A method and system for adaptive data analysis and model management

Abstract

The present application relates to a method and system for adaptive data analysis and model management, comprising: obtaining multiple sample data associated with a target business; setting labels for the multiple sample data; using any one of the multiple sample data as the target sample data, extracting a words from the target sample data and performing part-of-speech analysis on the a words; screening b words with noun parts of speech from the a words; screening c words with matching data values ​​from the b words; creating sample features and corresponding feature values ​​for deep learning training; training a neural network model; extracting multiple features and matching feature values ​​from the business data according to the input layer of the neural network model, inputting the feature values ​​of the multiple features into the neural network model, and obtaining expected result data of the target business execution output by the neural network model. The present invention does not require manual processing of sample data for training the neural network model, and is more efficient than existing technologies.

Description

Technical Field

[0001] The present invention relates to the field of deep learning, and more specifically, to a method and system for adaptive data analysis and model management. Background Art

[0002] Deep learning is a new research direction in machine learning, introduced to bring it closer to its original goal: artificial intelligence. Deep learning involves learning the inherent patterns and representational hierarchies of sample data. The information gained from this learning process significantly aids in interpreting data such as text, images, and sound. Its ultimate goal is to enable machines to acquire human-like analytical and learning capabilities, enabling them to recognize data such as text, images, and sound. Deep learning requires training a neural network model, an algorithmic mathematical model that mimics the behavioral characteristics of animal neural networks and performs distributed parallel information processing. This network relies on the complexity of the system to adjust the interconnections between its numerous nodes to achieve its information processing goals.

[0003] In existing technologies, different types of businesses require different neural network models to be trained. Therefore, different sample data must be prepared for each business. This sample data must then be manually broken down and analyzed before neural network model training can begin. This process of analyzing and breaking down large amounts of sample data is time-consuming and inefficient, hindering the generation and application of neural network models. Therefore, a technical solution is needed that can automatically analyze and process sample data and efficiently complete neural network model training. Summary of the Invention

[0004] In order to solve the above technical problems, the present application is proposed to provide a method and system for adaptive data analysis and model management that can automatically analyze and process sample data and efficiently complete neural network model training.

[0005] In a first aspect, the present invention provides a method for adaptive data analysis and model management, comprising: for a target business to be analyzed, obtaining a plurality of sample data associated with the target business; setting labels for the plurality of sample data, wherein the labels reflect expected result data obtained by analyzing the plurality of sample data; taking any one of the plurality of sample data as target sample data, extracting a words from the target sample data and performing part-of-speech analysis on the a words; screening out b words whose part of speech is noun from the a words; detecting from the target sample data whether the b words have matching data values; screening out c words with matching data values ​​from the b words and retaining them. words; creating sample features and corresponding feature values ​​for deep learning training based on the c words and the matching data values; training a neural network model based on the sample features and labels corresponding to the multiple sample data, wherein the input layer of the neural network model is determined based on the sample features corresponding to the multiple sample data, and the output layer of the neural network model is determined based on the labels corresponding to the multiple sample data; obtaining the business data generated during the execution of the target business; extracting multiple features and matching feature values ​​from the business data based on the input layer of the neural network model, inputting the feature values ​​of the multiple features into the neural network model, and obtaining the expected result data of the target business execution output by the neural network model.

[0006] Preferably, the aforementioned method of adaptive data analysis and model management, the step of creating sample features and corresponding feature values ​​for deep learning training based on the c words and matching data values ​​includes: taking any word of the c words as a candidate sample feature, comparing the name of the candidate sample feature with the names of multiple standard features recorded in a preset data dictionary corresponding to the target business, and when the name of the candidate sample feature is the same as the name of any standard feature in the data dictionary or the similarity is higher than a preset threshold, modifying the name of the candidate sample feature to the name of the standard feature and recording it as the sample feature of the target sample data, and its feature value is the data value corresponding to the candidate sample feature, wherein the formula for calculating the similarity is

[0007] ,

[0008] represents the candidate sample features, represents any standard feature in the data dictionary, express and The similarity between express The characters, express The characters, express length, express length, Indicates that The front characters are changed to The front The number of operations required to generate a character.

[0009] Preferably, the aforementioned method of adaptive data analysis and model management, the step of modifying the name of the candidate sample feature to the name of the standard feature and recording it as the sample feature of the target sample data, whose characteristic value is the data value corresponding to the candidate sample feature, also includes: judging whether the characteristic value corresponding to the sample feature complies with the characteristic value rule pre-recorded in the data dictionary and corresponding to the standard feature, and deleting the sample feature when the judgment result is no.

[0010] Preferably, in the aforementioned method for adaptive data analysis and model management, the step of setting labels for the plurality of sample data includes: for each sample data among the plurality of sample data, obtaining result data for the sample data submitted by a plurality of users; and calculating the difference between the result data submitted by each of the plurality of users and the result data submitted by other users:

[0011] ,

[0012] in, represents result data submitted by one of the multiple users, represents result data submitted by any other user among the multiple users, Indicates that among the multiple users The difference between the submitted result data and the result data of other users, represents the number of other users among the multiple users; and sets the result data corresponding to the minimum gap value as the value of the label corresponding to the corresponding sample data.

[0013] Preferably, in the aforementioned method of adaptive data analysis and model management, the step of screening and retaining c words with matching data values ​​from the b words also includes: calculating the frequency of occurrence of each word in the c words in the multiple sample data, and selecting one or more words from the c words to filter out according to the frequency of occurrence.

[0014] Preferably, in the aforementioned method for adaptive data analysis and model management, the step of selecting one or more words from the c words and filtering them out according to their frequency of occurrence includes: calculating the number of words filtered out from the c words according to the time limit for analyzing data by the neural network model. , where z is a constant and t is the maximum time limit for the neural network model to analyze data.

[0015] Preferably, in the aforementioned method for adaptive data analysis and model management, the step of selecting and retaining c words with matching data values ​​from the b words further comprises: selecting any word that appears more than once in the plurality of sample data from the c words; calculating the dispersion according to the plurality of data values ​​corresponding to the word; ,in, is the oth data value of the word, is the average value of multiple data values ​​corresponding to the word; when the dispersion of the word is higher than a preset threshold, the word is filtered out from the c words.

[0016] Preferably, in the aforementioned method of adaptive data analysis and model management, the step of training the neural network model according to the sample features and labels corresponding to the plurality of sample data includes: obtaining the CPU resources and memory resources currently used to train the neural network model; calculating the number of iterations in the training process of the neural network model according to the CPU resources and memory resources of the neural network model; ,in, Indicates the current CPU resources, Indicates the current memory resources, h and is the weight coefficient.

[0017] In a second aspect, the present invention provides a system for adaptive data analysis and model management, comprising: a sample data acquisition module, for acquiring, for a target business to be analyzed, a plurality of sample data associated with the target business; a label setting module, for setting labels for the plurality of sample data, the labels reflecting expected result data obtained by analysis based on the plurality of sample data; a word segmentation module, for taking any one of the plurality of sample data as target sample data, extracting a words from the target sample data and performing part-of-speech analysis on the a words; a first screening module, for screening out b words whose part of speech is noun from the a words; a data value detection module, for detecting whether the b words have matching data values ​​from the target sample data; a second screening module, for screening out the words with matching data values ​​from the b words. a feature creation module, which creates sample features and corresponding feature values ​​for deep learning training based on the c words and the matching data values; a model training module, which trains a neural network model based on the sample features and labels corresponding to the multiple sample data, wherein the input layer of the neural network model is determined based on the sample features corresponding to the multiple sample data, and the output layer of the neural network model is determined based on the labels corresponding to the multiple sample data; a business data acquisition module, which acquires the business data generated during the execution of the target business; a result output module, which extracts multiple features and matching feature values ​​from the business data based on the input layer of the neural network model, inputs the feature values ​​of the multiple features into the neural network model, and obtains the expected result data of the target business execution output by the neural network model.

[0018] The above one or more technical solutions of the present invention have at least one or more of the following beneficial effects:

[0019] The technical solution of the present invention does not manually disassemble and analyze the sample data like the existing technical solution. Instead, after obtaining the sample data corresponding to the target business and setting labels, it first performs word segmentation and part-of-speech analysis. Since the name of the feature is often a noun, words with a noun part of speech can be filtered out. Secondly, since the feature must have a matching feature value, the remaining noun words can be screened and filtered for a second time. Through two screening and filtering, the sample features and the corresponding feature values ​​can be automatically extracted from the sample data. Combined with the labels, the input layer and output layer of the neural network model can be clarified and trained. Based on the trained neural network model, the business data generated during the business execution process can be analyzed and processed in real time and the predicted result data can be output. The process of training the neural network model using sample data does not require human participation, which is more efficient than the existing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 is a flow chart of adaptive data analysis and model management according to an embodiment of the present application;

[0022] Figure 2 is a partial flow chart of adaptive data analysis and model management according to an embodiment of the present application;

[0023] Figure 3 is a partial flow chart of adaptive data analysis and model management according to an embodiment of the present application;

[0024] Figure 4 is a partial flow chart of adaptive data analysis and model management according to an embodiment of the present application;

[0025] Figure 5 is a partial flow chart of adaptive data analysis and model management according to an embodiment of the present application;

[0026] Figure 6 4 is a block diagram of a system for adaptive data analysis and model management according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] Some embodiments of the present invention are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0028] like Figure 1 As shown, in one embodiment of the present invention, a method for adaptive data analysis and model management is provided, comprising:

[0029] Step S110 : For a target business to be analyzed, a plurality of sample data that are associated with the target business are obtained.

[0030] In this embodiment, different sample data need to be obtained for different types of services, and the sample data is often in text form.

[0031] Step S120 : setting labels for the plurality of sample data, where the labels reflect expected result data obtained by analyzing the plurality of sample data.

[0032] In this embodiment, for different types of services, the content reflected by the label of each sample data is also different. For example, it can be the classification of the content of each sample data, the expected service execution status, etc.

[0033] Step S130 : Taking any sample data from the plurality of sample data as target sample data, extracting a words from the target sample data and performing part-of-speech analysis on the a words.

[0034] In this embodiment, word segmentation and part-of-speech analysis may be performed on the target sample data based on NLP (natural language processing) technology.

[0035] Step S140 , selecting b words whose part of speech is noun from the a words.

[0036] In this embodiment, since the names of features are often nouns, words whose part of speech is noun can be filtered out accordingly.

[0037] Step S150 , detecting whether b words have matching data values ​​from the target sample data.

[0038] Step S160 , screening and retaining c words with matching data values ​​from the b words.

[0039] In this embodiment, since the feature must have a matching feature value, the remaining noun words can be subjected to secondary screening and filtering based on this.

[0040] Step S170: Create sample features and corresponding feature values ​​for deep learning training based on the c words and matching data values.

[0041] In this embodiment, the sample features and corresponding feature values ​​can be automatically extracted from the sample data through two screening and filtering operations, which significantly improves the efficiency compared to the existing technical solutions.

[0042] Step S180: training a neural network model based on sample features and labels corresponding to the plurality of sample data, wherein an input layer of the neural network model is determined based on the sample features corresponding to the plurality of sample data, and an output layer of the neural network model is determined based on the labels corresponding to the plurality of sample data.

[0043] Step S190: Acquire business data generated during the execution of the target business.

[0044] Step S1100, extracting multiple features and matching feature values ​​from the business data according to the input layer of the neural network model, inputting the feature values ​​of the multiple features into the neural network model, and obtaining the expected result data of the target business execution output by the neural network model.

[0045] According to the technical solution of this embodiment, after obtaining the sample data corresponding to the target business and setting the label, word segmentation and part-of-speech analysis are first performed. Since the name of the feature is often a noun, the words with the part of speech being a noun can be filtered out based on this. Secondly, since the feature must have a characteristic value that matches it, the remaining noun words can be screened and filtered for the second time. Through two screening and filtering, the sample features and the corresponding characteristic values ​​can be automatically extracted from the sample data. Combined with the label, the input layer and output layer of the neural network model can be clarified and trained. Based on the trained neural network model, the business data generated during the business execution process can be analyzed and processed in real time and the predicted result data can be output. The process of training the neural network model using sample data does not require human participation, which is more efficient than the existing technology.

[0046] like Figure 2 As shown, another embodiment of the present invention further provides a method for adaptive data analysis and model management. Compared with the above embodiment, the method for adaptive data analysis and model management of this embodiment includes step S170:

[0047] In step S210 , any word among the c words is used as a candidate sample feature, and the name of the candidate sample feature is compared with the names of multiple standard features recorded in a preset data dictionary corresponding to the target business.

[0048] In this embodiment, a digital dictionary is pre-set, and the data dictionary records the standard feature names when performing model training for the target business, which serves as a basis for extracting features and feature values ​​for training from the c words.

[0049] Step S220: When the name of the candidate sample feature is the same as the name of any standard feature in the data dictionary or the similarity is higher than a preset threshold, the name of the candidate sample feature is modified to the name of the standard feature and recorded as the sample feature of the target sample data. Its feature value is the data value corresponding to the candidate sample feature. The formula for calculating the similarity is:

[0050] ,

[0051] represents the candidate sample features, represents any standard feature in the data dictionary, express and The similarity between express The characters, express The characters, express length, express length, Indicates that The front characters are changed to The front The number of operations required to generate a character.

[0052] In this embodiment, considering that there are cases where the descriptions of words in the sample data are not standardized, similarity analysis is performed on c words based on a digital dictionary, and the words that can be used for model training are found and modified according to the standard feature names. After modification, they are used as features to be used for training and their feature values ​​are recorded.

[0053] Step S230 , judging whether the feature value corresponding to the sample feature meets the feature value rule pre-recorded in the data dictionary and corresponding to the standard feature, and deleting the sample feature when the judgment result is no.

[0054] In this embodiment, the feature value rules corresponding to different standard features are also recorded in the data dictionary. The feature value rules stipulate the reasonable range of feature values. Therefore, if the feature value corresponding to the feature extracted from the sample data does not comply with the feature value rules, it means that the feature value is unreasonable or inaccurate, and it is necessary to delete it.

[0055] like Figure 3 As shown, another embodiment of the present invention further provides a method for adaptive data analysis and model management. Compared with the above embodiment, the method for adaptive data analysis and model management of this embodiment includes step S120:

[0056] Step S310 : For each sample data among the plurality of sample data, obtain result data for the sample data submitted by a plurality of users.

[0057] In this embodiment, for each sample data, different users can submit different result data as a basis for generating a label.

[0058] Step S320: Calculate the difference between the result data submitted by each user and the result data submitted by other users:

[0059] ,

[0060] in, represents result data submitted by one of the multiple users, represents result data submitted by any other user among the multiple users, Indicates that among the multiple users The difference between the submitted result data and the result data of other users, Indicates the number of other users in the plurality of users.

[0061] Step S330 , setting the result data corresponding to the minimum gap value as the value of the label corresponding to the corresponding sample data.

[0062] In this embodiment, the result data submitted by a user with the smallest gap relative to the result data submitted by other users is screened out from the result data submitted by multiple users. The smallest gap often indicates that the result data submitted by the user is closest to the accurate result value. Therefore, a label can be set according to the result data corresponding to the smallest gap value.

[0063] Another embodiment of the present invention further provides a method for adaptive data analysis and model management. Compared to the aforementioned embodiment, the method for adaptive data analysis and model management of this embodiment further includes, in step S160:

[0064] Calculate the frequency of each word in the c words in multiple sample data, and select one or more words from the c words to filter out according to the frequency of occurrence.

[0065] In this embodiment, words with low frequency of occurrence in the sample data lack sufficient feature values ​​when used as sample features for model training, so it is necessary to filter them out.

[0066] Among them, the number of words filtered out from c words can be calculated based on the time limit of the neural network model to analyze the data , where z is a constant and t is the maximum time limit for the neural network model to analyze data.

[0067] In this embodiment, requirements are put forward for the analysis efficiency of the neural network model according to business needs, and the analysis efficiency of the neural network model is directly related to the number of features input into the input layer. Therefore, based on the analysis time limit of the neural network model, the number of input features of the neural network model can be controlled, thereby controlling the number of filtered c words.

[0068] like Figure 4 As shown, another embodiment of the present invention further provides a method for adaptive data analysis and model management. Compared with the above embodiment, the method for adaptive data analysis and model management in this embodiment further includes, in step S160:

[0069] Step S410 : selecting any word that appears more than once in a plurality of sample data from the c words.

[0070] Step S420: Calculate the dispersion based on the multiple data values ​​corresponding to the word. ,in, is the oth data value of the word, is the average value of multiple data values ​​corresponding to the word;

[0071] When the dispersion of the word is higher than a preset threshold, the word is filtered out from the c words.

[0072] Step S430: When the dispersion of the word is higher than a preset threshold, the word is filtered out from the c words.

[0073] In this embodiment, the data values ​​of the same word in multiple sample data are comprehensively analyzed to analyze the discreteness of the data values ​​of the word. The higher the discreteness, the more dispersed the data values ​​corresponding to the word are, the worse the stability is, and the lower the contribution to the neural network training is. In this case, it is necessary to filter it out.

[0074] like Figure 5 As shown, another embodiment of the present invention further provides a method for adaptive data analysis and model management. Compared with the above embodiment, the method for adaptive data analysis and model management in this embodiment includes step S180:

[0075] Step S510, obtaining the CPU resources and memory resources currently used to train the neural network model.

[0076] Step S520: Calculate the number of iterations during the neural network model training process based on the CPU resources and memory resources of the neural network model. ,in, Indicates the current CPU resources, Indicates the current memory resources, h and is the weight coefficient.

[0077] In this embodiment, the neural network model training will occupy a large amount of CPU resources and memory resources. Too many iterations during the training process will cause insufficient CPU and memory resources. Therefore, the number of iterations of the neural network model training can be reasonably controlled according to the currently available CPU resources and memory resources.

[0078] like Figure 6 As shown, in one embodiment of the present invention, a system for adaptive data analysis and model management is provided, comprising:

[0079] The sample data acquisition module 610 acquires a plurality of sample data associated with a target business to be analyzed.

[0080] In this embodiment, different sample data need to be obtained for different types of services, and the sample data is often in text form.

[0081] The label setting module 620 sets labels for the plurality of sample data, where the labels reflect expected result data obtained by analyzing the plurality of sample data.

[0082] In this embodiment, for different types of services, the content reflected by the label of each sample data is also different. For example, it can be the classification of the content of each sample data, the expected service execution status, etc.

[0083] The word segmentation module 630 takes any sample data from the plurality of sample data as target sample data, extracts a words from the target sample data, and performs part-of-speech analysis on the a words.

[0084] In this embodiment, word segmentation and part-of-speech analysis may be performed on the target sample data based on NLP (natural language processing) technology.

[0085] The first screening module 640 screens out b words whose part of speech is noun from the a words.

[0086] In this embodiment, since the names of features are often nouns, words whose part of speech is noun can be filtered out accordingly.

[0087] The data value detection module 650 detects whether b words in the target sample data have matching data values.

[0088] The second screening module 660 screens out the b words and retains c words with matching data values.

[0089] In this embodiment, since the feature must have a matching feature value, the remaining noun words can be subjected to secondary screening and filtering based on this.

[0090] The feature creation module 670 creates sample features and corresponding feature values ​​for deep learning training based on the c words and matching data values.

[0091] In this embodiment, the sample features and corresponding feature values ​​can be automatically extracted from the sample data through two screening and filtering operations, which significantly improves the efficiency compared to the existing technical solutions.

[0092] The model training module 680 trains a neural network model based on the sample features and labels corresponding to the multiple sample data, wherein the input layer of the neural network model is determined based on the sample features corresponding to the multiple sample data, and the output layer of the neural network model is determined based on the labels corresponding to the multiple sample data.

[0093] The business data acquisition module 690 acquires business data generated during the execution of the target business.

[0094] The result output module 6100 extracts multiple features and matching feature values ​​from the business data according to the input layer of the neural network model, inputs the feature values ​​of the multiple features into the neural network model, and obtains the expected result data of the target business execution output by the neural network model.

[0095] According to the technical solution of this embodiment, after obtaining the sample data corresponding to the target business and setting the label, word segmentation and part-of-speech analysis are first performed. Since the name of the feature is often a noun, the words with the part of speech being a noun can be filtered out based on this. Secondly, since the feature must have a characteristic value that matches it, the remaining noun words can be screened and filtered for the second time. Through two screening and filtering, the sample features and the corresponding characteristic values ​​can be automatically extracted from the sample data. Combined with the label, the input layer and output layer of the neural network model can be clarified and trained. Based on the trained neural network model, the business data generated during the business execution process can be analyzed and processed in real time and the predicted result data can be output. The process of training the neural network model using sample data does not require human participation, which is more efficient than the existing technology.

[0096] The basic principles of the present application have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in this application are merely illustrative and not restrictive, and it should not be assumed that these advantages, strengths, and effects are required of each embodiment of this application. In addition, the specific details disclosed above are merely illustrative and facilitating understanding, and are not restrictive. The above details do not limit this application to necessarily being implemented using the above specific details.

[0097] The block diagrams of the devices, devices, equipment, and systems involved in this application are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, or configured in any manner. Words such as "include," "comprise," "have," and the like are open-ended words, meaning "including but not limited to," and can be used interchangeably therewith. The words "or" and "and" used herein refer to the words "and / or" and can be used interchangeably therewith, unless the context clearly indicates otherwise. The word "such as" used herein refers to the phrase "such as but not limited to," and can be used interchangeably therewith.

[0098] It should also be noted that in the apparatus, device, and method of the present application, each component or each step can be decomposed and / or recombined, and such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0099] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0100] The above description has been provided for the purpose of illustration and description. Furthermore, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although a number of example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations thereof.

Claims

1. A method for adaptive data analysis and model management, characterized in that: include: For a target business to be analyzed, a plurality of sample data items are obtained that are associated with the target business; Setting labels for the plurality of sample data, wherein the labels reflect expected result data obtained by analyzing the plurality of sample data; Taking any one of the plurality of sample data as target sample data, extracting a words from the target sample data and performing part-of-speech analysis on the a words; Filtering b words whose part of speech is noun from the a words; Detecting whether the b words have matching data values ​​from the target sample data; Filter and retain c words with matching data values ​​from the b words; Creating sample features and corresponding feature values ​​for deep learning training based on the c words and the matching data values; Training a neural network model based on the sample features and labels corresponding to the plurality of sample data, wherein an input layer of the neural network model is determined based on the sample features corresponding to the plurality of sample data, and an output layer of the neural network model is determined based on the labels corresponding to the plurality of sample data; Acquire business data generated during the execution of the target business; Extracting multiple features and matching feature values ​​from the business data according to the input layer of the neural network model, inputting the feature values ​​of the multiple features into the neural network model, and obtaining expected result data of the business execution output by the neural network model, The step of "creating sample features and corresponding feature values ​​for deep learning training based on the c words and matching data values" includes: Take any word of the c words as a candidate sample feature, compare the name of the candidate sample feature with the names of multiple standard features recorded in the preset data dictionary corresponding to the target business, and when the name of the candidate sample feature is the same as the name of any standard feature in the data dictionary or the similarity is higher than a preset threshold, modify the name of the candidate sample feature to the name of the standard feature and record it as the sample feature of the target sample data, and its feature value is the data value corresponding to the candidate sample feature, wherein the formula for calculating the similarity is: ； represents the candidate sample features, represents any standard feature in the data dictionary, express and The similarity between express The characters, express The characters, express length, express length, Indicates that The front characters are changed to The front The number of operations required to generate a character.

2. The method of adaptive data analysis and model management according to claim 1, characterized in that: The step of “changing the name of the candidate sample feature to the name of the standard feature and recording it as the sample feature of the target sample data, whose feature value is the data value corresponding to the candidate sample feature” includes: According to the feature value rule corresponding to the standard feature pre-recorded in the data dictionary, it is judged whether the feature value corresponding to the sample feature meets the feature value rule. When the judgment result is no, the sample feature is deleted.

3. The method of adaptive data analysis and model management according to claim 1, characterized in that: The step of “setting labels for the plurality of sample data” includes: For each sample data among the plurality of sample data, obtaining result data for the sample data submitted by a plurality of users; Calculate the gap between the result data submitted by each user and the result data submitted by other users: ； in, represents result data submitted by one of the multiple users, represents result data submitted by any other user among the multiple users, Indicates that among the multiple users The difference between the submitted result data and the result data of other users, represents the number of other users in the plurality of users; Set the result data corresponding to the minimum gap value to the value of the label corresponding to the corresponding sample data.

4. The method of adaptive data analysis and model management according to claim 1, characterized in that: The step of "screening and retaining c words with matching data values ​​from the b words" includes: The frequency of each of the c words appearing in the plurality of sample data is calculated, and one or more words are selected from the c words and filtered out according to the frequency of appearance.

5. The method of adaptive data analysis and model management according to claim 4, characterized in that: The step of "selecting one or more words from the c words and filtering them out according to their frequency of occurrence" includes: Calculate the number of words filtered out from the c words based on the time limit for analyzing data by the neural network model ,in, is a constant, The maximum time limit for analyzing data for the neural network model.

6. The method of adaptive data analysis and model management according to claim 1, characterized in that: The step of "screening and retaining c words with matching data values ​​from the b words" includes: Selecting any word that appears more than once in the plurality of sample data from the c words; Calculate the dispersion based on the multiple data values ​​corresponding to the word ,in, is the oth data value of the word, is the average value of multiple data values ​​corresponding to the word; When the dispersion of the word is higher than a preset threshold, the word is filtered out from the c words.

7. The method of adaptive data analysis and model management according to claim 1, characterized in that: The step of “training a neural network model based on the sample features and labels corresponding to the plurality of sample data” includes: Obtaining the CPU resources and memory resources currently used to train the neural network model; Calculate the number of iterations during the training of the neural network model based on the CPU resources and memory resources of the neural network model ,in, Indicates the current CPU resources, Indicates the current memory resources, and is the weight coefficient.

8. A system for adaptive data analysis and model management, characterized in that: include: A sample data acquisition module, for a target business to be analyzed, acquires a plurality of sample data associated with the target business; a label setting module, configured to set labels for the plurality of sample data, wherein the labels reflect expected result data obtained by analyzing the plurality of sample data; A word segmentation module, taking any sample data from the plurality of sample data as target sample data, extracting a words from the target sample data and performing part-of-speech analysis on the a words; A first screening module selects b words whose part of speech is noun from the a words; A data value detection module detects whether the b words have matching data values ​​from the target sample data; A second screening module, screening and retaining c words with matching data values ​​from the b words; A feature creation module, which creates sample features and corresponding feature values ​​for deep learning training based on the c words and matching data values; a model training module for training a neural network model based on the sample features and labels corresponding to the plurality of sample data, wherein an input layer of the neural network model is determined based on the sample features corresponding to the plurality of sample data, and an output layer of the neural network model is determined based on the labels corresponding to the plurality of sample data; A business data acquisition module, which acquires business data generated during the execution of the target business; A result output module extracts multiple features and matching feature values ​​from the business data according to the input layer of the neural network model, inputs the feature values ​​of the multiple features into the neural network model, and obtains the expected result data of the business execution output by the neural network model. Among them, "creating sample features and corresponding feature values ​​for deep learning training based on the c words and matching data values" includes: Take any word of the c words as a candidate sample feature, compare the name of the candidate sample feature with the names of multiple standard features recorded in the preset data dictionary corresponding to the target business, and when the name of the candidate sample feature is the same as the name of any standard feature in the data dictionary or the similarity is higher than a preset threshold, modify the name of the candidate sample feature to the name of the standard feature and record it as the sample feature of the target sample data, and its feature value is the data value corresponding to the candidate sample feature, wherein the formula for calculating the similarity is: ； represents the candidate sample features, represents any standard feature in the data dictionary, express and The similarity between express The characters, express The characters, express length, express length, Indicates that The front characters are changed to The front The number of operations required to generate a character.

Images