step S5 of adjusting the preliminary report according to the information input by the current user to obtain the final report includes:

S501. Adjust the chart and text in the preliminary report according to the chart adjustment information, chart data content information, and text part adjustment information input by the current user;

S502. Fill the text content input by the current user into the text portion of the preliminary report to obtain a final report.

As described above, the preliminary report is adjusted according to the information input by the current user, thereby obtaining the final report. As mentioned earlier, preliminary reports have been obtained. However, some detailed parameters and specific data content of the preliminary report are not supplemented. According to this, adjustments are made according to the specific instructions of the current user, wherein the chart adjustment information includes: adjusting the size of the chart, the data display parameters of the chart (such as the unit time length of the time axis, etc.), etc.; the chart data content information includes : Chart data (such as the data points of the graph, the proportion of each block of the pie chart, etc.); the adjustment information of the text part includes: adjusting the font size and color. Then, the text content input by the current user is filled in the text part to obtain a final report.

In one embodiment, the method for obtaining the report type prediction model includes:

S201. Obtain a training set including a specified amount of sample data, where the sample data includes user feature information and a report type corresponding to the user feature information;

S202. Input the sample data of the training set into a neural network model for training, in which a stochastic gradient descent method is used in the training process, and the parameters of each layer of the neural network model are updated using the reverse conduction law to obtain preliminary training. model;

S203. Record the preliminary training model as the report type prediction model.

As mentioned above, the acquisition of the report type prediction model is realized. The characteristic information of the user refers to information that can reflect the characteristics of the user, such as the occupation of the user, the type of report currently used recently (within a specified time), the age of the user, and the gender of the user. The machine learning in this embodiment uses a neural network model, such as the VGG-F model, VGG16 model, InceptionV3 model, Xception model, AlexNet model, etc., and then includes user feature information and a report corresponding to the user feature information The type of sample data trains the neural network model, wherein the more sample data, the more accurate the prediction model trained. When there are too many sample data, it is preferable to use the stochastic gradient descent method for training, that is, to randomly sample some training data to replace the entire training set, thereby increasing the training speed. The reverse conduction law (BP) is based on the gradient descent method, which is essentially a mapping relationship: the function of an n-input m-output BP neural network is from n-dimensional Euclidean space to m-dimensional Euclidean Continuous mapping of a finite field in the Hurst space to update the parameters.

In one embodiment, the step 203 of recording the preliminary training model as the report type prediction model includes:

S2031: Obtain a verification set including a specified amount of sample data, where the sample data of the verification set includes user feature information and a report type corresponding to the user feature information;

S2032. Use the sample data of the verification set to verify the preliminary training model;

S2033. If the verification is passed, the preliminary training model is recorded as the report type prediction model.

As mentioned above, the acquisition of the report type prediction model is realized. The characteristic information of the user refers to information that can reflect the characteristics of the user, such as the occupation of the user, the type of report currently used recently (within a specified time), the age of the user, and the gender of the user. The sample data of the verification set is used to predict the report type of the current user in order to verify the preliminary training model, so the verification set of the specified amount of sample data is preferably sample data related to the current user, such as samples with the same occupation and the same age data. When the verification passes, it indicates that the preliminary training model is available, and accordingly the preliminary training model is recorded as the report type prediction model.

S211: Obtain a specified amount of sample data, where the sample data includes user feature information and a report type corresponding to the user feature information;

S212. Input the sample data of the training set into the CHAID decision tree model for training to obtain a preliminary CHAID decision tree;

S213. Record the preliminary CHAID decision tree as the report type prediction model.

As described above, the report type prediction model is obtained. The decision tree is classified according to the user's characteristic information to predict the type of report that the user will adopt. The characteristic information of the user refers to information that can reflect the characteristics of the user, such as the occupation of the user, the type of report currently used recently (within a specified time), the age of the user, and the gender of the user. The CHAID decision tree model refers to the decision tree model that uses the chi-square automatic interactive detection method CHAID. Here is a brief introduction to the principles of the CHAID decision tree: 1. Group values in the merged group that have no significant difference in decision variables; 2. Select the variable with the largest chi-square value as the tree classification variable; 3. Repeat steps 1 and 2 Choose the chi-square value greater than a certain value or P value is no longer less than a critical value, or the sample is less than a certain number. Among them, the modeling criteria of the CHAID decision tree model is, for example, the maximum number of layers of the tree is 3, the significance level of the subdivision of the parent node is 0.05, the minimum number of samples included in the parent node is 100, and the minimum value included in the child nodes is The number of samples is 50. Further, the recording of the preliminary CHAID decision tree as the report type prediction model further includes: verifying the preliminary CHAID decision tree using a verification set composed of sample data obtained in advance; if the verification passes, then the The preliminary CHAID decision tree is recorded as the report type prediction model.

In one embodiment, the step S212 of inputting the sample data of the training set into the CHAID decision tree model to obtain a preliminary CHAID decision tree includes:

S2121. Set the modeling standard parameters of the CHAID decision tree model. The modeling standard parameters include the maximum number of decision trees, the subdividable significance level of the parent node, the minimum number of samples included by the parent node, and the The minimum number of samples;

S2122. Input the sample data of the training set into the CHAID decision tree model established by the chi-square automatic interactive detection method for training to obtain a preliminary CHAID decision tree.

As mentioned above, the preliminary CHAID decision tree is achieved. The CHAID decision tree model can only be determined by setting standard modeling parameters of the CHAID decision tree model. The modeling standard parameters include the maximum number of decision trees, the subdividable significance level of the parent node, the minimum number of samples contained by the parent node and the minimum number of samples contained by the child nodes, for example, the maximum number of layers of the tree is 3-5 The significance level of the subdivision of the parent node is 0.05, the minimum number of samples included in the parent node is 100-200, and the minimum number of samples included in the child node is 50-100.

The machine learning-based report generation method of this application uses a machine-learning-based report type prediction model to predict the type of report that the current user will use, and then retrieves a preset preliminary report from the database according to the report type. The preliminary report is adjusted to obtain the final report, thereby avoiding the user's tedious operations, improving the user experience, and improving the efficiency of report completion.

Referring to FIG. 2, an embodiment of the present application provides a report generation device based on machine learning, including:

The characteristic information obtaining unit 10 is configured to obtain characteristic information of the current user, and the characteristic information of the current user includes at least the occupation information of the current user;

A report type prediction model calculation unit 20, configured to input the feature information into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model is composed of user feature information and the user feature information Trained by the training data composed of the corresponding report types;

The report type prediction unit 30 is used to output the predicted report type to be used by the current user;

The preliminary report retrieval unit 40 is configured to retrieve a preset preliminary report from the database according to the type of report to be used, wherein the type of the preliminary report is the same as the type of report to be used;

The final report obtaining unit 50 is configured to adjust the preliminary report according to the information input by the current user, thereby obtaining the final report.

As described in the above unit 10, the characteristic information of the current user is obtained, and the characteristic information of the current user includes at least the occupation information of the current user. Wherein, the characteristic information of the current user refers to information that can reflect the characteristics of the current user, such as the occupation of the current user, the type of report used by the current user (within a specified time), the age of the current user, and the gender of the current user. Among them, the occupation information of the current user has a greater relationship with the type of report that may be used, so the characteristic information of the current user includes at least the occupation information of the current user, so as to improve the accuracy of the prediction of the report type. For example, stock commentators in the financial industry have a higher probability of using stock statements. The process of acquiring the characteristic information of the current user includes: extracting the characteristic information of the current user from the registered account information of the current user.

As described in the above unit 20, the feature information is input into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model passes user feature information and a report type corresponding to the user feature information Trained. The report type prediction model based on machine learning, through continuous self-learning, improves the accuracy of prediction and avoids predicting the wrong report type. The report type prediction model can be generated based on any machine learning model, such as a neural network model and a classification tree model, and then trained through training data. Wherein, the report type prediction model is trained by training data composed of user characteristic information and the report type corresponding to the user characteristic information. The user characteristic information may include any information that can reflect the characteristics of the user, such as the occupation of the user, the type of report recently used by the user (within a specified time), the age of the user, and the gender of the user. Among them, the user refers to a user who uses a report. Wherein, the report type prediction model is used to predict the report type to be used by the user based on the user characteristic information.

As described in the above unit 30, the predicted report type to be used by the current user is output. The report type can be classified into any classification method. For example, it can be divided into the following chart types: pie chart report (including pie chart in the report), curve chart report, and histogram report; : Financial statements, financial statements, statistical analysis reports. Through the report type prediction model, the predicted report type to be used by the current user can be output.

As described in the above unit 40, according to the report type to be used, a preset preliminary report is retrieved from the database, wherein the type of the preliminary report is the same as the type of the report to be used. There are pre-stored preliminary reports of different report types in the preset database, and there can be multiple preliminary reports of the same report type for users to choose. For example, the preliminary report of the pie chart report can include three types of reports: the pie chart is at the top of the report, the middle of the report, and the end of the report; the financial preliminary report can include the report with the K chart, the report with the histogram, and the report with the curve chart. Reports, etc. These preliminary reports can be directly selected by the user, thereby eliminating the user's trivial process of creating reports from scratch, and improving the speed of report formation. Wherein, the preset preliminary report may be a completed complete report, or the current user may select the exact chart template and text part template from multiple alternative chart templates and text part templates. The combined report.

As described in the above unit 50, the preliminary report is adjusted according to the information input by the current user, thereby obtaining the final report. From the foregoing, a preliminary report is provided, and then the final report can be obtained based on the information input by the current user. Wherein, the information input by the current user includes at least one of information for adjusting the parameters of the preliminary report and text content information for describing the preliminary report. Wherein, the final report includes at least a chart part, and further, may include a text part.

In one embodiment, the preliminary report retrieval unit 40 includes:

The drawing template and text part template retrieval subunit is used to retrieve a preset plurality of chart templates and a plurality of text part templates from a preset database according to the report type to be used;

A combination subunit, configured to combine the graph template and the text part template selected by the current user into the preliminary report;

The calling subunit is used for calling the preliminary report.

In one embodiment, the final report obtaining unit 50 includes:

An adjustment subunit for adjusting the chart and text in the preliminary report according to the chart adjustment information, chart data content information and text part adjustment information input by the current user;

The final report obtaining subunit is used to fill the text content input by the current user into the text portion of the preliminary report to obtain the final report.

As described above, the preliminary report is adjusted according to the information input by the current user, thereby obtaining the final report. As mentioned earlier, preliminary reports have been obtained. However, some detailed parameters and specific data content of the preliminary report are not supplemented. According to this, the adjustment is made according to the specific instructions of the current user, wherein the chart adjustment information includes: adjusting the size of the chart, the data display parameters of the chart (such as the unit time length of the time axis, etc.), etc.; the chart data content information includes : Chart data (such as the data points of the graph, the proportion of each block of the pie chart, etc.); the adjustment information of the text part includes: adjusting the font size and color. Then, the text content input by the current user is filled in the text part to obtain a final report.

In one embodiment, the device includes a report type prediction model acquisition unit, and the report type prediction model acquisition unit includes:

A training set acquisition subunit for acquiring a training set including a specified amount of sample data, wherein the sample data includes user feature information and a report type corresponding to the user feature information;

The neural network model training subunit is used to input the sample data of the training set into the neural network model for training, in which the stochastic gradient descent method is used in the training process, and the neural network model is updated using the reverse conduction law The parameters of the layer to get the preliminary training model;

The report type prediction model marking subunit is used to record the preliminary training model as the report type prediction model.

In one embodiment, the report type prediction model marking subunit includes:

A verification set obtaining module, configured to obtain a verification set including a specified amount of sample data, wherein the sample data of the verification set includes user characteristic information and a report type corresponding to the user characteristic information;

A verification module, configured to use the sample data of the verification set to verify the preliminary training model;

The report type prediction model marking module is used to record the preliminary training model as the report type prediction model if the verification is passed.

As described above, the acquisition of the report type prediction model is realized. The characteristic information of the user refers to information that can reflect the characteristics of the user, such as the occupation of the user, the type of report currently used recently (within a specified time), the age of the user, and the gender of the user. The sample data of the verification set is used to predict the report type of the current user in order to verify the preliminary training model, so the verification set of the specified amount of sample data is preferably sample data related to the current user, such as samples with the same occupation and the same age data. When the verification passes, it indicates that the preliminary training model is available, and accordingly the preliminary training model is recorded as the report type prediction model.

A sample data obtaining subunit, configured to obtain a specified amount of sample data, wherein the sample data includes user characteristic information and a report type corresponding to the user characteristic information;

The decision tree model training subunit is used to input the sample data of the training set into the CHAID decision tree model for training to obtain a preliminary CHAID decision tree;

The decision tree marking subunit is used to record the preliminary CHAID decision tree as the report type prediction model.

In one embodiment, the decision tree model training subunit includes:

The modeling standard parameter setting module is used to set the modeling standard parameters of the CHAID decision tree model, the modeling standard parameters include the maximum number of decision trees, the subdividable significance level of the parent node, the parent node contains The minimum number of samples and the minimum number of samples contained in the child node;

The preliminary CHAID decision tree obtaining module is used for inputting the sample data of the training set into the CHAID decision tree model established by the chi-square automatic interactive detection method for training to obtain a preliminary CHAID decision tree.

The report generation device based on machine learning of this application uses the report type prediction model based on machine learning to predict the report type that the current user will use, and then retrieves the preset preliminary report from the database according to the report type. The preliminary report is adjusted to obtain the final report, thereby avoiding the user's tedious operations, improving the user experience, and improving the efficiency of report completion.

Referring to FIG. 3, an embodiment of the present invention further provides a computer device. The computer device may be a server, and its internal structure may be as shown in the figure. The computer device includes a processor, memory, network interface, and database connected by a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer-readable instructions, and a database. The memory device provides an environment for the operation of the operating system and computer-readable instructions in the non-volatile storage medium. The database of the computer device is used to store data used in the report generation method based on machine learning. The network interface of the computer device is used to communicate with external terminals through a network connection. When the computer-readable instructions are executed by the processor, a report generation method based on machine learning is realized.

The above processor executes the above machine learning-based report generation method, including the following steps: acquiring the current user's feature information, the current user's feature information includes at least the current user's occupation information; and entering the feature information into a preset machine-based report Operation in the learned report type prediction model, where the report type prediction model is trained by training data composed of user characteristic information and the report type corresponding to the user characteristic information; the predicted current user will output The type of report used; according to the type of report to be used, a preset preliminary report is retrieved from the database, wherein the type of the preliminary report is the same as the type of report to be used; according to the information entered by the current user , Adjust the preliminary report to obtain the final report.

In one embodiment, the step of retrieving a preset preliminary report from the database according to the type of report to be used, wherein the step of the type of the preliminary report being the same as the type of report to be used includes: The report type to be used, extracting preset multiple chart templates and multiple text part templates from a preset database; combining the current user selected chart template and text part template into the preliminary report; Recall the preliminary report.

In one embodiment, the step of adjusting the preliminary report according to the information input by the current user to obtain the final report includes: adjusting information, content information of the chart data, and text according to the chart input by the current user Adjust the information, adjust the chart and text parts in the preliminary report; fill the text part of the current user into the text part of the preliminary report to obtain the final report.

In one embodiment, the method for obtaining a report type prediction model includes: obtaining a training set including a specified amount of sample data, wherein the sample data includes user feature information and a report corresponding to the user feature information Type; input the sample data of the training set into the neural network model for training, in which the stochastic gradient descent method is used in the training process, and the parameters of each layer of the neural network model are updated by the reverse conduction method to obtain preliminary training Model; record the preliminary training model as the report type prediction model.

In one embodiment, the step of recording the preliminary training model as the report type prediction model includes: obtaining a verification set including a specified amount of sample data, wherein the sample data of the verification set includes user characteristics Information, and the report type corresponding to the user feature information; verify the preliminary training model using the sample data of the verification set; if the verification passes, then record the preliminary training model as the report type prediction model.

In one embodiment, the method for obtaining a report type prediction model includes: obtaining a specified amount of sample data, wherein the sample data includes user feature information and a report type corresponding to the user feature information; The sample data of the training set is input into the CHAID decision tree model for training to obtain a preliminary CHAID decision tree; the preliminary CHAID decision tree is recorded as the report type prediction model.

In one embodiment, the step of inputting the sample data of the training set into the CHAID decision tree model to obtain a preliminary CHAID decision tree includes: setting modeling standard parameters of the CHAID decision tree model. The modeling standard parameters include the maximum number of decision trees, the subdividable significance level of the parent node, the minimum number of samples included in the parent node and the minimum number of samples included in the child node; input the sample data of the training set to the chi-square automatic The CHAID decision tree model established by the interactive detection method is trained to obtain a preliminary CHAID decision tree.

Those skilled in the art can understand that the structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.

The computer equipment of this application uses the report type prediction model based on machine learning to predict the report type that the current user will use, and then retrieves the preset preliminary report from the database according to the report type and adjusts the preliminary report In order to get the final report, so as to avoid the user's tedious operations, improve the user experience, and improve the efficiency of report completion.

An embodiment of the present application also provides a computer-readable storage medium on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, a method for generating a report form based on machine learning includes the following steps: obtaining the current user’s Feature information, the feature information of the current user includes at least the occupation information of the current user; the feature information is input into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model is determined by the user feature information , And training data composed of the report type corresponding to the user characteristic information is trained; output the predicted report type that the current user will use; according to the report type to be used, the preset is retrieved from the database The preliminary report, wherein the type of the preliminary report is the same as the type of the report to be used; according to the information input by the current user, the preliminary report is adjusted to obtain the final report. The computer-readable storage medium is, for example, a non-volatile computer-readable storage medium or a volatile computer-readable storage medium.

The computer-readable storage medium of the present application predicts the report type that the current user will use through a report type prediction model based on machine learning, and then retrieves the preset preliminary report from the database according to the report type After adjustment, the final report is obtained, thereby avoiding the user's tedious operations, improving the user experience, and improving the efficiency of report completion.

A person of ordinary skill in the art may understand that all or part of the process in the method of the foregoing embodiments may be completed by instructing relevant hardware through computer-readable instructions, and the computer-readable instructions may be stored in a non-volatile computer In a readable storage medium, when the computer-readable instructions are executed, they may include the processes of the foregoing method embodiments. Wherein, any reference to the memory, storage, database or other media provided in the present application and used in the embodiments may include non-volatile and/or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

It should be noted that in this article, the terms "include", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements It also includes other elements that are not explicitly listed, or include elements inherent to such processes, devices, objects, or methods. Without more restrictions, the element defined by the sentence "include one..." does not exclude that there are other identical elements in the process, device, article or method that includes the element.

The above are only the preferred embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by the description and drawings of this application, or directly or indirectly used in other related In the technical field, the same reason is included in the scope of patent protection of this application.

Claims

A report generation method based on machine learning is characterized by including:

Acquiring characteristic information of the current user, the characteristic information of the current user includes at least the occupation information of the current user;

The feature information is input into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model is trained by training data composed of user feature information and a report type corresponding to the user feature information Made

Output the predicted report type that the current user will use;

According to the type of report to be used, a preset preliminary report is retrieved from the database, wherein the type of the preliminary report is the same as the type of report to be used;

According to the information input by the current user, the preliminary report is adjusted to obtain a final report.
The method for generating a report based on machine learning according to claim 1, characterized in that the preset preliminary report is retrieved from the database according to the type of report to be used, wherein the type of the preliminary report and all Describe the same steps as the type of report to be used, including:

According to the type of report to be used, retrieve preset multiple chart templates and multiple text part templates from a preset database;

Combining the chart template and the text part template selected by the current user into the preliminary report;

Recall the preliminary report.
The method for generating a report based on machine learning according to claim 1, wherein the step of adjusting the preliminary report according to the information input by the current user to obtain the final report includes:

Adjust the graph and text parts in the preliminary report according to the graph adjustment information, graph data content information and text part adjustment information input by the current user;

The text content input by the current user is filled into the text portion of the preliminary report to obtain the final report.
The method for generating a report form based on machine learning according to claim 1, wherein the method for obtaining a report type prediction model includes:

Obtaining a training set including a specified amount of sample data, where the sample data includes user feature information and a report type corresponding to the user feature information;

Input the sample data of the training set into the neural network model for training, in which the stochastic gradient descent method is used in the training process, and the parameters of each layer of the neural network model are updated by the reverse conduction rule to obtain the preliminary training model;

The preliminary training model is recorded as the report type prediction model.
The method for generating a report based on machine learning according to claim 4, wherein the step of recording the preliminary training model as the report type prediction model includes:

Obtaining a verification set including a specified amount of sample data, wherein the sample data of the verification set includes user feature information and a report type corresponding to the user feature information;

Verify the preliminary training model using the sample data of the verification set;

If the verification is passed, the preliminary training model is recorded as the report type prediction model.
The method for generating a report form based on machine learning according to claim 1, wherein the method for obtaining a report type prediction model includes:

Obtain a specified amount of sample data, where the sample data includes user feature information and a report type corresponding to the user feature information;

Input the sample data of the training set into the CHAID decision tree model for training to obtain a preliminary CHAID decision tree;

The preliminary CHAID decision tree is recorded as the report type prediction model.
The method for generating a report based on machine learning according to claim 6, wherein the step of inputting sample data of the training set into a CHAID decision tree model for training to obtain a preliminary CHAID decision tree includes:

Set the modeling standard parameters of the CHAID decision tree model, the modeling standard parameters include the maximum number of decision trees, the subdividable significance level of the parent node, the minimum number of samples contained in the parent node and the minimum samples contained in the child node number;

The sample data of the training set is input into the CHAID decision tree model established by the chi-square automatic interactive detection method for training to obtain a preliminary CHAID decision tree.
A report generation device based on machine learning is characterized by including:

A characteristic information obtaining unit, configured to obtain characteristic information of the current user, the characteristic information of the current user includes at least the occupation information of the current user;

Report type prediction model operation unit, used for inputting the feature information into a preset report type prediction model based on machine learning, wherein the report type prediction model is determined by user feature information and corresponding to the user feature information The training data composed of the report type is trained;

A report type prediction unit, used to output the predicted report type that the current user will use;

A preliminary report retrieval unit, configured to retrieve a preset preliminary report from the database according to the type of report to be used, wherein the type of the preliminary report is the same as the type of report to be used;

The final report obtaining unit is used to adjust the preliminary report according to the information input by the current user, so as to obtain the final report.
The report generation device based on machine learning according to claim 8, wherein the preliminary report retrieval unit includes:

The drawing template and text part template retrieval subunit is used to retrieve a preset plurality of chart templates and a plurality of text part templates from a preset database according to the report type to be used;

A combination subunit, configured to combine the graph template and the text part template selected by the current user into the preliminary report;

The calling subunit is used for calling the preliminary report.
The apparatus for generating a report based on machine learning according to claim 8, wherein the final report obtaining unit includes:

An adjustment subunit for adjusting the chart and text in the preliminary report according to the chart adjustment information, chart data content information and text part adjustment information input by the current user;

The final report obtaining subunit is used to fill the text content input by the current user into the text portion of the preliminary report to obtain the final report.
The report generation device based on machine learning according to claim 8, characterized in that the device includes a report type prediction model acquisition unit, and the report type prediction model acquisition unit includes:

A training set acquisition subunit for acquiring a training set including a specified amount of sample data, wherein the sample data includes user feature information and a report type corresponding to the user feature information;

The neural network model training subunit is used to input the sample data of the training set into the neural network model for training, in which the stochastic gradient descent method is used in the training process, and the neural network model is updated using the reverse conduction law The parameters of the layer to get the preliminary training model;

The report type prediction model marking subunit is used to record the preliminary training model as the report type prediction model.
The apparatus for generating a report based on machine learning according to claim 11, wherein the report type prediction model marking subunit includes:

A verification set obtaining module, configured to obtain a verification set including a specified amount of sample data, wherein the sample data of the verification set includes user characteristic information and a report type corresponding to the user characteristic information;

A verification module, configured to use the sample data of the verification set to verify the preliminary training model;

The report type prediction model marking module is used to record the preliminary training model as the report type prediction model if the verification is passed.
The report generation device based on machine learning according to claim 8, characterized in that the device includes a report type prediction model acquisition unit, and the report type prediction model acquisition unit includes:

A sample data obtaining subunit, configured to obtain a specified amount of sample data, wherein the sample data includes user characteristic information and a report type corresponding to the user characteristic information;

The decision tree model training subunit is used to input the sample data of the training set into the CHAID decision tree model for training to obtain a preliminary CHAID decision tree;

The decision tree marking subunit is used to record the preliminary CHAID decision tree as the report type prediction model.
The report generation device based on machine learning according to claim 13, wherein the decision tree model training subunit includes:

The modeling standard parameter setting module is used to set the modeling standard parameters of the CHAID decision tree model, the modeling standard parameters include the maximum number of decision trees, the subdividable significance level of the parent node, the parent node contains The minimum number of samples and the minimum number of samples contained in the child node;

The preliminary CHAID decision tree obtaining module is used for inputting the sample data of the training set into the CHAID decision tree model established by the chi-square automatic interactive detection method for training to obtain a preliminary CHAID decision tree.
A computer device includes a memory and a processor. The memory stores computer-readable instructions. The processor is characterized in that when the processor executes the computer-readable instructions, a report generation method based on machine learning is implemented. Learned report generation methods, including:

Acquiring characteristic information of the current user, the characteristic information of the current user includes at least the occupation information of the current user;

The feature information is input into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model is trained by training data composed of user feature information and a report type corresponding to the user feature information Made

Output the predicted report type that the current user will use;

According to the type of report to be used, a preset preliminary report is retrieved from the database, wherein the type of the preliminary report is the same as the type of report to be used;

According to the information input by the current user, the preliminary report is adjusted to obtain a final report.
The computer device according to claim 15, wherein the preset preliminary report is retrieved from a database according to the type of report to be used, wherein the type of the preliminary report and the report to be used Steps of the same type, including:

According to the type of report to be used, retrieve preset multiple chart templates and multiple text part templates from a preset database;

Combining the chart template and the text part template selected by the current user into the preliminary report;

Recall the preliminary report.
The computer device according to claim 15, wherein the step of adjusting the preliminary report based on the information input by the current user to obtain the final report includes:

Adjust the graph and text parts in the preliminary report according to the graph adjustment information, graph data content information and text part adjustment information input by the current user;

The text content input by the current user is filled into the text portion of the preliminary report to obtain the final report.
A computer-readable storage medium on which computer-readable instructions are stored, characterized in that, when the computer-readable instructions are executed by a processor, a report generation method based on machine learning is realized, and the report generation method based on machine learning ,include:

Acquiring characteristic information of the current user, the characteristic information of the current user includes at least the occupation information of the current user;

The feature information is input into a preset report type prediction model based on machine learning for calculation, wherein the report type prediction model is trained by training data composed of user feature information and a report type corresponding to the user feature information Made

Output the predicted report type that the current user will use;

According to the type of report to be used, a preset preliminary report is retrieved from the database, wherein the type of the preliminary report is the same as the type of report to be used;

According to the information input by the current user, the preliminary report is adjusted to obtain a final report.
The computer-readable storage medium according to claim 18, wherein the preset preliminary report is retrieved from a database according to the type of report to be used, wherein the type of the preliminary report and the Use the same steps for the report type, including:

According to the type of report to be used, retrieve preset multiple chart templates and multiple text part templates from a preset database;

Combining the chart template and the text part template selected by the current user into the preliminary report;

Recall the preliminary report.
The computer-readable storage medium of claim 18, wherein the step of adjusting the preliminary report based on the information input by the current user to obtain a final report includes:

Adjust the graph and text parts in the preliminary report according to the graph adjustment information, graph data content information and text part adjustment information input by the current user;

The text content input by the current user is filled into the text portion of the preliminary report to obtain the final report.