Data display method, device, equipment and storage medium

Abstract

The embodiment of the present application provides a data display method, device, equipment and storage medium. The method includes: receiving input operations on a target data set and a target model in a display interface, the target data set includes at least one data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension; displaying weight data of the processing result data in the display interface; wherein the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one feature dimension to the processing result data, which can facilitate users to select different target models and target data sets to be predicted according to their needs to interpret the weights of feature dimensions, and facilitate users to intuitively view the contribution of feature dimensions to the prediction results, and provide more information support for user decision-making.

Description

Technical Field

[0001] The present application relates to the field of communication technology, and in particular to a data display method and device, an electronic device and a storage medium. Background Art

[0002] With the development of technology, the application of machine learning models is increasing, and it can provide more accurate processing result data. For example, in tax audits, machine learning models can predict the high risk of a certain user. For auxiliary diagnosis and treatment, such as the diagnosis of patients with new coronary pneumonia, machine learning models can also be used to predict the risk of a certain user getting sick.

[0003] However, due to the black-box nature of machine learning models, decision makers can only input the predicted data to be analyzed into the machine learning model and then obtain the processed result data. This method only knows the input and output of the model, and does not make it convenient for users to know why there is this output, so there is little information for users to make decisions. Summary of the invention

[0004] The embodiment of the present application provides a data display method to provide users with more information in a visual manner for the processing result data of the target model.

[0005] Correspondingly, the embodiment of the present application also provides a data display device, an electronic device and a storage medium to ensure the implementation and application of the above method.

[0006] In order to solve the above problems, the present application discloses a data display method, including:

[0007] Receiving input operations on a target data set and a target model in a display interface, wherein the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension;

[0008] The weight data of the processing result data are displayed in the display interface; wherein the processing result data include: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data include: the target weight of at least one of the feature dimensions to the processing result data.

[0009] The present application also discloses a data display method, including:

[0010] Providing a display interface to the client; receiving a target data set and a target model input by the client through the display interface; the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension;

[0011] Send weight data of the processing result data to the client for display in the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data.

[0012] The present application also discloses a data display method, including:

[0013] Acquire a display interface from a server; receive an input operation on a target data set and a target model in the display interface, and send the target data set and the target model to the server;

[0014] Receive weight data for the processing result data returned by the server and display it on the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data.

[0015] The present application also discloses a data display and analysis device, which includes:

[0016] A first receiving module is used to receive an input operation on a target data set and a target model in a display interface, wherein the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension;

[0017] The first display module is used to display weight data of the processing result data on the display interface; wherein the processing result data includes: the processing results obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions to the processing result data.

[0018] The present application also discloses a data display and analysis device, which is applied to an interpretation server, and the device includes:

[0019] A page providing module is used to provide a display interface to the client; a second receiving module is used to receive a target data set and a target model input by the client through the display interface; the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension;

[0020] A result sending module is used to send weight data of the processing result data to the client for display in the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data.

[0021] The present application also discloses a data display and analysis device, which is applied to a client, and the device includes:

[0022] a page acquisition module, used to acquire a display interface from a server; a third receiving module, used to receive input operations on a target data set and a target model on the display interface, and send the target data set and the target model to the server;

[0023] The second display module is used to receive weight data of the processing result data returned by the server and display it on the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data.

[0024] An embodiment of the present application further discloses an electronic device, including: a processor; and a memory, on which executable code is stored. When the executable code is executed, the processor executes one or more methods described in the embodiments of the present application.

[0025] The embodiments of the present application also disclose one or more machine-readable media on which executable codes are stored. When the executable codes are executed, the processor executes one or more methods described in the embodiments of the present application.

[0026] Compared with the prior art, the embodiments of the present application have the following advantages:

[0027] In an embodiment of the present application, first, input operations on a target data set and a target model are received in a visual display interface, so that the user can select different target models and target data sets to be predicted according to needs to interpret the weights of feature dimensions.

[0028] Secondly, displaying the weight of at least one feature dimension of the data to be analyzed on the processing result data in a visual display interface can facilitate users to intuitively view the contribution of the feature dimension to the prediction result and provide more information support for user decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic diagram of the architecture of a data display system according to an embodiment of the present application;

[0030] Figure 2is a flowchart of another data display method embodiment of the present application;

[0031] Figure 3 is a schematic diagram of a display interface of the present application;

[0032] Figure 4 is a schematic diagram of another display interface of the present application;

[0033] Figure 5 is a schematic diagram of another display interface of the present application;

[0034] Figure 6 is a schematic diagram of another display interface of the present application;

[0035] Figure 7 is a schematic diagram of another display interface of the present application;

[0036] Figure 8 is a schematic diagram of another display interface of the present application;

[0037] Fig. 9 It is a flowchart of another data display method embodiment of the present application on the server side;

[0038] Fig.10 is a flowchart of another embodiment of a method for displaying data on a client side of the present application;

[0039] Fig.11 is a structural block diagram of an embodiment of a data display device of the present application;

[0040] Fig.12 is a structural block diagram of another data display device embodiment of the present application;

[0041] Fig.13 is a structural block diagram of another data display device embodiment of the present application;

[0042] Fig.14 It is a schematic diagram of the structure of a device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0044] In the prior art, many scenarios use machine learning models to predict the data to be analyzed, but the machine learning model is a black box for users. After the user inputs the data to be analyzed, a processing result data is obtained. The user does not know which feature in the data to be analyzed is important to the processing result data. In some scenarios, the black box nature of this machine learning model will lead to insufficient information dimensions when users make decisions. For example, in tax audits, machine learning models predict that the enterprise has high risks based on the relevant data of the enterprise, but audits must pay attention to evidence. It is not enough for the model to only indicate high risks. Important risk points must also be given and evidence must be provided from the data. For example, from the various feature dimensions of the enterprise's relevant data, determine which feature dimensions have high weights. The high weight means that the feature dimension contributes greatly to the high risk. It can be understood that the above-mentioned screening of feature dimensions with high weights can extract from the data which means of the enterprise that produced the high-risk result. For example, for auxiliary diagnosis and treatment, machine learning models can make diagnoses based on the patient's monitoring indicators, but doctors still have to explain the various indicators given in the diagnosis results, which will increase doctors' trust in the machine learning model and better serve patients. For example, which indicators of the patient have high weights? This high weight means that the indicator contributes greatly to the diagnosis result, so the doctor can explain to the patient based on this indicator with a large contribution.

[0045] Reference Figure 1 , which shows a data display system architecture diagram of an embodiment of the present application. In the embodiment of the present application, it may include: a client 101, an interpretation server 102, and a data source, such as a tax data service end 103, a patient data service end 104, a customs data service end 108, and an industrial and commercial data service end 109.

[0046] It should be noted that the data source may be a database, and the client may be a device with a display screen, such as a personal computer or a mobile terminal, which is not limited in the embodiments of the present application.

[0047] It should be noted that, in the embodiment of the present application, the data in the data source can be obtained after the consent of the data owner. For example, various characteristic data of the patient can be obtained after the consent of the patient, and other situations are similar and will not be described in detail again.

[0048] Of course, other service terminals may also be included, such as a customs declaration data service terminal in a customs scenario, an enterprise behavior data service terminal in an industrial and commercial scenario, etc., which are not limited by the embodiments of the present application.

[0049] In the embodiment of the present application, the above data source may store a target data set to be analyzed, and the target data set may include at least one piece of data to be analyzed, and the data to be analyzed may include at least one feature dimension. Figure 1The patient data server 104 in FIG. 1 stores characteristic data of multiple patients, and the characteristic dimensions of the characteristic data include user ID, gender, age, data of various symptoms, etc. For another example, the tax data server 103 in FIG. 1 stores characteristic dimensions of the characteristic data such as tax amount, tax time, various tax types, etc. for tax audit. Figure 1 In the customs data server 108, for customs declaration, the characteristic dimensions of the characteristic data include the name of the declared commodity, the type of commodity, the quantity of commodity, the weight of commodity, the size of commodity, the time of declaration, the place of declaration, the person who declares the customs, etc. For example, Figure 1 The industrial and commercial data service end 109 in the monitoring of corporate behavior includes characteristic dimensions of the characteristic data such as corporate registration behavior, various business operations of the enterprise, and so on.

[0050] It should be noted that the target model in the embodiment of the present application can be a machine learning model pre-trained using corresponding training data, such as a neural network model, a support vector machine model, etc., and the implementation of this application does not limit it. For example, for the aforementioned tax audit, the tax data of the enterprise can be used in advance to construct a training set, and then the training set is used to train a machine learning model for predicting the tax risk of the enterprise. For example, for the aforementioned auxiliary diagnosis, the relevant data of the patient can be used in advance to construct a training set, and then the training set is used to train a machine learning model for identifying the patient's disease results. In auxiliary diagnosis, for example, to identify the probability of a user suffering from new coronary pneumonia, the various characteristic data of the patient and the corresponding conclusion of whether it is confirmed to be new coronary pneumonia can be used to construct a training set, and then the training set is used to train a machine learning model for identifying whether the target population is at risk of new coronary pneumonia.

[0051] It should be noted that in actual scenarios, there can be multiple data sources, and there may be multiple data sources for the same scenario. For example, for tax audits, there may be a database in multiple regions, and each database stores the tax data of enterprises in that region. For auxiliary diagnosis, there may also be multiple regions, and each region also has a corresponding database, which stores the patient data of that region.

[0052] It should be noted that in the embodiments of the present application, different machine learning models may be trained for different data sources, such as the machine learning model used by data source 1 is model A, and the machine learning model used by data source 2 is model B. Of course, for the same data source, multiple machine learning models may also be used, and each machine learning model corresponds to analyzing different types of data in the data source. For example, for data source A, there are 3 types of data Q1, Q2, and Q3, and the machine learning models A1, A2, and A3 are used for processing during prediction. Among them, different types of data in the same data source can be stored in different data tables.

[0053] like Figure 1 In an embodiment of an information display method of the present application, the method may include the following process:

[0054] A1, a user initiates an operation of accessing an information display page in the client 101.

[0055] exist Figure 1 In the present invention, the user can initiate an operation of accessing the information display page in the client, such as triggering access to a preset URL in a browser or other application.

[0056] It should be noted that in the case of a client using a mouse or keyboard input method, the user can use the mouse or keyboard to initiate an operation of accessing the information display page. In the case of a client using a touch screen input method, the user can use the touch screen to initiate an operation of accessing the information display page.

[0057] It should be noted that the display interface in the embodiment of the present application can be the information display page.

[0058] A2, the client 101 requests the information display page from the interpretation server.

[0059] After receiving the user's operation for accessing the information display page, the client requests the information display page from the interpretation server.

[0060] In actual applications, the browser may send a page request to the interpretation server based on the link corresponding to the information display page, and the interpretation server may then return the information display page based on the page request.

[0061] A3, explaining that the server 102 returns the information display page to the client.

[0062] The explanation server 102 returns the information display page to the client according to the above request.

[0063] After the client 101 obtains the information display page, it displays the information display page on the client screen, thereby providing the user with a display interface for receiving input operations on the target data set and the target model, and for displaying weight data of the processing result data.

[0064] Among them, the information display page can be as follows Figure 1 In the example 105, the information display page 105 may include a first interface area 106 and a second interface area 107, wherein the first interface area may be used to receive input operations on the target data set and the target model, and the second interface area may display weight data for the processing result data.

[0065] The weight data may include an indication of the weight of each feature dimension for obtaining the processing result data. Of course, the weight data may also include other data, such as the association between the weight and the dimension, etc. The embodiment of the present application does not limit it.

[0066] It should be noted that the positions of the first interface area 106 and the second interface area 107 can be arranged according to actual needs, such as side by side on the left and right, etc., and the embodiment of the present application does not limit it. Of course, the information display page may not clearly divide the first interface area 106 and the second interface area 107. It is sufficient to set controls for receiving the above input operations and an area for displaying processing results in the information display page, and the embodiment of the present application does not limit it.

[0067] A4, receiving the target model and target data set input by the user in the first interface area 106 of the information display page 105 displayed on the client 101.

[0068] It should be noted that the target model may be the aforementioned machine learning model for predicting the processing result data corresponding to the data to be analyzed.

[0069] As described above, there may be multiple data sources, and there may also be multiple data tables in the data source, resulting in multiple data sets to be analyzed and multiple machine learning models. Therefore, users can select target data sets and target models according to their needs.

[0070] like Figure 1 The first interface area in the information display page 105 may include: a model identifier input control, a data set identifier input control, an explanation control, etc. The user may input the model identifier AAA of the target model in the model identifier input control, input the target data set identifier BBB in the data set identifier input control, and then click the explanation control to send the above information.

[0071] It should be noted that the target data set includes at least one of a tax data set, a customs declaration data set, a medical data set, and a corporate behavior data set; the target model includes: a model for predicting the taxpayer's tax risk based on the tax data set, a model for predicting the customs declaration risk of the customs declaration person based on the customs declaration data set, a model for predicting the user's illness based on the medical data set, and a model for predicting the corporate behavior risk based on the corporate behavior data.

[0072] The tax data set may include tax behavior data of taxpayers under various feature dimensions, etc. The model for predicting the tax risk of the taxpayer based on the tax data set may predict the tax risk of the taxpayer, such as tax evasion risk, tax false reporting risk, etc.

[0073] The customs declaration data set may include the declaration behavior data of the customs declarant under various feature dimensions, various attribute information of the declared items, etc., such as product name, model, declared quantity, etc. The model for predicting the declaration risk of the customs declarant based on the customs declaration data set can predict the declaration risk of the customs declarant, such as the risk of carrying counterfeit goods, the risk of smuggling, etc.

[0074] The medical data set may include patient-related information under various feature dimensions. The model for predicting the user's disease condition based on the medical data set may be used to predict the risk of the patient suffering from a certain disease.

[0075] The enterprise behavior data may include enterprise registration behavior, various business behavior data of the enterprise, etc. At least one of the models for predicting enterprise behavior risks based on enterprise behavior data can predict whether the enterprise behavior complies with regulations, thereby facilitating the supervision of enterprise behavior.

[0076] In actual applications, the data set and the model are set together. For example, if a tax data set is set, a model for predicting the taxpayer's tax risk based on the tax data set can be set simultaneously; if a customs declaration data set is set, a model for predicting the customs declaration risk of the customs declarant based on the customs declaration data set can be set. Other situations are similar and will not be described in detail here.

[0077] It can be understood that for a certain scenario, there can be at least one data set and at least one model. For example, for a tax scenario, multiple tax data sets can be set by region. Correspondingly, multiple models can be set for predicting the tax risk of taxpayers based on the tax data set, so that it is convenient to choose according to the needs of users. For medical scenarios, multiple medical data sets can also be set by region, hospital, disease type, etc. For example, a patient data set is obtained for a disease such as new coronavirus pneumonia. Correspondingly, multiple models can be set for predicting the user's illness based on the medical data set. Other situations are similar and will not be described in detail here.

[0078] It should be noted that the embodiments of the present application can be applied to taxation, medical care, customs, industry and commerce and other government affairs scenarios, and can also be applied to other scenarios that can be identified using machine learning models, such as fiscal government affairs scenarios, predicting the risks of fiscal policies of units at all levels based on fiscal data sets of units at all levels, and land government affairs scenarios, predicting land supply risks based on land-related data sets of units at all levels, etc. The embodiments of the present application are not limited to them.

[0079] A5, the client sends the target dataset identifier BBB and the target model identifier AAA to the interpretation server.

[0080] It can be understood that after the user clicks the above explanation control, the client can send the target data set identifier BBB and the target model identifier AAA to be explained to the explanation server.

[0081] A6, explaining that the server 102 obtains the target data set from the data source.

[0082] like Figure 1 After receiving the target dataset identifier BBB and the target model identifier AAA, the interpretation server can use the target dataset identifier BBB from the tax data server after receiving the target dataset identifier BBB and the target model identifier AAA. Then the interpretation server can call the target model corresponding to the target model identifier AAA to predict the data to be analyzed in the target dataset identifier and obtain the prediction results of each data to be analyzed.

[0083] The target model can be preset on the interpretation server, and the interpretation server can directly call the corresponding target model according to the target model identifier to process the target data set to obtain the processing result data. The target model can also be preset on the data server where the target data set is located, and the interpretation server can obtain the target model while obtaining the target data set from the data server, and then call the target model to process the target data set to obtain the processing result data.

[0084] It should be noted that in the embodiment of the present application, the process of interpreting the processing result data to obtain weight data may include the following steps:

[0085] B1, obtain the explanation dataset corresponding to the target model;

[0086] It should be noted that the explanation dataset can be obtained in a variety of ways, such as the following two ways:

[0087] Method 1:

[0088] 1. Obtain the distribution Dt_statistics of feature values ​​in each feature dimension in the training data set Dt used to train the target model;

[0089] 2. Determine the explanation data set Ds_X; among the N explanation data in the explanation data set, each feature dimension satisfies the distribution Dt_statistics.

[0090] Assume that the target dataset is Dp_X, and the feature dimensions [f1, f2, ..., fn] in Dp_X are the same as the feature dimensions [f1, f2, ..., fn] in the training dataset Dt. The feature dimensions in Ds_X can be identified by words that are easier for users to understand intuitively. For example, for the electrocardiogram feature dimension in Dp_X, the feature dimension is identified by the string XXX in the data source. When determining the interpretation dataset Ds_X, the string XXX is replaced with the word "electrocardiogram".

[0091] Method 2:

[0092] 1. For the target dataset Dp_X, an explanation dataset De_X associated one-to-one with Dp_X can be pre-constructed.

[0093] It should be noted that the data to be analyzed in the target data set can be input into the target model trained with the training data set Dt to obtain processing result data.

[0094] The feature dimensions [fe1, fe2, ..., fem] of De_X may be part or all of the feature dimensions [fe1, fe2, ..., fem] in Dt. Of course, if the user is more concerned about the feature dimensions he needs, [fe1, fe2, ..., fem] may be selected as needed, and the embodiment of the present application does not limit it.

[0095] It should be noted that the identification of the feature dimensions in De_X may also adopt the aforementioned words that are easier for users to understand intuitively.

[0096] B2, for the data to be analyzed in the target data set, after the target model processes the data to be analyzed to obtain prediction result data, determine the distance between each piece of explanation data in the explanation data set and the data to be analyzed, and use the distance as the first weight of the piece of explanation data;

[0097] For example, the data to be analyzed p1 in the aforementioned Dp_X is input into the target model for processing to obtain the processing result data, such as y1. Then for the above two methods, there can be two interpretation processes respectively:

[0098] Based on the interpretation data set obtained in the first method, the process of obtaining the first weight may include:

[0099] The data to be analyzed p1 in Dp_X is used as the target data to be analyzed;

[0100] The distance between each interpretation data ti in Ds_X and p1 is calculated to obtain the first weight, where i=1, 2, ..., n, n is an integer greater than 0, and n is the total number of interpretation data in Ds_X.

[0101] The other data to be analyzed in Dp_X can refer to the calculation process of the above method 1.

[0102] Based on the explanation data set obtained in the second method, the process of obtaining the first weight may include:

[0103] The parsed data t1 corresponding to the data to be analyzed p1 in De_X is used as the target data to be analyzed;

[0104] The processing result data y1 of the data to be analyzed p1 is used as the processing result data corresponding to t1;

[0105] The distance between each interpretation data ti in De_X and t1 is calculated respectively to obtain the first weight.

[0106] For other data to be analyzed in De_X, the calculation process of the above method 2 can be referred to.

[0107] B3, constructing an explanation training data set based on each explanation data, the prediction result data of the data to be analyzed, and the first weight of the explanation data, and training the explanation model according to the explanation training data set.

[0108] As in the process B2 mentioned above, whether it is the calculation process of method one or method two, the first weight corresponding to the n explanatory data of the explanatory data set is obtained. Then, each explanatory data can be multiplied by the first weight, and the processing result corresponding to the explanatory data can be multiplied by the first weight to obtain n weighted explanatory data and the weighted processing result.

[0109] Then, the N feature dimensions of the weighted explanation data are used as the input of the explanation model, the coefficients of each feature dimension are used as the coefficients to be trained for the explanation model, and the weighted processing results are used as the output of the explanation model to train the explanation model.

[0110] After training, the coefficients of the feature dimensions in the explanatory model are the target weights of the corresponding feature dimensions.

[0111] B4, outputs the target weights for at least some of the feature dimensions in the explanation model.

[0112] The above B1-B4 is the overall logic of the interpretation process, and the specific interpretation principles will be introduced in detail in the subsequent sections.

[0113] A7. Explain that the server 102 returns weight data for the processing result data.

[0114] For example, in the aforementioned process, the target data set includes 1,000 pieces of data to be analyzed, each piece of data to be analyzed includes feature data of 100 feature dimensions. After the interpretation server obtains the processing result data of part or all of the 1,000 pieces of data to be analyzed in the above manner, the interpretation server can call the preset parsing model to parse each piece of processing result data, and obtain the weights of all the 100 feature dimensions of each piece of data to be analyzed or the weights of some feature dimensions that are ranked first after the absolute values ​​are sorted from large to small. Then the server can return the weights of the above feature dimensions for these processing result data to the client.

[0115] A8. The client 101 displays weight data of the processing result data in the second interface area.

[0116] After receiving the weights of the feature dimensions of the processing result data sent by the interpretation server 102, the client 101 may display the weights of the feature dimensions of the processing result data in the second interface area 106 of the information display page 105. Figure 1 For the target data set BBB, the result after data 1 is processed by the target model AAA is 1. The weight data after interpretation may include: feature 1: weight x, feature 2: weight y, etc. The processing result data and weight data of other data to be analyzed may also be displayed.

[0117] It should be noted that in the above A1, the client can also follow the preset application, in which the information display page is preset, and the user can directly enter the target data set and the target model in the first interface area of ​​the information display page, and the client receives the user's input operation on the target data set and the target model through the first interface area. In this case, the client does not need to request the information display page from the interpretation server, and the server does not need to display the information display page according to the request. It can be understood that the specific way in which the client displays the information display page is not limited by the embodiment of the present application.

[0118] It should be noted that, in the embodiment of the present application, the above-mentioned first interface area 106 and the second interface area 107 may be in the same focus interface or may not be in the same focus interface. The focus interface can be understood as the interface currently being displayed on the display screen. When the first interface area 106 and the second interface area 107 are not in the same focus interface, the focus interface of the first interface area 106 can be displayed first, and after the user inputs the above-mentioned target data set and target model and other contents, after the user clicks the explanation control, it is switched to the focus interface of the second interface area 107. When the user wants to re-enter other target data sets and target models and other contents, return to the focus interface of the first interface area 106.

[0119] It should be noted that if the server does not query the target data set input by the user, or the server does not query the target model input by the user, it is impossible to execute the process of obtaining the processing results by processing the data to be analyzed according to the target model. The user can be prompted that the input is wrong in the above information display page, and the user can also be prompted to re-enter the information.

[0120] Therefore, through the above method, in the embodiment of the present application, the user can select the target data set and target model according to their needs, and then can intuitively see the weights of some or all of the feature dimensions, so as to provide more information for user decision-making. For example, for tax audits, users can intuitively see which behavioral characteristics in high-risk enterprises cause the high weight of the high risk. For example, for auxiliary diagnosis, doctors can intuitively see which patient characteristics are likely to cause the patient to suffer from a certain disease when the probability is high.

[0121] Reference Figure 2 , which shows a data display method of an embodiment of the present application, which may include the following steps:

[0122] Step 201: receiving an input operation on a target data set and a target model on a display interface, wherein the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension;

[0123] In the embodiment of the present application, as described above, there may be multiple data sources, and each data source may have multiple data tables.

[0124] Then, in the embodiment of the present application, if there is a data source and there are multiple data tables in the data source, the target data set can be a certain data table.

[0125] If there are multiple data sources and each data source has a data table, the target data set can be the data table of the data source.

[0126] If there are multiple data sources and each data source has multiple data tables, the target data set can be a data table in a data source.

[0127] It should be noted that in the embodiment of the present application, the machine learning model can be trained in advance with corresponding training data for each of the aforementioned data tables. A machine learning model can also be trained for the same type of data tables. For example, if the tax data of various regions are stored in databases of different regions, and the characteristic dimensions of the tax data of various regions are the same, a machine learning model can be trained, and then the machine learning model can be used to identify the data tables in the databases of different regions, respectively, to reduce the number of machine learning models and save system resources.

[0128] The target data set may include at least one piece of data to be analyzed, and one piece of data to be analyzed may include at least one characteristic dimension. For example, in a tax audit, the tax data of an enterprise may be a piece of data to be analyzed, and the tax data may include characteristic dimensions of various tax behaviors. For a patient, the relevant information of the patient may be a piece of data to be analyzed, and the relevant information may include characteristic dimensions such as various detection indicators.

[0129] It can be understood that in one embodiment of the present application, a Figure 1 The information display page 105 may be the above-mentioned display interface.

[0130] It can be understood that the information display page 105 displays the first interface area 106, and the user can input the target data set and the target model in the first interface area 106. Of course, the above input operation can also be received at other locations of the information display page 105, and the embodiment of the present application does not limit it.

[0131] In actual applications, what the user inputs in the above display interface is the target data set identifier and the target model identifier. The device that performs subsequent predictions and interpretations can obtain the corresponding target data set based on the target data set identifier, and can also obtain the corresponding target model based on the target model identifier. Then input the target data set into the target model to obtain the processing result data. For example, if the target model is a model for predicting the tax risk of an enterprise, and the data to be analyzed in the target data set is the tax data of the enterprise, then the data to be analyzed can be input into the target model to obtain the tax risk level of the enterprise, such as high risk, low risk, etc. For another example, if the target model is to predict the probability that a person has new coronary pneumonia, and the data to be analyzed in the target data set is the test data of the person, etc., then the data to be analyzed can be input into the target model to obtain the probability that the person has new coronary pneumonia.

[0132] In the embodiment of the present application, after receiving the input operation, the user does not need to perform other operations, and part or all of the target data set can be directly input into the target model to obtain the processing result data. The processing process can also be executed after the user triggers a certain control. For example, a prediction control is also provided in the information display page. After the user clicks the prediction control, part or all of the target data set can be input into the target model to obtain the processing result.

[0133] After obtaining part or all of the processing result data of the data to be analyzed, the processing result data can be interpreted to obtain weight data, which can include the target weight of at least one of the aforementioned feature dimensions for the processing result data, and the target weight can be understood as a contribution value.

[0134] It can be understood that the device receiving the user's input operation can be the aforementioned client. If the client has sufficient computing power, the client can perform the aforementioned process of inputting part or all of the target data set into the target model to obtain the processing result data, and interpreting the processing result data to obtain the weight data. If the client has insufficient computing power, such as Figure 1 , the client can send the target data set and target model to the interpretation server, which performs the above two processes and then sends the processing result data and weight data to the client for display.

[0135] Step 202, displaying weight data of the processing result data in the display interface; wherein the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions to the processing result data.

[0136] It can be understood that in the embodiment of the present application, after the aforementioned processing result data and weight data are obtained, they can be displayed in the display interface.

[0137] Optionally, an embodiment of the present application includes the above-mentioned display interface including a first interface area and a second interface area. The above-mentioned step 201 may include: receiving input operations for the target data set and the target model in the first interface area. The above-mentioned step 202 may include: displaying weight data for the processing result data in the second interface area. In this way, the area where the input operations of the target data set and the target model are located is separated from the data display area, and the interface is more concise and intuitive, which is convenient for users to operate and view data.

[0138] like Figure 1The information display page 105 may further include a second interface display area 107. After obtaining the weight data of the processing result data, the weight data of the processing result data may be displayed in the second interface area.

[0139] It should be noted that when displaying the weight data of the processing result data in the display interface, the processing result data and the weight data can be displayed together. The processing result data and the weight data can also be displayed together in the second interface display area 107 in the display interface.

[0140] The weight data may include a target weight of at least one characteristic dimension. For example, for the aforementioned tax audit, the risk level of the enterprise and the target weights of some or all characteristic dimensions may be displayed.

[0141] Of course, in the case where the size of the display interface for displaying data is fixed, if the display interface cannot display multiple rows of data, for example, the size of the second interface area is fixed and cannot display the above-mentioned processing result data and weight data, the first slider can be set along the row arrangement direction to display the row data by operating the slider, or paging can be used to display the data, with a maximum of a preset number of rows per page, and the row data can be displayed by switching the paging. In addition, if the second interface area cannot display multiple columns of data, a second slider can be set along the column arrangement direction to display the column data by operating the second slider.

[0142] Optionally, in an embodiment of the present application, before step 202, the method further includes:

[0143] Step S11, receiving an input operation on an explanation data set in the display interface; the explanation data set includes explanation data corresponding to the data to be analyzed respectively; wherein the explanation data includes feature data of at least one target feature dimension among at least one feature dimension of the data to be analyzed; wherein, for the same feature dimension, a first feature dimension identifier of the explanation data is different from a second feature dimension identifier of the corresponding data to be analyzed; wherein an association relationship exists between at least one of the first feature dimension identifiers and a second feature dimension identifier belonging to the same feature dimension.

[0144] It can be understood that in the embodiment of the present application, an input operation for interpreting the data set can also be received in the first interface area.

[0145] It should be noted that if the aforementioned first interface area is provided in the display interface, in the embodiment of the present application, an input operation for interpreting the data set can also be received in the first interface area.

[0146] The explanatory data includes feature data of at least one target feature dimension in at least one feature dimension of the data to be analyzed. For example, if there are 1,000 pieces of data to be analyzed in the target data set, then there are 1,000 pieces of explanatory data corresponding to the data to be analyzed in the explanatory data set. For example, if there is one piece of data to be analyzed of user A in the target data set, then there is one piece of explanatory data of user A in the explanatory data set. In addition, for example, if there are data of 100 feature dimensions for the piece of data to be analyzed of user A, then there can be data of at least one of the 100 feature dimensions for the piece of explanatory data of user A, for example, there are 20 feature dimensions and their data among the 100 feature dimensions for the piece of explanatory data of user A.

[0147] Among them, for the same feature dimension, the first feature dimension identifier of the explained data is different from the second feature dimension identifier of the corresponding data to be analyzed. For example, in the target data set, for the electrocardiogram feature dimension, the identifier of the feature dimension is the string XXX, and in the determined explained data set, the identifier of the feature dimension is the word "electrocardiogram". The identifier of the feature dimension in the explained data set is more easy to understand.

[0148] In actual applications, due to historical issues of the data source, the naming of the feature dimension may not be easy to understand through the identifier, so this application changes the naming of the feature dimension in the interpretation data set to a word that is easier for users to understand its meaning.

[0149] There is an association relationship between at least one of the first feature dimension identifiers and a second feature dimension identifier belonging to the same feature dimension.

[0150] This association relationship allows the user to match the processing result data of a piece of data to be analyzed with the corresponding interpretation data. For example, if the user ID is used as the association primary key, then after the data to be analyzed of user A is input into the target model and the processing result data is obtained, the processing result data can be matched with the interpretation data of user A according to user A to avoid the inability to perform the interpretation process.

[0151] This solution of step S11 can solve the problem of inconsistency between the characteristic dimension identification of the data for training the target model and the characteristic dimension identification of the explanatory data required by the user, so that the user can set the identification of the characteristic dimension of the explanatory data according to his own needs.

[0152] Optionally, in an embodiment of the present application, the first interface area includes: an explanation data source selection control, an explanation data table selection control, and a feature dimension association control; the step S11 includes:

[0153] Step S111, receiving an interpretation data source identifier in the interpretation data source selection control; the interpretation data source identifier is used to access an interpretation data source corresponding to the interpretation data source identifier;

[0154] Step S111, receiving a selection operation of an explanation data table identifier in the explanation data table selection control; the explanation data table identifier is used to obtain an explanation data set corresponding to the explanation data table identifier from an explanation data source;

[0155] Step S113, receiving, in the explained data table selection control, a setting operation for associating the first characteristic dimension identifier in the explained data set with the second characteristic dimension identifier in the target data set.

[0156] like Figure 3 , which shows a schematic diagram of a display interface of an embodiment of the present application. Figure 3 In the display interface 110 , a first interface area A11 and a second interface area A12 are included.

[0157] Figure 3 In the example, "Select Explanation Data" may be included for selecting an explanation data set. In the embodiment of the present application, for the explanation data, there may also be multiple data sources storing the explanation data and multiple data tables. Therefore, in the embodiment of the present application, "Select Explanation Data" may include two controls: an explanation data source selection control 114 for selecting a data source for the explanation data, and an explanation data table selection control 115 for selecting a data table for the explanation data.

[0158] Since the user additionally selects an explanation data table, the identifier of the characteristic dimension of the explanation data table may be different from the identifier of the characteristic dimension of the target data table where the data to be analyzed is located. In order to make the processing result data of each data in the target data table correspond to the data in the explanation data table, the embodiment of the present application further sets a characteristic dimension association control "Select Association" in the first interface area A11, such as Figure 3 In order to facilitate the user to associate the key fields of the target data table with the key fields of the explanation data table, a control 116 for selecting the primary key of the prediction data table and a control 117 for selecting the primary key of the explanation data table are provided. It should be noted that the control 116 can use the identifier of the characteristic dimension that uniquely identifies a piece of data to be analyzed as a selection item, and the control 117 can also use the identifier of the characteristic dimension that uniquely identifies a piece of explanation data as a selection item, so that one piece of data to be analyzed will not be associated with multiple pieces of explanation data, nor will one piece of explanation data be associated with multiple pieces of data to be analyzed.

[0159] Then, the user can select the explained data source in control 114 and the explained data table in control 115. Then, after the user selects the target data table in control 113, the primary key of the target data table can be selected in 116 and the primary key of the explained data table can be selected in 117 so that the device can associate the primary keys of the two tables.

[0160] It should be noted that the display interface of the embodiment of the present application may not be divided into the first interface area A11 and the second interface area A12. It is sufficient to provide a control for receiving input operations for interpreting data sources in the display interface, and the embodiment of the present application does not limit this.

[0161] Optionally, in an embodiment of the present application, the display interface includes a model identification input control, a prediction data source identification input control, and a prediction data table selection control; the step 201 includes:

[0162] Step S21, receiving a target model identifier input in the model identifier input control; the target model identifier is used to call the target model;

[0163] Step S22, receiving a predicted data source identifier input into the data source identifier input control; the predicted data source identifier is used to access the predicted data source;

[0164] Step S23, receiving a selection operation on a target data table identifier in the prediction data table selection control; the target data table identifier is used to obtain the target data set from the prediction data source.

[0165] It can be understood that the model identification input control, the prediction data source identification input control, and the prediction data table selection control can be set in the above-mentioned first interface area A11.

[0166] like Figure 3 and Figure 4 In the embodiment of the present application, in the first interface area A11 of the information display page 110, there is a model identification input control "model identification" 111 for inputting the target model, and there is a control "select predicted data" for inputting the target data set.

[0167] like Figure 3 and Figure 4For the data to be analyzed, when there are multiple data sources and multiple data tables, "Select predicted data" includes two controls: a predicted data source identifier input control 112 for selecting the data source where the data to be analyzed is located, and a predicted data table selection control 113 for selecting the data table where the data to be analyzed is located. It should be noted that in an embodiment of the present application, the available data sources and data tables can be pre-set, and then the data source and data table can be selected by using drop-down menus such as 112 and 113. It can be understood that control 113 can be dependent on 112. After the user selects a data source in 112, the embodiment of the present application will provide the data table corresponding to the data source to the drop-down menu 113 for the user to select, thereby reducing the user's selection range and improving selection efficiency.

[0168] Then the user can select the target data source where the data to be analyzed is located through the control 112 , and select the target data table where the data to be analyzed is located through the control 113 , and the target data table is used as the target data set.

[0169] Optionally, in an embodiment of the present application, before step 102, the method further includes:

[0170] Step S31, providing a prediction control; the prediction control is used to display the processing result data in the display interface when a trigger operation on the prediction control is received.

[0171] It can be understood that in the embodiment of the present application, a prediction control can be displayed in the display interface for user operation, and after the user triggers the prediction control, the processing result can be displayed in the display interface.

[0172] like Figure 3 and Figure 4 In the embodiment of the present application, a prediction control 119 may also be provided in the first interface area A11. After the user enters the model identifier aaa in 111, selects library 1 in 112, and selects table 1 in 113, if the user clicks the prediction control 119, Figure 5 , the processing result data is displayed row by row in the second interface area. Figure 5 In the figure, the prediction result of each piece of data to be analyzed is shown, including the prediction probability, feature dimensions XXX, YYY, ZZZ, etc. It can be understood that the prediction result can be a processing result obtained by processing the data to be analyzed by the target model.

[0173] It should be noted that Figures 3 to 8 The naming of the controls in is only an example and is not limited to the embodiments of the present application.

[0174] Optionally, in an embodiment of the present application, before step 102, the method further includes:

[0175] Step S32, providing an explanation object selection control for each processing result data in the display interface; the explanation object selection control is used to determine the processing result data to be explained.

[0176] In the embodiment of the present application, an explanation object selection control may be provided in the display interface, and a processing result data may correspond to an explanation object selection control one by one. The user may decide to select or not select the processing result data by operating the explanation object selection control. Of course, a select all control may also be provided, and after the user triggers the select all control, all the processing result data may be selected.

[0177] It should be noted that, when the display interface includes the aforementioned second interface area A12, an interpretation object selection control may be provided in the second interface area A12.

[0178] like Figure 5 In the embodiment of the present application, an explanation object selection control 123 may be provided for each row of data in the second interface area A12. The user may select the processing result data that he wants to explain through the explanation object selection control 123.

[0179] Optionally, in an embodiment of the present application, before step 102, the method further includes:

[0180] Step S33, providing an explanation control; the explanation control is used to obtain weight data of the processing result data when a trigger operation on the explanation control is received.

[0181] In the embodiment of the present application, an explanation control may be added to the display interface, and the explanation control may start the explanation process of the processing result data to obtain weight data of the processing result data.

[0182] like Figure 5 In the embodiment of the present application, an explanation control 120 may also be provided in the first interface area A11. It is understood that if the user Figure 5 In the example, the two data of sequence 1 and 2 are selected, and the explanation control 120 is clicked to explain the three data and obtain the weight data. Figure 5 , an “*” is added to the box of the selected data to indicate that it has been selected. Of course, other symbols can also be used to indicate that it has been selected.

[0183] Optionally, in one embodiment of the present application, the display interface includes: a fixed column and a sliding column, the fixed column includes: a button column for displaying an interpretation object selection control, an interpretation status column for displaying whether it has been interpreted, a result display column for displaying a processing result, and a probability display column for displaying at least one of the predicted probability of the processing result; the sliding column includes: a feature data column corresponding to at least one of the feature dimensions.

[0184] In the embodiment of the present application, in order to conveniently display the above data, such as displaying the above processing result data or the above weight data, two types of columns are set, namely fixed columns and slideable columns. The fixed column includes: a button column for displaying an interpretation object selection control, an interpretation status column for displaying whether the interpretation has been completed, a result display column for displaying the processing result, and a probability display column for displaying the predicted probability of the processing result; the slideable column includes: at least one feature data column corresponding to the feature dimension.

[0185] Optionally, when the display interface includes the aforementioned second interface area A12, the fixed column and the slidable column can be set in the second interface area A12.

[0186] like Figure 5 In an embodiment of the present application, when displaying in the second interface area A2, multiple columns can be set. The above columns can be divided into fixed columns and movable columns. The fixed columns are fixed and cannot be moved. The movable columns can swap the positions of the columns between movable columns. The fixed columns may include: a button column for displaying an interpretation object selection control, an interpretation status column for displaying whether the interpretation has been completed, a result display column for displaying a processing result, and a probability display column for displaying at least one of the predicted probabilities of the processing results. Among them, the button column for displaying the interpretation object selection control is as follows: Figure 5 The column where the box is located is used to show whether the explanation has been completed. For example, Figure 5 The column where "Explanation" is located is used to display the result display column of the processing result. Figure 5 The "Prediction Result" column is used to display the predicted probability of the processing result. Figure 5 The column where "predicted probability" is located. The movable column is as follows Figure 5 The columns where the feature dimensions XXX, YYY, and ZZZ are located. Any two feature dimension columns can be swapped, for example, feature dimension XXX can be swapped with feature dimension ZZZ. A feature dimension column can also be inserted between any two other adjacent columns.

[0187] Optionally, in an embodiment of the present application, after step S33, the method further includes:

[0188] Step S34, when the processing result data does not have weight data, setting the interpretation status corresponding to the processing result data in the interpretation status column to an uninterpreted mark;

[0189] Step S35, when the processing result data has weight data, setting the interpretation status corresponding to the processing result data in the interpretation status column to an interpreted mark.

[0190] for example Figure 5 In the example, the user selects two rows of data, 1 and 2, through the control 123. After clicking the explanation control 120, the two rows of data 1 and 2 are explained. After the explanation is completed, Figure 6 , the "Explanation" column of the two rows of data 1 and 2 is marked with an explained mark "!" to indicate that the explanation has been given.

[0191] like Figure 6 , the uninterpreted third row of data is left blank, which can be understood as an uninterpreted mark. Of course, other visual marks can also be used as uninterpreted marks, such as "x" as an uninterpreted mark.

[0192] Optionally, in an embodiment of the present application, after step S35, the method further includes:

[0193] Step S36, when the processing result data has weight data, setting the selection control corresponding to the processing result data in the option button column to be unselectable.

[0194] for example Figure 6 If the two rows of data in sequence 1 and 2 have been explained, there is no need to select the two rows for repeated explanation, and the selection control corresponding to the processing result data in the selection button column can be set to be unselectable. Then, when the user selects the unexplained row and clicks the explanation control again, only the newly selected row can be explained.

[0195] Optionally, in an embodiment of the present application, step 202 further includes:

[0196] Step S41, displaying weight data of the processing result data in a list in the display interface;

[0197] Or in step S41, weight data of the processing result data is displayed graphically in the display interface.

[0198] In the embodiments of the present application, the weight data of the processing result data can be displayed in a variety of ways, such as in a list or in a graphical form. In the case where the weight has positive and negative values, a graphical comparison display can be used, which is more intuitive. Among them, a positive value can indicate a positive correlation with the processing result. The larger the positive value, the stronger the positive impact on the processing result. Among them, a negative value can indicate a negative correlation with the processing result. The larger the negative value, the stronger the negative impact on the processing result. Figure 7 The weight data of this row of data in sequence 1 can be displayed in a graphical way. Figure 7 In the figure, the left side of the center line is the feature dimension that has a negative impact on the prediction result, and the right side of the center line is the feature dimension that has a positive impact on the prediction result. The length of the line corresponding to the feature dimension indicates its importance or weight.

[0199] like Figure 8 , the weight data of this row of data in sequence 1 can be displayed in a table. Figure 8 In the table, feature dimensions that have a negative impact on the prediction results are placed in one table, and feature dimensions that have a positive impact on the prediction results are placed in another table. Write the value and target weight after the identifier of the feature dimension.

[0200] from Figure 7 , Figure 8 It can be seen that the labels of feature dimensions use words that are easy for users to understand, and the corresponding values ​​and weights are displayed corresponding to the corresponding feature dimension labels, providing users with more intuitive reference information.

[0201] Optionally, lists and graphs can be displayed based on the user's choice. Figure 7 and Figure 8 In the example, there are two labels, "icon" and "list". When the user clicks the "icon" label, Figure 7 The weight data of the processing result data is displayed in a graphic. When the user clicks "List", Figure 8 The weight data of the processing result data is displayed in a list.

[0202] Optionally, in an embodiment of the present application, step 202 includes:

[0203] Step S43, when the processing result data is displayed in the display area and weight data already exists in the processing result data, if an expansion operation of the processing result data is received, the weight data is displayed in the display area at a position corresponding to the processing result data.

[0204] It is understandable that after explaining the processing result data, the weight data of the processing result data may not be displayed first, because if the weight data of each processing result data is displayed, the interface may be more disordered, and the interval between the weight data of multiple processing result data is too long, and the user needs to drag more frequently to view the weight data they want to see. Therefore, on the basis of displaying the processing result data first, after receiving the user's expansion operation on a certain processing result data, the weight data of the processing result data is expanded, and the weight data required by the user can be displayed in a targeted manner. The interface is more concise and more convenient for users to browse.

[0205] like Figure 6 After interpreting the processing result data, you can add an interpreted status mark to the interpreted processing result data instead of expanding the weight data. Figure 6 The sequence 1 and 2 are added with the interpreted status "!". Then when the user clicks on a row, the weight data corresponding to the row is displayed. Figure 7 or Figure 8 .

[0206] It should be noted that if Figure 3 , Figure 4 , the first interface area may also include a processing result data quantity selection control "Return result filter" control 118, the user can slide 118 to set the number of processing results displayed after the target data set is processed. In actual applications, it can be displayed from high to low according to the category that the user is concerned about, according to the probability of belonging to the category. For example, for patients with new coronary pneumonia, if the user is concerned about the result of new coronary pneumonia, the results identified as new coronary pneumonia will be displayed in order from high to low by probability. For other display methods, the embodiments of the present application do not limit them.

[0207] It should be noted that controls such as a save control for saving the interpretation result to a set storage area and a return control for exiting the information display page may also be provided in the display interface.

[0208] For example Figure 3 , Figure 4 A save control 121 is also provided in the first interface area A11. After the user clicks the save control 121, the interpretation result is saved to the set storage area.

[0209] For example Figure 3 , Figure 4 A return control 122 is also provided in the first interface area A11. After the user clicks the return control 122, the user can exit the information display page.

[0210] For example Figure 3 , Figure 4, the first interface area A11 also provides a non-mandatory item collapse control 124. In the embodiment of the present application, the model identifier input control, the prediction data source identifier input control, and the prediction data table selection control can be set as mandatory items. The explanation data source selection control, the explanation data table selection control, and the feature dimension association control can be set as non-mandatory items. Then the user Figure 4 After clicking control 124, enter Figure 3 , showing the aforementioned explanation data source selection control, explanation data table selection control, and feature dimension association control. Figure 3 After clicking control 124, enter Figure 4 , and the aforementioned explanation data source selection control, explanation data table selection control, and characteristic dimension association control are folded. In this way, the display area of ​​the second interface display area A12 can be expanded without the explanation data source selection control, explanation data table selection control, and characteristic dimension association control.

[0211] The following describes the process of interpreting the processing result data.

[0212] Optionally, in an embodiment of the present application, before step 202, the method further includes:

[0213] Step C11, taking each piece of interpretation data in the interpretation data set as target data to be analyzed;

[0214] For example, the target model M is trained using the training data set Dt, and the target data set to be predicted is Dp_X, and the feature dimension of Dp_X is the same as the feature dimension of Dt.

[0215] An explanation data set De_X is set in advance for Dp_X, and each row of data in De_X is associated with each row of data in Dp_X one by one. It can be understood that Dp_X and De_X are associated with the row of data for the same object. For example, a row of tax data for enterprise A in Dp_X is associated with a row of tax data for enterprise A in De_X.

[0216] In the embodiment of the present application, Dp_X may be first input into the target model M to obtain the processing result data Dp_Y, for example, obtaining the processing result data yi of each row of data xi in Dp_X.

[0217] Among them, the feature dimensions [fe1, fe2, …, fem] of De_X are part or all of the feature dimensions [f1, f2, …, fn] of Dp_X.

[0218] In an embodiment of the present invention, in order to explain the scenario where the feature dimension identifier of Dt and the feature dimension identifier of the explanatory data required by the user are inconsistent when training the target model M, or the number of feature dimensions required by the user is inconsistent with the number of feature dimensions of Dt, the processing result can also be explained. The embodiment of the present application additionally sets the above-mentioned feature dimension identifier or an explanatory data set De_X including some feature dimensions and data in Dp_X. Of course, the number of entries in the explanatory data set De_X is the same as the number of entries in Dp_X.

[0219] In practical applications, in order to obtain more accurate weight data, the amount of data in Dp_X should be at least greater than 100.

[0220] It can be understood that in the embodiment of the present application, each piece of explanation data in the explanation data set is taken as target data to be analyzed, that is, the feature dimension [fe1, fe2, ..., fem] in De_X is taken as the explanation object.

[0221] Step C12, according to the association relationship between the first feature dimension identifier and the second feature dimension identifier belonging to the same feature dimension, using the processing result of the corresponding data to be analyzed in the target data set as the processing result of the explained data;

[0222] For example, when associating, the user ID feature dimension is used to associate Dp_X with De_X. Then, since , the processing result y of each row of data x in Dp_X can be associated with De_X, and the y can be corresponded to the row of De_X with the same user ID.

[0223] Then even if De_X is preprocessed as described above, each row of data of the preprocessed De_X" can correspond to the corresponding y.

[0224] Step C13, respectively determining the distance between the explained data in the explained data set and the target data to be analyzed, and obtaining a first weight of the explained data according to the distance;

[0225] In the embodiment of the present invention, for a piece of data x in De_X, the distances between each piece of data in De_X and the piece of data x may be calculated respectively, and then the first weight of each piece of data in De_X may be obtained.

[0226] Optionally, in order to facilitate calculation, in the embodiment of the present application, De_X may also be preprocessed, and the preprocessing process may include:

[0227] 1. The numerical feature dimensions in De_X are divided according to a preset numerical range, and each value in each numerical range is changed to a new value according to a uniform distribution or a Gaussian distribution.

[0228] Among them, the numerical range can be understood as a bucket. Each feature dimension can be divided into multiple buckets according to needs. The specific range of each bucket can be set according to actual needs, and the embodiment of the present application is not limited to this.

[0229] Among them, we can first construct a uniform distribution U(E, σ) based on the mean value E and variance σ, where the mean value of the uniform distribution is E and the variance is σ. Then, we use the uniform distribution U(E, σ) to randomly generate k values ​​in the range (a, b), and the k and the eigenvalues ​​in the explanatory data set fall within the range of the explanatory data.

[0230] In addition, we can first construct a Gaussian distribution N(E, σ) based on the mean value E and variance σ, where the mean value of the Gaussian distribution is E and the variance is σ. Then, we randomly generate k values ​​in the range (a, b) using the Gaussian distribution N(E, σ), and the k and the eigenvalues ​​in the explanatory data set fall within the range to explain the number of data items.

[0231] For example, for the age feature dimension, it can be divided into several age groups, 0-20 years old, 20-40 years old, 40-60 years old, 60-80 years old, and over 80 years old. Then, for the values ​​in each age group, after obtaining the maximum value b, minimum value a, average value E and variance σ, they are changed to new values ​​using uniform distribution or Gaussian distribution.

[0232] For example, for 0 to 20 years old, there are 20 data of ages that fall into this age group, and then this age group has 20 age values. Then get the maximum value, minimum value, average value and variance of these 20 ages. According to the maximum value a, minimum value b, average value E and variance σ, randomly select 20 age values ​​with uniform distribution or Gaussian distribution, replace the original 20 age values ​​and write them into the age feature dimension of the 20 explanatory data.

[0233] For the category row feature dimension, it can remain unchanged.

[0234] After De_X is preprocessed as above, an explanation data set De_X' is obtained. Each explanation data in De_X' is also used as data to be analyzed.

[0235] Then, for each piece of data x' to be analyzed in De_X', the distances between each piece of data in De_X' and the piece of data x' are calculated respectively, and then the first weight of each piece of data in De_X' is obtained.

[0236] In practical applications, the value of each piece of data under each characteristic dimension of De_X' can be compared with the value of the characteristic dimension in x'; if they are the same, the value is changed to 1, and if they are different, the value is changed to 0. In this way, De_X' is processed into a binary array De_X", and x' is processed as x" = [1,1,...1].

[0237] Then calculate the Euclidean distance between each row of data in De_X" and x", and get d = [d1, d2, ..., dN]. Where N is the number of rows in De_X".

[0238] Then according to d=[d 1 , d 2 , …, dn], and the first weight of each row of data is obtained by formula (1).

[0239]

[0240] Among them sw i is the weight of the i-th row, d i is the aforementioned Euclidean distance, and n is the total number of rows of De_X” data.

[0241] In this way, we get the weight sw = [sw 1 , sw 2 ,…,sw n ],

[0242] It should be noted that the positions of the above 0 and 1 can also be interchanged, and the embodiments of the present application are not limited thereto.

[0243] Step C14, training an explanation model according to the explanation data, the processing result of the explanation data, and the first weight of the explanation data;

[0244] It can be understood that through the aforementioned steps C11-C14, the preprocessed explanation data De_X", the corresponding processing result data Dp_Y, and the first weight sw of n rows of explanation data are obtained, and the explanation model can be trained based on the three data.

[0245] For example, the explanation data De_X" has 100 rows, and each row of data has a processing result data yi, and each row of data has a first weight sw i , then we can first construct an explanation training data set. For example, multiply each value in the explanation data [x11,x12,…,x1m] of a row by the weight sw of the row 1 , multiply the processing result data y1 corresponding to the row by the weight sw 1 , and then get a training data {[x11*sw 1 ,x12*sw 1 ,…,x1m*sw1 ],y1*sw 1}, so n training data can be obtained.

[0246] Then, the n training data are used to train the explanation model built based on m feature dimensions [fe1, fe2,…, fem]. After the model training is completed, the parameter corresponding to each feature dimension is the weight of the corresponding feature dimension.

[0247] Of course, we can also directly select a row of data x from De_X as the target data to be analyzed, and then calculate the distance d between each row of data and x, and then calculate sw according to the above formula (1), and then build a training set based on each row of explained data xi, yi corresponding to the explained data xi, and swi of the explained data xi to train the explanation model.

[0248] Step C15, obtaining the target weight of each target feature dimension according to the explained model.

[0249] In the embodiment of the present invention, part or all of the feature dimensions and their weights may be output as required.

[0250] It should be noted that, in the above steps C15 to C16, when training the target model M, the feature dimension identifiers of Dt and the feature dimension identifiers of the explanatory data required by the user are inconsistent, or the number of feature dimensions required by the user is inconsistent with the number of feature dimensions of Dt. Through the above process, the processing results can also be explained.

[0251] Optionally, in an embodiment of the present application, before step 202, the method further includes:

[0252] Step S51, obtaining the distribution of feature data of each feature dimension in the training data set used to train the target model;

[0253] In an embodiment of the present application, without presetting an explanation dataset for the target dataset, a dataset can be constructed based on the distribution of data of the feature dimensions of the training data used to train the target model.

[0254] For example, the target model M is trained using the training data set Dt, and the target data set to be predicted is Dp_X, and the feature dimension of Dp_X is the same as the feature dimension of Dt.

[0255] Then for Dt, the data under each feature dimension is distributed. For example, Dt has 1000 data, and assuming there is a feature dimension XXX, then the 1000 data have 1000 data under the feature dimension XXX, then the 1000 data have a distribution.

[0256] Among them, the distribution of feature data of each feature dimension in the above-mentioned training data set can be predetermined by the data source and then stored in the corresponding target model according to the data source. The embodiment of the present application obtains the above-mentioned distribution corresponding to the target model in the data source. The distribution of feature data of each feature dimension in the above-mentioned training data set can be predetermined by the data source and then stored in the above-mentioned interpretation server. When used, the distribution is obtained locally from the interpretation server according to the target model. Of course, other methods can also be used to obtain the distribution, and the embodiments of the present application are not limited thereto.

[0257] Step S52, determining an explanation data set according to the distribution; the distribution of feature data of each feature dimension of the explanation data set is the same as the distribution of feature data of each feature dimension in the training data set.

[0258] In the embodiment of the present application, the number of entries q in the interpretation data set Ds_X can be determined according to the number of interpretation data q required for interpretation accuracy. In this way, the present application generates q interpretation data according to the distribution of the feature data of the aforementioned feature dimensions. The number of feature dimensions in Ds_X can be the same as that of the target data set Dp_X, and the values ​​of each feature dimension in Ds_X meet the aforementioned feature dimension.

[0259] Optionally, in an embodiment of the present application, step S51 includes:

[0260] Step S511: for a numerical feature dimension, according to M preset numerical ranges, respectively count the first quantity of values ​​in each numerical range to obtain a first quantity distribution under the numerical feature dimension;

[0261] Optionally, the embodiment of the present application may determine the distribution of data of the feature dimension according to the type of the feature dimension.

[0262] For example, for a numerical feature dimension, the number of values ​​falling into the pre-divided numerical range can be counted. Then the number of each numerical range constitutes the distribution of the numerical range. For example, the age feature dimension can be divided according to age groups, and the numerical range can be divided into several age groups of 0-20 years old, 20-40 years old, 40-60 years old, 60-80 years old, and over 80 years old. Then count the number of each data in Dt that falls into the above age groups. Assuming that Dt has 1,000 data as mentioned above, for the age feature dimension, according to the above age group division method, for example, 100 for 0-20 years old, 200 for 20-40 years old, 200 for 40-60 years old, 300 for 60-80 years old, and 200 for over 80 years old, then the distribution of the feature dimension of this age group is 0-20 years old: 20-40 years old: 40-60 years old: 60-80 years old: over 80 years old = 1:2:2:3:2.

[0263] Optionally, the maximum, minimum, average and variance of the data within each numerical range can also be counted.

[0264] The step S52 comprises:

[0265] Step S521: Generate N numerical values ​​for the numerical feature dimension; wherein the number of first numerical values ​​within each of the preset numerical ranges satisfies the first numerical distribution.

[0266] In practical applications, when the distribution of data of the feature dimension is determined according to the type of the feature dimension, the generation of the value of the feature dimension of the interpretation data can also be performed according to the corresponding type.

[0267] For this numerical feature dimension, among the q data generated, the values ​​under this numerical feature dimension satisfy the distribution of this numerical feature dimension. Assume that for the accuracy of the explanation, q is set to 500. For the aforementioned age feature dimension, according to its distribution, 50 age values ​​can be randomly generated from 0 to 20 years old, 100 age values ​​from 20 to 40 years old, 100 age values ​​from 40 to 60 years old, 150 age values ​​from 60 to 80 years old, and 100 age values ​​over 80 years old. .

[0268] Optionally, in an embodiment of the present application, step S51 includes:

[0269] The step S512, for the categorical feature dimension, under the feature dimension, counts the second quantity with the same category value to obtain a second quantity distribution under the categorical feature dimension;

[0270] For example, for a categorical feature dimension, the distribution of the feature dimension can be obtained by counting the number of values ​​of the same category. For example, for the feature dimension of the electrocardiogram category, the category values ​​are normal waveform and abnormal waveform. Assuming that Dt has 1,000 data as mentioned above, its distribution is, for example, 300 normal waveforms and 700 abnormal waveforms, then the distribution of the feature dimension is normal waveform: abnormal waveform = 3:7.

[0271] The step S52 comprises:

[0272] Step S522: Generate N category values ​​for the categorical feature dimension; wherein the N category values ​​satisfy the second quantity distribution.

[0273] For example, for the categorical feature dimension, among the q pieces of data generated, the values ​​under the categorical feature dimension satisfy the distribution of the categorical feature dimension. Assume that for the accuracy of the explanation, q is set to 500. For the feature dimension of the electrocardiogram category, according to its distribution, 150 values ​​of normal waveform are set, and 350 values ​​of abnormal waveform are set.

[0274] Then, through the above method, the distribution of data of each feature dimension of the explained data set Ds_X is the same as the distribution of the feature dimension of Dt.

[0275] Optionally, in an embodiment of the present application, after step S52 or after step S11, the method further includes:

[0276] Step S61, processing the explained data set according to the target model to obtain processing result data corresponding to the explained data set;

[0277] For example, for the aforementioned target data set Dp_X, the embodiment of the present application can use the aforementioned Ds_X to interpret the data to be analyzed in the Dp_X.

[0278] In the embodiment of the present application, Ds_X may be first input into the target model M to obtain the processing result data Ds_Y, for example, obtaining the processing result data yi of each row of data f1 in Ds_X.

[0279] Step S62, respectively determining the distance between the explained data in the explained data set and the target data to be analyzed, and obtaining a first weight of the explained data according to the distance;

[0280] For example, for a piece of data xi in Dp_X, the distance between each piece of data in Ds_X and the xi can be calculated, and then the first weight of each piece of interpretation data can be obtained according to the distance.

[0281] Optionally, in order to facilitate calculation, in the embodiment of the present application, Ds_X and Dp_x may be preprocessed, and the preprocessing process may include:

[0282] First, the data of the numerical feature dimensions in Ds_X and Dp_X are replaced with discrete data, and the data of the categorical feature dimensions can remain unchanged, and Ds_X' and Dp_x' are obtained respectively.

[0283] In the aforementioned step S511, if the characteristic value E and variance σ of the data within each numerical range of the numerical feature dimension are obtained, further, in each numerical range, based on the maximum value a and minimum value b of the newly generated number within the numerical range, combined with the aforementioned average value E and variance σ, it can be changed to a new numerical value according to a uniform distribution or a Gaussian distribution.

[0284] Among them, we can first construct a uniform distribution U(E, σ) based on the mean value E and variance σ, where the mean value of the uniform distribution is E and the variance is σ. Then, k values ​​are randomly generated in the range (a, b) using the uniform distribution U(E, σ), and the k and the number of explanatory data that fall within the range of the eigenvalues ​​in the explanatory data set. The sum of the number of values ​​k in each range is the aforementioned q.

[0285] It can be understood that the same distribution can be used to determine new values ​​for Ds_X and Dp_X.

[0286] In addition, we can first construct a Gaussian distribution N(E, σ) based on the mean value E and variance σ, where the mean value of the Gaussian distribution is E and the variance is σ. Then, we use the Gaussian distribution N(E, σ) to randomly generate k values ​​in the range (a, b).

[0287] Then, for each piece of data x' to be analyzed in Dp_X', the distances between each piece of data in Ds_X' and the piece of data x' are calculated respectively, and then the first weight of each piece of data in Ds_X' is obtained.

[0288] In practical applications, the value of each piece of data under each characteristic dimension of Ds_X' can be compared with the value of the characteristic dimension in x'; if they are the same, the value is changed to 1, and if they are different, the value is changed to 0. In this way, Ds_X' is processed into a binary array Ds_X", and x' is processed as x" = [1,1,...1].

[0289] Then calculate the Euclidean distance between each row of data in Ds_X" and x", and get d = [d1, d2, ..., dN]. Where N is the number of rows in Ds_X".

[0290] Then according to d=[d 1 , d 2 , …, dn], and the first weight of each row of data is obtained by formula (1).

[0291]

[0292] Among them sw i is the weight of the i-th row, d i is the aforementioned Euclidean distance, e is a natural constant, and n is the total number of rows of De_X” data.

[0293] In this way, we get the weight sw = [sw 1 , sw 2 ,…,sw n ],

[0294] It should be noted that the positions of the above 0 and 1 can also be interchanged, and the embodiments of the present application are not limited thereto.

[0295] Step S63, training an explanation model according to the explanation data, the processing result of the explanation data, and the first weight of the explanation data;

[0296] It can be understood that, as in the aforementioned process, the preprocessed explanation data Ds_X", the corresponding processing result data Ds_Y, and the first weight sw of n rows of explanation data are obtained, and the explanation model can be trained based on the three data.

[0297] For example, the explanation data De_X" has 500 rows, and each row of data has a processing result data yi, and each row of data has a first weight sw i , then we can first construct an explanation training data set. For example, multiply each value in the explanation data [x11,x12,…,x1m] of a row by the weight sw of the row 1 , multiply the processing result data y1 corresponding to the row by the weight sw 1 , and then get a training data {[x11*sw 1 ,x12*sw 1 ,…,x1m*sw 1 ],y1*sw 1}, so n training data can be obtained.

[0298] Then, the n training data are used to train the explanation model built based on m feature dimensions [fe1, fe2,…, fem]. After the model training is completed, the parameter corresponding to each feature dimension is the weight of the corresponding feature dimension.

[0299] Of course, we can also directly select a row of data x from Dp_X as the target data to be analyzed, and then calculate the distance d between each row of data in Ds_X and x, and then calculate sw according to the above formula (1), and then construct a training set based on each row of explained data xi, the yi corresponding to the explained data xi, and the swi of the explained data xi to train the explanation model.

[0300] Step S64, obtaining the target weight of each target feature dimension according to the explained model.

[0301] In the embodiment of the present invention, part or all of the feature dimensions and their weights may be output as required.

[0302] Optionally, in an embodiment of the present application, step S63 includes:

[0303] Step S631, constructing a first interpretation data training set according to the interpretation data, the processing result of the interpretation data and the first weight of the interpretation data;

[0304] Step S632: training a linear regression model according to the first interpretation data training set; wherein the linear regression model includes feature data and a second weight of the feature data.

[0305] In the embodiment of the present application, based on the above {[x11*sw 1 ,x12*sw 1 ,…,x1m*sw 1 ],y1*sw 1}, train a linear regression model, and then the parameters of each feature dimension of the linear regression model are the parameters of each feature dimension of the data to be analyzed in Dp_X.

[0306] Optionally, in an embodiment of the present application, step 202 includes:

[0307] Step S71, in the second interface area, displaying the P feature dimensions with the top second weight sorting and the second weight corresponding to the processing result data; wherein the sorting is the sorting of the second weight from large to small.

[0308] It can be understood that the parameters of each feature dimension calculated in the above manner can be sorted from large to small, and then the top P second weights are selected as target weights for display.

[0309] Optionally, in an embodiment of the present application, after step S632, the method further includes:

[0310] Step S633, selecting target second weights of the top P feature dimensions from the second weights; wherein the sorting is sorting the second weights from large to small;

[0311] Step S634, constructing a second interpretation data training set according to the target feature data corresponding to the target second weight in the interpretation data, the processing result of the interpretation data and the first weight of the interpretation data;

[0312] Step S635: training a logistic regression model according to the second interpretation data training set; wherein the logistic regression model includes the target feature data and a third weight of the target feature data.

[0313] It is understandable that after the aforementioned training of the linear regression model, the difference between its weights may not be obvious, and it is not obvious enough when displayed in a graph or table. In the embodiment of the present application, P second weights and their characteristic dimensions can also be selected from the second weights of the linear regression model in a descending manner.

[0314] Then, the P feature dimension data in De_X” are combined with Ds_Y and the second weight to construct a training set again in a similar manner to the above, and then the logistic regression model is trained based on the training set.

[0315] Then the third weight of each feature dimension in the trained logistic regression model is used as the final target weight.

[0316] Optionally, in an embodiment of the present application, step 202 includes:

[0317] Step S72: In the second interface area, display the third weight corresponding to the processing result data.

[0318] As described above, the third weight of each feature dimension of the logistic regression is displayed corresponding to the corresponding feature dimension and processing structure.

[0319] Of course, in practical applications, the parameters of the logistic regression model may be positive and negative. In the embodiment of the present application, they can be classified and displayed according to positive and negative numbers respectively, and sorted by absolute value in each category. Figure 7 and Figure 8 , where the feature dimensions are sorted by absolute value.

[0320] In an embodiment of the present application, first, input operations on the target data set and the target model are received in the first visual interface area, so that the user can select different target models and target data sets to be predicted according to needs to interpret the weights of feature dimensions.

[0321] Secondly, displaying the weight of at least one characteristic dimension of the data to be analyzed on the processing result data in the visual second interface area can facilitate users to intuitively view the contribution of the characteristic dimension to the prediction result and provide more information support for user decision-making.

[0322] Reference Fig. 9 , which shows a data display method of an embodiment of the present application, applied to an interpretation server, and the method may include the following steps:

[0323] Step 301, providing a display interface to a client; Step 302, receiving a target data set and a target model input by the client through the display interface; the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension;

[0324] Step 303, sending weight data of the processing result data to the client for display in the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data.

[0325] The execution process of this embodiment refers to the aforementioned Figure 1 The execution process of the explanation server is similar in principle and will not be described in detail here.

[0326] Reference Fig.10 , which shows a data display method of an embodiment of the present application, applied to a client, and the method may include the following steps:

[0327] Step 401, obtaining a display interface from a server; Step 402, receiving an input operation on a target data set and a target model in the display interface, and sending the target data set and the target model to the server;

[0328] Step 403, receiving weight data for the processing result data returned by the server and displaying it on the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data.

[0329] The execution process of this embodiment refers to the aforementioned Figure 1 The execution process of the client is similar in principle and will not be described in detail here.

[0330] In an embodiment of the present application, first, input operations on a target data set and a target model are received in a visual display interface, so that the user can select different target models and target data sets to be predicted according to needs to interpret the weights of feature dimensions.

[0331] Secondly, displaying the weight of at least one feature dimension of the data to be analyzed on the processing result data in a visual display interface can facilitate users to intuitively view the contribution of the feature dimension to the prediction result and provide more information support for user decision-making.

[0332] It should be noted that, for the method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0333] On the basis of the above embodiments, this embodiment further provides a data display device, which is applied to electronic devices such as terminal devices and servers.

[0334] Reference Fig.11 , which shows a data display device of an embodiment of the present application, which may specifically include:

[0335] A first receiving module 511 is used to receive an input operation on a target data set and a target model in a display interface, wherein the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension;

[0336] The first display module 512 is used to display weight data of the processing result data in the display interface; wherein the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions to the processing result data.

[0337] Optionally, also include:

[0338] A fourth receiving module is used to receive an input operation on an explanation data set on the display interface; the explanation data set includes explanation data corresponding to the data to be analyzed respectively; wherein the explanation data includes feature data of at least one target feature dimension among at least one feature dimension of the data to be analyzed; wherein, for the same feature dimension, the first feature dimension identifier of the explanation data is different from the second feature dimension identifier of the corresponding data to be analyzed; wherein there is an association relationship between at least one of the first feature dimension identifiers and the second feature dimension identifier belonging to the same feature dimension.

[0339] Optionally, the display interface includes: an explanation data source selection control, an explanation data table selection control, and a feature dimension association control;

[0340] The fourth receiving module is also used to: receive an interpretation data source identifier in the interpretation data source selection control; the interpretation data source identifier is used to access the interpretation data source corresponding to the interpretation data source identifier; receive a selection operation on the interpretation data table identifier in the interpretation data table selection control; the interpretation data table identifier is used to obtain an interpretation data set corresponding to the interpretation data table identifier from the interpretation data source; and receive a setting operation in the interpretation data table selection control for associating a first feature dimension identifier in the interpretation data set with a second feature dimension identifier in the target data set.

[0341] Optionally, the display interface includes a model identification input control, a prediction data source identification input control, and a prediction data table selection control; and the first receiving module is further used to:

[0342] Receiving a target model identifier input in the model identifier input control; the target model identifier is used to call the target model;

[0343] Receiving a predicted data source identifier input in the data source identifier input control; the predicted data source identifier is used to access the predicted data source;

[0344] A selection operation on a target data table identifier is received in the prediction data table selection control; the target data table identifier is used to obtain the target data set from the prediction data source.

[0345] Optionally, also include:

[0346] The first providing module is used to provide a prediction control; the prediction control is used to display the processing result data on the display interface when a trigger operation on the prediction control is received.

[0347] Optionally, also include:

[0348] The second providing module is used to provide an explanation object selection control for each processing result data in the display interface; the explanation object selection control is used to determine the processing result data to be explained.

[0349] Optionally, also include:

[0350] The third providing module is used to provide an explanation control; the explanation control is used to obtain the weight data of the processing result data when a trigger operation on the explanation control is received.

[0351] Optionally, the display interface includes:

[0352] Fixed columns and slidable columns, the fixed columns include: a button column for displaying an interpretation object selection control, an interpretation status column for displaying whether the interpretation has been completed, a result display column for displaying a processing result, and a probability display column for displaying at least one of the predicted probabilities of the processing results; the slidable columns include: a feature data column corresponding to at least one of the feature dimensions.

[0353] Optionally, also include:

[0354] A status marking module is used to set the interpretation status corresponding to the processing result data in the interpretation status column to an uninterpreted mark when there is no weight data in the processing result data; and to set the interpretation status corresponding to the processing result data in the interpretation status column to an interpreted mark when there is weight data in the processing result data.

[0355] Optionally, also include:

[0356] A selection control setting module is used to set the selection control corresponding to the processing result data in the option button column to be unselectable when the processing result data has weight data.

[0357] Optionally, the first display module is further used for:

[0358] Displaying weight data of the processing result data in a list on the display interface;

[0359] Alternatively, the weight data of the processing result data is displayed graphically on the display interface.

[0360] Optionally, the first display module is also used to display processing result data on the display interface, and when weight data already exists for the processing result data, if an expansion operation on the processing result data is received, the weight data is displayed in the display interface at a position corresponding to the processing result data.

[0361] Optionally, the display interface includes a first interface area and a second interface area;

[0362] The receiving of input operations on the target data set and the target model in the display interface includes:

[0363] Receiving input operations on a target data set and a target model in the first interface area;

[0364] The displaying of weight data of the processing result data in the display interface includes:

[0365] The weighting data for the processing result data is displayed in the second interface area.

[0366] Optionally, also include:

[0367] A first target data determination module, configured to use each piece of interpretation data in the interpretation data set as target data to be analyzed;

[0368] A first result data determination module is used to use the processing result of the corresponding data to be analyzed in the target data set as the processing result of the explained data according to the association relationship between the first feature dimension identifier and the second feature dimension identifier belonging to the same feature dimension;

[0369] a first weight calculation module, used to respectively determine the distance between the explained data in the explained data set and the target data to be analyzed, and obtain a first weight of the explained data according to the distance;

[0370] A first model training module, used for training an explanation model according to the explanation data, a processing result of the explanation data, and a first weight of the explanation data;

[0371] The first target weight determination module is used to obtain the target weight of each target feature dimension according to the explained model.

[0372] Optionally, also include:

[0373] A training distribution acquisition module, used to acquire the distribution of feature data of each feature dimension in a training data set used to train the target model;

[0374] The explanation data generation module is used to determine an explanation data set according to the distribution; the distribution of the feature data of each feature dimension of the explanation data set is the same as the distribution of the feature data of each feature dimension in the training data set.

[0375] Optionally, the training distribution acquisition module is further used to: for a numerical feature dimension, according to M preset numerical ranges, respectively count the first number of values ​​in each numerical range to obtain a first quantity distribution under the numerical feature dimension;

[0376] The explanation data generation module is also used for:

[0377] For the numerical feature dimension, N numerical values ​​are generated; wherein the number of first numerical values ​​within each of the preset numerical ranges satisfies the first numerical distribution.

[0378] Optionally, the training distribution acquisition module is further used for:

[0379] A category distribution acquisition module, configured to count a second number of the same category value under a category-type feature dimension, and obtain a second quantity distribution under the category-type feature dimension;

[0380] The explanation data generation module is also used for:

[0381] For the categorical feature dimension, N category values ​​are generated; wherein the N category values ​​satisfy the second quantity distribution.

[0382] Optionally, also include:

[0383] a second result data determination module, processing the explained data set according to the target model to obtain processing result data corresponding to the explained data set;

[0384] a second weight obtaining module, used to respectively determine the distance between the explained data in the explained data set and the target data to be analyzed, and obtain a first weight of the explained data according to the distance;

[0385] A second training module, used for training an explanation model according to the explanation data, the processing result of the explanation data, and the first weight of the explanation data;

[0386] The second target weight acquisition module is used to obtain the target weight of each target feature dimension according to the explained model.

[0387] Optionally, the second training module is further used for:

[0388] constructing a first interpretation data training set according to the interpretation data, the processing result of the interpretation data and the first weight of the interpretation data;

[0389] A linear regression model is trained according to the first interpretation data training set; wherein the linear regression model includes feature data and a second weight of the feature data.

[0390] Optionally, the first display module is further used for:

[0391] In the display interface, the P feature dimensions with the highest second weight ranking and the second weight are displayed corresponding to the processing result data; wherein the ranking is a ranking from large to small according to the second weight.

[0392] Optionally, also include:

[0393] A weight selection module, used to select target second weights of the top P feature dimensions from the second weights; wherein the sorting is sorting the second weights from large to small;

[0394] A training construction module, used for constructing a second interpretation data training set according to target feature data corresponding to the target second weight in the interpretation data, a processing result of the interpretation data and the first weight of the interpretation data;

[0395] The third training module is used to train a logistic regression model according to the second explanation data training set; wherein the logistic regression model includes the target feature data and a third weight of the target feature data.

[0396] Optionally, the first display module is further used for:

[0397] In the display interface, the third weight is displayed corresponding to the processing result data.

[0398] In an embodiment of the present application, first, input operations on the target data set and the target model are received on a visual display interface, so that the user can select different target models and target data sets to be predicted according to needs to interpret the weights of feature dimensions.

[0399] Secondly, displaying the weight of at least one feature dimension of the data to be analyzed on the processing result data in a visual display interface can facilitate users to intuitively view the contribution of the feature dimension to the prediction result and provide more information support for user decision-making.

[0400] Reference Fig.12 It shows a data display device of an embodiment of the present application, which is applied to an interpretation server and may specifically include:

[0401] The page providing module 511 is used to provide a display interface to the client;

[0402] The second receiving module 512 is used to receive a target data set and a target model input by the client through the display interface; the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension;

[0403] The result sending module 513 is used to send weight data of the processing result data to the client for display in the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data.

[0404] Reference Fig.13 , which shows a data display device of an embodiment of the present application, applied to a client, and may specifically include:

[0405] The page acquisition module 521 is used to acquire the display interface from the server; the third receiving module 522 is used to receive the input operation of the target data set and the target model in the display interface, and send the target data set and the target model to the server;

[0406] The second display module 523 is used to receive weight data of the processing result data returned by the server and display it on the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data.

[0407] In an embodiment of the present application, first, input operations on the target data set and the target model are received on a visual display interface, so that the user can select different target models and target data sets to be predicted according to needs to interpret the weights of feature dimensions.

[0408] Secondly, displaying the weight of at least one feature dimension of the data to be analyzed on the processing result data in a visual display interface can facilitate users to intuitively view the contribution of the feature dimension to the prediction result and provide more information support for user decision-making.

[0409] The embodiment of the present application also provides a non-volatile readable storage medium, which stores one or more modules (programs). When the one or more modules are applied to a device, the device can execute instructions (instructions) of each method step in the embodiment of the present application.

[0410] The present application embodiment provides one or more machine-readable media on which instructions are stored, and when executed by one or more processors, an electronic device executes one or more of the methods described in the above embodiments. In the present application embodiment, the electronic device includes various types of devices such as terminal devices and servers (clusters).

[0411] The embodiments of the present disclosure may be implemented as a device configured as desired using any appropriate hardware, firmware, software, or any combination thereof, and the device may include electronic devices such as terminal devices and servers (clusters). Figure 7 An exemplary apparatus 700 that can be used to implement various embodiments described in this application is schematically illustrated.

[0412] For one embodiment, Figure 7 An exemplary apparatus 700 is shown having one or more processors 702, a control module (chip set) 704 coupled to at least one of the processor(s) 702, a memory 706 coupled to the control module 704, a non-volatile memory (NVM) / storage device 708 coupled to the control module 704, one or more input / output devices 710 coupled to the control module 704, and a network interface 712 coupled to the control module 704.

[0413] The processor 702 may include one or more single-core or multi-core processors, and the processor 702 may include any combination of general-purpose processors or special-purpose processors (such as graphics processors, application processors, baseband processors, etc.). In some embodiments, the device 700 can be used as a terminal device, server (cluster), etc. described in the embodiments of the present application.

[0414] In some embodiments, the apparatus 700 may include one or more computer-readable media (e.g., memory 706 or NVM / storage device 708) having instructions 714 and one or more processors 702 combined with the one or more computer-readable media and configured to execute the instructions 714 to implement a module to perform the actions described in the present disclosure.

[0415] For one embodiment, the control module 704 may include any suitable interface controller to provide any suitable interface to at least one of the processor(s) 702 and / or any suitable device or component in communication with the control module 704 .

[0416] The control module 704 may include a memory controller module to provide an interface to the memory 706. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0417] The memory 706 may be used, for example, to load and store data and / or instructions 714 for the device 700. For one embodiment, the memory 706 may include any suitable volatile memory, such as a suitable DRAM. In some embodiments, the memory 706 may include double data rate type four synchronous dynamic random access memory (DDR4 SDRAM).

[0418] For one embodiment, control module 704 may include one or more input / output controllers to provide an interface to NVM / storage device 708 and input / output device(s) 710 .

[0419] For example, NVM / storage 708 may be used to store data and / or instructions 714. NVM / storage 708 may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable non-volatile storage device(s) (e.g., one or more hard disk drives (HDDs), one or more compact disk (CD) drives, and / or one or more digital versatile disk (DVD) drives).

[0420] NVM / storage device 708 may include storage resources that are physically part of the device on which apparatus 700 is installed, or it may be accessible to the device without being part of the device. For example, NVM / storage device 708 may be accessed via input / output device(s) 710 over a network.

[0421] (One or more) input / output devices 710 may provide an interface for the apparatus 700 to communicate with any other appropriate device, and the input / output device 710 may include a communication component, an audio component, a sensor component, etc. The network interface 712 may provide an interface for the apparatus 700 to communicate through one or more networks, and the apparatus 700 may wirelessly communicate with one or more components of a wireless network according to any of one or more wireless network standards and / or protocols, for example, accessing a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G, 5G, etc., or a combination thereof for wireless communication.

[0422] For one embodiment, at least one of the processor(s) 702 may be packaged together with the logic of one or more controllers (e.g., a memory controller module) of the control module 704. For one embodiment, at least one of the processor(s) 702 may be packaged together with the logic of one or more controllers of the control module 704 to form a system-in-package (SiP). For one embodiment, at least one of the processor(s) 702 may be integrated on the same die with the logic of one or more controllers of the control module 704. For one embodiment, at least one of the processor(s) 702 may be integrated on the same die with the logic of one or more controllers of the control module 704 to form a system-on-chip (SoC).

[0423] In various embodiments, the apparatus 700 may be, but is not limited to, a terminal device such as a server, a desktop computing device, or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet computer, a netbook, etc.). In various embodiments, the apparatus 700 may have more or fewer components and / or a different architecture. For example, in some embodiments, the apparatus 700 includes one or more cameras, a keyboard, a liquid crystal display (LCD) screen (including a touch screen display), a non-volatile memory port, multiple antennas, a graphics chip, an application-specific integrated circuit (ASIC), and a speaker.

[0424] Among them, the main control chip can be used as a processor or control module in the device, sensor data, location information, etc. are stored in a memory or NVM / storage device, the sensor group can be used as an input / output device, and the communication interface may include a network interface.

[0425] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0426] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0427] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data analysis terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data analysis terminal device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0428] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data analysis terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0429] These computer program instructions can also be loaded onto a computer or other programmable data analysis terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0430] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0431] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0432] The data analysis method and device, an electronic device and a storage medium provided by the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A data display method, characterized in that: The method comprises: Receiving an input operation on a target data set and a target model in a display interface, the target data set including at least one piece of data to be analyzed, and the data to be analyzed including feature data of at least one feature dimension; The weight data of the processing result data are displayed in the display interface; wherein the processing result data include: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data include: the target weight of at least one of the feature dimensions for the processing result data, and the target weight is obtained by determining the distance between each explanatory data in the explanatory data set and the data to be analyzed after the data to be analyzed is processed by the target model, and obtaining the first weight of the explanatory data according to the distance, and training the explanatory model according to the explanatory data, the processing result of the data to be analyzed, and the first weight of the explanatory data, and after the training is completed, the explanatory model is obtained.

2. The method according to claim 1, characterized in that Also includes: An input operation for an explanation data set is received on the display interface; the explanation data set includes explanation data corresponding to the data to be analyzed respectively; wherein the explanation data includes feature data of at least one target feature dimension among at least one feature dimension of the data to be analyzed; wherein, for the same feature dimension, a first feature dimension identifier of the explanation data is different from a second feature dimension identifier of the corresponding data to be analyzed; wherein there is an association relationship between at least one of the first feature dimension identifiers and a second feature dimension identifier belonging to the same feature dimension.

3. The method according to claim 2, characterized in that The display interface includes: an explanation data source selection control, an explanation data table selection control, and a feature dimension association control; The receiving an input operation for explaining the data set in the display interface includes: receiving an interpretation data source identifier in the interpretation data source selection control; the interpretation data source identifier is used to access an interpretation data source corresponding to the interpretation data source identifier; A selection operation of an explanation data table identifier is received in the explanation data table selection control; the explanation data table identifier is used to obtain an explanation data set corresponding to the explanation data table identifier from an explanation data source; A setting operation for associating a first feature dimension identifier in the explained data set with a second feature dimension identifier in the target data set is received in the explained data table selection control.

4. The method according to any one of claims 1 to 3, characterized in that The display interface includes a model identification input control, a prediction data source identification input control, and a prediction data table selection control; the receiving of input operations for the target data set and the target model in the display interface includes: Receiving a target model identifier input in the model identifier input control; the target model identifier is used to call the target model; Receiving a predicted data source identifier input in the data source identifier input control; the predicted data source identifier is used to access the predicted data source; A selection operation on a target data table identifier is received in the prediction data table selection control; the target data table identifier is used to obtain the target data set from the prediction data source.

5. The method according to any one of claim 1, characterized in that: Also includes: A prediction control is provided; the prediction control is used to display the processing result data in the display interface when a trigger operation on the prediction control is received.

6. The method according to claim 5, characterized in that Also includes: Providing an explanation object selection control for each piece of processing result data in the display interface; The explained object selection control is used to determine the processing result data to be explained.

7. The method according to claim 5, characterized in that Also includes: An explanation control is provided; the explanation control is used to obtain weight data of the processing result data when a trigger operation on the explanation control is received.

8. The method according to claim 6, characterized in that The display interface includes: Fixed columns and slidable columns, the fixed columns include: a button column for displaying an interpretation object selection control, an interpretation status column for displaying whether the interpretation has been completed, a result display column for displaying a processing result, and a probability display column for displaying at least one of the predicted probabilities of the processing results; the slidable columns include: a feature data column corresponding to at least one of the feature dimensions.

9. The method according to claim 8, characterized in that Also includes: In the case where the processing result data does not have weight data, setting the interpretation status corresponding to the processing result data in the interpretation status column to an uninterpreted mark; In the case where the processing result data has weight data, the interpretation status corresponding to the processing result data in the interpretation status column is set to an interpreted flag.

10. The method according to claim 9, characterized in that Also includes: In the case that the processing result data has weight data, the selection control corresponding to the processing result data in the option button column is set to be unselectable.

11. The method according to any one of claims 1-3, 5-10, characterized in that: The displaying of the weight data of the processing result data in the display interface includes: Displaying weight data of the processing result data in a list in the display interface; Alternatively, the weight data of the processing result data is displayed graphically in the display interface.

12. The method according to claim 11, characterized in that The displaying of weight data of the processing result data in the display interface includes: When the processing result data is displayed in the display interface and weight data already exists in the processing result data, if an expansion operation on the processing result data is received, the weight data is displayed in the display interface at a position corresponding to the processing result data.

13. The method according to claim 1, characterized in that The display interface includes a first interface area and a second interface area; The receiving of input operations on the target data set and the target model in the display interface includes: Receiving input operations on a target data set and a target model in the first interface area; The displaying of weight data of the processing result data in the display interface includes: The weighting data for the processing result data is displayed in the second interface area.

14. The method according to claim 2, characterized in that The determining of the distance between each piece of explanation data in the explanation data set and the data to be analyzed, and obtaining a first weight of the explanation data according to the distance, comprises: Taking each piece of explanation data in the explanation data set as target data to be analyzed; According to the association relationship between the first feature dimension identifier and the second feature dimension identifier belonging to the same feature dimension, the processing result of the corresponding data to be analyzed in the target data set is used as the processing result of the explained data; The distances between the explained data in the explained data set and the target data to be analyzed are respectively determined, and a first weight of the explained data is obtained according to the distances.

15. The method according to claim 1, characterized in that Also includes: Obtaining the distribution of feature data of each feature dimension in a training data set used to train the target model; According to the distribution, an explanation data set is determined; the distribution of the feature data of each feature dimension of the explanation data set is the same as the distribution of the feature data of each feature dimension in the training data set.

16. The method according to claim 15, characterized in that The obtaining the distribution of feature data of each feature dimension in the training data set used to train the target model includes: For a numerical feature dimension, according to M preset numerical ranges, first quantities of numerical values ​​in each numerical range are counted respectively to obtain a first quantity distribution under the numerical feature dimension; Determining the explanation data set according to the distribution includes: For the numerical feature dimension, N numerical values ​​are generated; wherein the number of first numerical values ​​within each of the preset numerical ranges satisfies the first numerical distribution.

17. The method according to claim 15, characterized in that The obtaining the distribution of feature data of each feature dimension in the training data set used to train the target model includes: For a categorical feature dimension, under the feature dimension, a second quantity with the same category value is counted to obtain a second quantity distribution under the categorical feature dimension; Determining the explanation data set according to the distribution includes: For the categorical feature dimension, N category values ​​are generated; wherein the N category values ​​satisfy the second quantity distribution.

18. The method according to any one of claims 15 to 17, characterized in that: Also includes: Processing the explained data set according to the target model to obtain processing result data corresponding to the explained data set; respectively determining the distances between the explained data in the explained data set and the target data to be analyzed, and obtaining a first weight of the explained data according to the distances; training an explanation model according to the explanation data, the processing result of the explanation data, and the first weight of the explanation data; According to the explained model, the target weight of each target feature dimension is obtained.

19. The method according to claim 14 or 17, characterized in that: The step of training the explanation model according to the explanation data, the processing result of the explanation data, and the first weight of the explanation data comprises: constructing a first interpretation data training set according to the interpretation data, the processing result of the interpretation data and the first weight of the interpretation data; A linear regression model is trained according to the first interpretation data training set; wherein the linear regression model includes feature data and a second weight of the feature data.

20. The method according to claim 19, characterized in that The displaying of weight data of the processing result data in the display interface includes: In the display interface, the P feature dimensions with the highest second weight ranking and the second weight are displayed corresponding to the processing result data; wherein the ranking is a ranking from large to small according to the second weight.

21. The method according to claim 19, characterized in that Also includes: From the second weights, select the target second weights of the top P feature dimensions; wherein the sorting is sorting the second weights from large to small; constructing a second interpretation data training set according to the target feature data corresponding to the target second weight in the interpretation data, the processing result of the interpretation data and the first weight of the interpretation data; A logistic regression model is trained based on the second explanation data training set; wherein the logistic regression model includes the target feature data and a third weight of the target feature data.

22. The method according to claim 21, characterized in that The displaying of weight data of the processing result data in the display interface includes: In the display interface, the third weight is displayed corresponding to the processing result data.

23. The method according to any one of claims 1-3, 5-10, 13-17, characterized in that: The target data set includes at least one of a tax data set, a customs declaration data set, a medical data set, and a corporate behavior data set; The target model includes: at least one of a model for predicting a taxpayer's tax risk based on a tax data set, a model for predicting a customs declaration risk based on a customs declaration data set, a model for predicting a user's illness based on a medical data set, and a model for predicting corporate behavior risk based on corporate behavior data.

24. A data display method, applied to an interpretation server, characterized in that: The method comprises: Providing a display interface to the client; receiving a target data set and a target model input by the client through the display interface; the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension; Send weight data of the processing result data to the client for display in the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data, and the target weight is determined after the data to be analyzed is processed by the target model, and the distance between each explanation data in the explanation data set and the data to be analyzed is determined, and the first weight of the explanation data is obtained according to the distance; the explanation model is trained according to the explanation data, the processing result of the data to be analyzed, and the first weight of the explanation data, and after the training is completed, the weight is obtained according to the explanation model.

25. A data display method, applied to a client, characterized in that: The method comprises: Acquire a display interface from a server; receive input operations on a target data set and a target model in the display interface, and send the target data set and the target model to the server, wherein the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension; Receive weight data for the processing result data returned by the server and display it in the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data, and the target weight is determined after the data to be analyzed is processed by the target model, and the distance between each explanation data in the explanation data set and the data to be analyzed is determined, and the first weight of the explanation data is obtained according to the distance; the explanation model is trained according to the explanation data, the processing result of the data to be analyzed, and the first weight of the explanation data, and after the training is completed, the weight is obtained according to the explanation model.

26. A data display device, characterized in that: The device comprises: A first receiving module is used to receive an input operation on a target data set and a target model in a display interface, wherein the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension; A first display module is used to display weight data of processing result data in the display interface; wherein, the processing result data includes: processing results obtained by processing the data to be analyzed according to the target model, and the weight data includes: target weight of at least one of the feature dimensions for the processing result data, and the target weight is obtained by determining the distance between each explanatory data in the explanatory data set and the data to be analyzed after the data to be analyzed is processed by the target model, and obtaining the first weight of the explanatory data according to the distance, and training the explanatory model according to the explanatory data, the processing results of the data to be analyzed, and the first weight of the explanatory data, and after the training is completed, the explanatory model is obtained.

27. A data display device, applied to an interpretation server, characterized in that: The device comprises: A page providing module is used to provide a display interface to the client; a second receiving module is used to receive a target data set and a target model input by the client through the display interface; the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension; A result sending module is used to send weight data of the processing result data to the client for display in the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data, and the target weight is determined after the data to be analyzed is processed by the target model, and the distance between each explanation data in the explanation data set and the data to be analyzed is determined, and the first weight of the explanation data is obtained according to the distance, and the explanation model is trained according to the explanation data, the processing result of the data to be analyzed, and the first weight of the explanation data, and after the training is completed, the weight is obtained according to the explanation model.

28. A data display device, applied to a client, characterized in that: The device comprises: a page acquisition module, configured to acquire a display interface from a server; a third receiving module, configured to receive input operations on a target data set and a target model in the display interface, and send the target data set and the target model to the server, wherein the target data set includes at least one piece of data to be analyzed, and the data to be analyzed includes feature data of at least one feature dimension; The second display module is used to receive weight data of the processing result data returned by the server and display it on the display interface; the processing result data includes: the processing result obtained by processing the data to be analyzed according to the target model, and the weight data includes: the target weight of at least one of the feature dimensions for the processing result data, and the target weight is determined after the data to be analyzed is processed by the target model, and the distance between each explanation data in the explanation data set and the data to be analyzed is determined, and the first weight of the explanation data is obtained according to the distance, and the explanation model is trained according to the explanation data, the processing result of the data to be analyzed, and the first weight of the explanation data, and after the training is completed, the weight is obtained according to the explanation model.

29. An electronic device, characterized in that: include: processor; and A memory having executable codes stored thereon, which, when executed, causes the processor to perform the method according to any one of claims 1 to 25.

30. One or more machine-readable media having executable codes stored thereon, which, when executed, cause a processor to perform the method according to any one of claims 1 to 25.

Images