An AI large model-based application development data management system and method

Abstract

The application discloses an application development data management system and method based on an AI large model, relates to the technical field of data management, and comprises the following steps: multi-dimensional data collection is performed on an application development process, corresponding application development records are generated, and a high-dimensional feature set is obtained; multi-dimensional data in any application development record is mined, and conflicts existing between various features are identified; the correlation between different features is analyzed, and a weight coefficient is set for each feature; the corresponding data of any feature in each application development record is analyzed, and the conflict proportion of any feature is obtained; a multi-target joint model is constructed to evaluate the development of any application development record; the high-dimensional feature set of real-time development records at the current moment is obtained, preliminary evaluation is performed through the multi-target joint model; the expected change of each feature is analyzed, and risk early warning is performed on the expected conflict.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of data management, and particularly relates to an application development data management system and method based on an AI large model. BACKGROUND

[0002] With the rapid development of artificial intelligence technology, especially the application of large models, data management systems have become increasingly important in various industries; in the field of application development, the importance of large models in data management cannot be underestimated;

[0003] Existing application development data management systems often rely on manual rules and simple statistical analysis, which are difficult to handle complex data relationships and high-dimensional features, and are more dependent on shallow features in project recommendation, which is difficult to effectively integrate multi-source heterogeneous data, and is prone to target conflicts in the recommended results, such as technical adaptation and resource overallocation; at the same time, the rule-based system lacks dynamic optimization capability, and does not establish a historical conflict analysis mechanism, which leads to repeated resource allocation errors and cannot realize self-adaptive optimization. SUMMARY

[0004] The application aims to provide an application development data management system and method based on an AI large model to solve the problems in the prior art.

[0005] To achieve the above-mentioned purpose, the application provides the following technical scheme: an application development data management method based on an AI large model, the management method comprising the following steps:

[0006] Step S100: Multi-dimensional data collection is performed on the application development process, corresponding application development records are generated, and a high-dimensional feature set of the application development records is obtained; the multi-dimensional data in any application development record is mined to identify conflicts between features;

[0007] Step S200: Based on the conflict between different features, the association between different features is analyzed; according to the association between different features, a weight coefficient is set for each feature;

[0008] Step S300: The corresponding data of any feature in each application development record is analyzed to obtain the conflict proportion of any feature; by combining the corresponding weight coefficient, a multi-objective joint model is constructed to evaluate the development of any application development record;

[0009] Step S400: Obtain the high-dimensional feature set of the real-time development record at the current time, and perform preliminary evaluation through the multi-objective joint model; extract the weight coefficient of any feature at the current time, analyze the expected change of each feature, and perform risk warning on the expected conflict.

[0010] Further, step S100 includes the following steps:

[0011] Step S101: Pre-set several data collection methods for application development process data collection, obtain several dimensions of development data sets, wherein one data collection method corresponds to one dimension of development data set; summarize the several dimensions of development data sets to generate an application development record; data collection for application development process can be implemented through API interface, log embedding, sensor and third party data platform, etc. The development data can be divided into user behavior data, development log data, resource data and external environment data, etc.

[0012] Step S102: Arbitrarily select a development data set of any dimension in an application development record, if the development data in the development data set is structured data, extract the data type and data field of the development data to obtain a feature group, if the development data of the selected dimension is unstructured data, use AI large model to perform semantic analysis on the development data and extract several features to obtain a feature group; summarize the feature groups of each dimension to obtain a high-dimensional feature set of the selected application development record; the purpose of structured data division is to ensure the accuracy of feature extraction, for example, keywords in demand documents, code annotations and other unstructured data need to be extracted as features, while resource utilization, progress ratio, demand submission frequency and other structured data, in addition to the data type itself, the corresponding numerical changes also need to be analyzed as features;

[0013] Step S103: Pre-construct a conflict type database, the conflict type database contains several conflict types, and each conflict type is assigned a corresponding conflict feature set, arbitrarily select a feature group from the high-dimensional feature set, and arbitrarily select a conflict type, compare the selected feature group with the conflict feature set of the selected conflict type, if there is the same feature, set the selected feature group as the same feature group; the conflict type includes technical conflict such as incompatibility of development tools, resource conflict such as imbalance of budget allocation and running overload, time sequence conflict such as parallel task competition, process delay, etc.

[0014] Step S104: Count the number of same feature groups as n1, set the number of conflict features in the conflict feature set as N, calculate the proportion of the number of same feature groups as α = n1 / N, pre-set a number proportion threshold α th , if α ≥ α thIf the conflict type corresponding to the conflict feature set is selected as one conflict type of the application development record, the conflict type is identified by the same feature group, and the purpose is to exclude the interference of the numerical value pair feature comparison. If the specific numerical value is also used as a feature, the conflict type will be different between different records, which will affect the conflict identification. Therefore, by setting the feature group, the feature group with the same feature is used as the judgment basis for identifying the conflict type, which can effectively identify the conflict type.

[0015] Further, step S200 includes the following steps:

[0016] Step S201: Arbitrarily select a conflict type in an application development record, and arbitrarily select two conflict features from the selected conflict type, and set them as the ith conflict feature and the jth conflict feature, respectively. An association set R(i,j) between the two conflict features is established.

[0017] Step S202: Obtain the number of conflict types containing the association set R(i,j) in each conflict type, which is m1 (i,j) , set the total number of conflict types as M1, calculate the conflict number ratio of the association set R(i,j) as β1=m1 (i,j) / M1; count the number of records containing the association set R(i,j) in each application development record as m2 (i,j) , set the number of application development records as M2, calculate the conflict frequency of the association set R(i,j) as β2=m2 (i,j) / M2; according to the formula:

[0018] ;

[0019] The association value G(i,j) between the ith conflict feature and the jth conflict feature is calculated. The association between the two features mainly considers the number of conflict types involved by the two features and the conflict frequency in the historical records, and both directions can directly reflect the association;

[0020] Step S203: Arbitrarily select the ith conflict feature, and obtain the association value between the ith conflict feature and the remaining conflict features, according to the formula:

[0021] ;

[0022] Wherein, e is the total number of conflict features; the feature coefficient P i of the ith conflict feature is calculated; the feature coefficients of each conflict feature are obtained, and the formula is:

[0023]

[0024] The weight coefficient Q of the ith conflict feature is calculatedi .

[0025] Further, the step S300 comprises the following steps:

[0026] Step S301: arbitrarily selecting an i-th conflict feature in the application development record, obtaining a development data set in the dimension where the i-th conflict feature is located, if the development data set is structured data and there is a numerical range corresponding to the data field, then presetting a conflict numerical range (d min ,d max ) for the i-th conflict feature; obtaining the numerical range (d1 i ,d2 i ) of the i-th conflict feature in the selected application development record, if d1 i <d min <d2 i <d max , obtaining a first conflict proportion η1=(d2 i -d min ) / (d max -d min ), if d min <d1 i <d2 i <d max , obtaining a first conflict proportion η1=(d2 i -d1 i ) / (d max -d min ), if d min <d1 i <d max <d2 i , obtaining a first conflict proportion η1=(d max -d1 i ) / (d max -d min ); the first conflict proportion is when the conflict feature has a numerical range, it is necessary to consider whether the numerical fluctuation range in the actual development process is in the normal interval, so as to reflect the influence range of the existence of the conflict feature;

[0027] Step S302: if the development data set does not have a numerical range, obtaining the number of the same features v in the development data set, setting the number of features in the development data set as v total , obtaining a second conflict proportion η2=v / v total ; the second conflict proportion represents the proportion of the number of keywords involved when there is no numerical range, reflecting the correlation between the specific conflict type;

[0028] Step S303: setting the feature label of the i-th conflict feature as Flag iIf the i-th conflict feature has a numerical range, then Flag i =1, if the i-th conflicting feature does not have a numerical range, then Flag i =0; Construct a multi-objective joint model to evaluate any application development record:

[0029] ;

[0030] Where, η1 i η2 represents the proportion of the first conflict for the i-th conflict feature. i Let Z be the second conflict proportion of the i-th conflict feature; input each conflict feature in the selected application development record into the multi-objective joint model to obtain the evaluation value Z of the selected application development record.

[0031] Furthermore, step S400 includes the following steps:

[0032] Step S401: Collect development data under each dimension in the real-time development record, and obtain feature groups of arbitrary dimensions respectively, generate a high-dimensional feature set of the real-time development record, compare the high-dimensional feature set with the conflict feature sets of each conflict type, and obtain a set of identical features; arbitrarily select the u-th feature from the set of identical features, and obtain the weight coefficient of the u-th feature at the current time as (Q u ) now ;

[0033] Step S402: Obtain all development data corresponding to the u-th feature. If the development data of the u-th feature has a numerical range, then obtain the first conflict ratio η1 of the u-th feature. u If no numerical range exists, then the second conflict proportion η2 is obtained. u ; Input all features into the multi-objective joint model to obtain the real-time evaluation value Z of the real-time development record. now ;

[0034] Step S403: Preset an anomaly assessment threshold Z th If Z now ≥Z th If Z... now <Z th Then the number of features contained in the same feature set is n. now According to the formula:

[0035] ;

[0036] Here, Select() is the selection function. If the development data of the u-th feature has a numerical range, then Select(u) = η1. u, otherwise, Select(u) = η2 u ; calculate the expected weight coefficient (Q u ) now ; if (Q u ) now < Q u ) now , set the u-th feature as an abnormal feature;

[0037] Step S404: count the number of abnormal features in the same feature set n ’ , and obtain the abnormal proportion τ = n ’ / n now ; if τ ≥ τ th , a risk warning is given to the real-time development record, wherein τ th is an abnormal proportion threshold.

[0038] In order to better implement the above method, an application development data management system is also proposed, which includes a multi-dimensional feature analysis module, a feature correlation analysis module, a model weight analysis module and a real-time abnormality identification module.

[0039] The multi-dimensional feature analysis module is used for multi-dimensional data collection of the application development process, generating corresponding application development records and obtaining a high-dimensional feature set of the application development records; and mining multi-dimensional data in any application development record to identify conflicts between features.

[0040] The feature correlation analysis module is used for analyzing the correlation between different features based on the conflicts between different features; and setting a weight coefficient for each feature according to the correlation between different features.

[0041] The model weight analysis module is used for analyzing the corresponding data of any feature in each application development record to obtain the conflict proportion of any feature; and constructing a multi-objective joint model to evaluate the development of any application development record by combining the corresponding weight coefficient.

[0042] The real-time abnormality identification module is used for obtaining a high-dimensional feature set of real-time development records at the current time, and performing preliminary evaluation through a multi-objective joint model; extracting the weight coefficient of any feature at the current time, analyzing the expected change of each feature, and giving a risk warning for the expected conflict.

[0043] Further, the multi-dimensional feature analysis module includes a high-dimensional feature generation unit and a feature conflict analysis unit.

[0044] The high-dimensional feature generation unit is configured to perform multi-dimensional data collection on the application development process, generate corresponding application development records, and obtain a high-dimensional feature set of the application development records.

[0045] Further, the feature correlation analysis module comprises a feature correlation analysis unit and a weight coefficient setting unit.

[0046] The feature correlation analysis unit is configured to analyze the correlation between different features based on the conflict between the different features, and the weight coefficient setting unit is configured to set a weight coefficient for each feature according to the correlation between the different features.

[0047] Further, the model weight analysis module comprises a conflict proportion analysis unit and a joint model construction unit.

[0048] The conflict proportion analysis unit is configured to analyze the corresponding data of any feature in each application development record to obtain the conflict proportion of the any feature, and the joint model construction unit is configured to construct a multi-objective joint model to evaluate the development of any application development record by combining the corresponding weight coefficients.

[0049] Further, the real-time anomaly identification module comprises a preliminary evaluation analysis unit and a target conflict prediction unit.

[0050] The preliminary evaluation analysis unit is configured to obtain the high-dimensional feature set of the real-time development record at the current time and perform preliminary evaluation through the multi-objective joint model, and the target conflict prediction unit is configured to extract the weight coefficient of any feature at the current time, analyze the expected change of each feature, and perform risk warning on the expected conflict.

[0051] Compared with the prior art, the present application has the following advantages:

[0052] 1. The present application automatically extracts the features of multi-source heterogeneous data by efficiently processing multi-dimensional data and constructing a high-dimensional feature set, which solves the problem of insufficient high-dimensional feature integration compared with manual rules and shallow feature analysis in traditional methods, and lays a foundation for subsequent conflict identification and dynamic optimization.

[0053] 2. The present application can quickly identify abnormal features and trigger risk warning by real-time analysis of feature conflict proportion and correlation, construction of an evaluation model based on historical data, real-time collection of development records and comparison of conflict feature sets, effectively avoiding target conflict problems such as technical adaptation and resource over-provisioning, and significantly improving the adaptability and risk control capability of the development process.

[0054] 3、The application can effectively optimize resource allocation strategies, reduce the frequency of conflicts in the development process, thereby reducing the cost of manual intervention, and comprehensively improving the efficiency and reliability of application development. BRIEF DESCRIPTION OF DRAWINGS

[0055] Fig. 1 A step schematic diagram of an application development data management method based on an AI large model;

[0056] Fig. 2 A structure schematic diagram of an application development data management system based on an AI large model. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the application.

[0058] Embodiment: As shown in the figure, the application provides an application development data management method based on an AI large model, and the management method comprises the following steps: Figs. 1-2

[0059] Step S100: Multi-dimensional data collection is performed on the application development process, corresponding application development records are generated, and a high-dimensional feature set of the application development records is obtained; and the multi-dimensional data in any application development record is mined to identify the conflict between the features.

[0060] The step S100 comprises the following steps:

[0061] Step S101: A plurality of data collection methods are preset to collect data on the application development process, and a plurality of sets of development data are obtained, wherein one data collection method corresponds to one set of development data; and the plurality of sets of development data are summarized to generate one application development record.

[0062] Step S102: Any set of development data of any dimension in an application development record is selected, if the development data in the set of development data is structured data, the data type and data field of the development data are extracted to obtain a feature group, if the development data of the selected dimension is unstructured data, the development data is semantically analyzed and a plurality of features are extracted using an AI large model to obtain a feature group; and the feature groups of the dimensions are summarized to obtain a high-dimensional feature set of the selected application development record. ​

[0063] Step S103: A conflict type database is constructed in advance, the conflict type database contains a plurality of conflict types, and each conflict type is assigned a corresponding conflict feature set. A feature set is randomly selected from the high-dimensional feature set, a conflict type is randomly selected, and the selected feature set is compared with the conflict feature set of the selected conflict type. If there is a same feature, the selected feature set is set as the same feature set.

[0064] Step S104: The number of same feature sets is n1, the number of conflict features in the conflict feature set is N, the proportion of the number of same feature sets is a = n1 / N, and a number proportion threshold a is preset. th If a ≥ a th , the conflict type corresponding to the conflict feature set is taken as a conflict type of the selected application development record.

[0065] Step S200: Based on the conflict between different features, the association between different features is analyzed, and the weight coefficient of each feature is set according to the association between different features.

[0066] Step S200 includes the following steps:

[0067] Step S201: A conflict type in an application development record is randomly selected, two conflict features are randomly selected from the selected conflict type, and the selected conflict features are set as the ith conflict feature and the jth conflict feature, respectively. An association set R(i,j) between the two conflict features is established.

[0068] Step S202: The number of conflict types containing the association set R(i,j) in each conflict type is m1 (i,j) , the total number of conflict types is M1, and the conflict proportion of the association set R(i,j) is calculated as β1 = m1 (i,j) / M1; the number of records containing the association set R(i,j) in each application development record is m2 (i,j) , the number of application development records is M2, and the conflict frequency of the association set R(i,j) is calculated as β2 = m2 (i,j) / M2; according to the formula:

[0069]

[0070] The association value G(i,j) between the ith conflict feature and the jth conflict feature is calculated.

[0071] ​Example 1: Extract the association set R(i,j) of the i-th conflict feature and the j-th conflict feature. Find that the number of association sets R(i,j) contained in all conflict types is 5, and the total number of conflict types is 10. This indicates that the conflict rate of association set R(i,j) is 50%. Simultaneously, count that the number of application development records containing association sets R(i,j) is 10, and the total number of records is 50. This indicates that the conflict occurrence frequency of association set R(i,j) is 20%. Calculate the association value G(i,j) of association set R(i,j) = 50% × 20% = 10%.

[0072] Step S203: Randomly select the i-th conflicting feature, and obtain the association values ​​between the i-th conflicting feature and the other conflicting features, according to the formula:

[0073] ;

[0074] Where e is the total number of conflict features; the feature coefficient P of the i-th conflict feature is calculated. i Obtain the feature coefficients of each conflict feature according to the formula:

[0075]

[0076] The weight coefficient Q of the i-th conflict feature is calculated. i ;

[0077] Example 2: Obtain the correlation values ​​between the i-th conflicting feature and the other three conflicting features as 10%, 20%, and 20%, respectively, and calculate the feature coefficient P of the i-th conflicting feature. i =10%×20%×20%=0.4%; Similarly, the feature coefficients of the remaining conflict features are 0.5%, 0.5% and 0.6% respectively, and the weight coefficient Q of the i-th conflict feature is calculated. i =0.4% / 2%=20%.

[0078] Step S300: Analyze the corresponding data of any feature in each application development record to obtain the conflict ratio of any feature; by combining the corresponding weight coefficients, construct a multi-objective joint model to evaluate the development status of any application development record;

[0079] Step S300 includes the following steps:

[0080] Step S301: Randomly select the i-th conflict feature in an application development record, obtain the development data set of the dimension containing the i-th conflict feature. If the development data set is structured data and the corresponding data field has a numerical range, then preset a conflict numerical range (d) for the i-th conflict feature. min ,dmax ); Obtain the numerical range (d1) of the i-th conflict feature in the selected application development record. i ,d2 i If d1 i <d min <d2 i <d max The proportion of the first conflict is obtained as η1 = (d2) i -d min ) / (d max -d min If d min <d1 i <d2 i <d max The proportion of the first conflict is obtained as η1 = (d2) i -d1 i ) / (d max -d min If d min <d1 i <d max <d2 i The proportion of the first conflict η1 is obtained as follows: (d max -d1 i ) / (d max -d min );

[0081] Step S302: If the development dataset does not have a numerical range, then obtain the number of identical features in the development dataset as v, and set the number of features in the development dataset as v. total The proportion of the second conflict is obtained as η2 = v / v total ;

[0082] Step S303: Set the feature label of the i-th conflicting feature as Flag. i If the i-th conflict feature has a numerical range, then Flag i =1, if the i-th conflicting feature does not have a numerical range, then Flag i =0; Construct a multi-objective joint model to evaluate any application development record:

[0083] ;

[0084] Where, η1 i η2 represents the proportion of the first conflict for the i-th conflict feature. i Let Z be the second conflict proportion of the i-th conflict feature; input each conflict feature in the selected application development record into the multi-objective joint model to obtain the evaluation value Z of the selected application development record;

[0085] Embodiment 3: Set 3 conflict features exist in the selected application development record, 1 conflict feature resource utilization exists in the numerical range (40%, 60%), and the preset conflict numerical range is (0%, 50%), the first conflict ratio η1=(50%-40%) / 50%=20% is calculated; and the remaining 2 conflict features such as R&D log information and requirement document do not exist in the numerical range, then feature extraction is performed, and the same feature quantity is 5 and 7 respectively, it is set that the number of features corresponding to the two conflict features in the development data set is 15 and 20 respectively, and two second conflict ratios of 33.3% and 35% are obtained respectively; the weight coefficients of the three conflict features are set as 20%, 40% and 40% respectively, and the evaluation value Z of the selected application development record is calculated as Z=20%×20%+40%×33.3%+40%×35%=4%+13.32%+14%=0.3132.

[0086] Step S400: Obtain the high-dimensional feature set of the real-time development record at the current time, and preliminarily evaluate through the multi-objective joint model; extract the weight coefficient of any feature at the current time, analyze the expected change of each feature, and give a risk warning for the conflict that exists in expectation;

[0087] Among them, step S400 includes the following steps:

[0088] Step S401: Collect the development data in each dimension of the real-time development record, and obtain the feature group of any dimension respectively, generate the high-dimensional feature set of the real-time development record, compare the high-dimensional feature set with the conflict feature set of each conflict type, and obtain a same feature set; select the u-th feature from the same feature set, and obtain the weight coefficient of the u-th feature at the current time as (Q u ) now ;

[0089] Step S402: Obtain all development data corresponding to the u-th feature, if the development data of the u-th feature exists in the numerical range, then obtain the first conflict ratio η1 u of the u-th feature, if there is no numerical range, then obtain the second conflict ratio η2 u ; input all features into the multi-objective joint model to obtain the real-time evaluation value Z now of the real-time development record;

[0090] Step S403: preset an abnormal evaluation threshold Z th , if Z now ≥Z th , an abnormal reminder is given to the real-time development record, if Z now <Z th , the number of features contained in the same feature set is nnow , according to the formula:

[0091] ;

[0092] wherein Select() is a selection function, if the development data of the u-th feature exists a value range, then Select(u)=η1 u , otherwise, Select(u)=η2 u ; the expected weight coefficient (Q u ) now of the u-th feature is calculated; if (Q u ) now < (Q u ) now , the u-th feature is set as an abnormal feature;

[0093] Step S404: the number of abnormal features in the same feature set is counted as n ’ , and the abnormal proportion τ=n ’ / n now is obtained; if τ≥τ th , a risk warning is given to the real-time development record, wherein τ th is an abnormal proportion threshold.

[0094] An application development data management system, the management system comprising a multi-dimensional feature analysis module, a feature correlation analysis module, a model weight analysis module and a real-time abnormality identification module;

[0095] The multi-dimensional feature analysis module is configured to collect multi-dimensional data in the application development process, generate corresponding application development records, and obtain a high-dimensional feature set of the application development records; and mine the multi-dimensional data in any application development record and identify conflicts between features.

[0096] The feature correlation analysis module is configured to analyze correlations between different features based on the conflicts between the different features; and set weight coefficients for the features based on the correlations between the different features.

[0097] The model weight analysis module is configured to analyze corresponding data of any feature in each application development record to obtain a conflict proportion of the feature; and construct a multi-objective joint model to evaluate the development of any application development record by combining the corresponding weight coefficients.

[0098] The real-time abnormality identification module is configured to obtain a high-dimensional feature set of a real-time development record at a current time, preliminarily evaluate the real-time development record by the multi-objective joint model, extract weight coefficients of any feature at the current time, analyze expected changes of the features, and give a risk warning for expected conflicts.

[0099] The multi-dimensional feature analysis module comprises a high-dimensional feature generation unit and a feature conflict analysis unit.

[0100] The high-dimensional feature generation unit is configured to collect multi-dimensional data of the application development process, generate corresponding application development records, and obtain a high-dimensional feature set of the application development records.

[0101] The feature correlation analysis module comprises a feature correlation analysis unit and a weight coefficient setting unit.

[0102] The feature correlation analysis unit is configured to analyze the correlation between different features based on the conflict between the different features. The weight coefficient setting unit is configured to set a weight coefficient for each feature according to the correlation between the different features.

[0103] The model weight analysis module comprises a conflict proportion analysis unit and a joint model construction unit.

[0104] The conflict proportion analysis unit is configured to analyze the corresponding data of any feature in each application development record to obtain the conflict proportion of the feature. The joint model construction unit is configured to construct a multi-objective joint model by combining the corresponding weight coefficients to evaluate the development of any application development record.

[0105] The real-time anomaly recognition module comprises a preliminary evaluation analysis unit and a target conflict prediction unit.

[0106] The preliminary evaluation analysis unit is configured to obtain a high-dimensional feature set of a real-time development record at the current time and perform preliminary evaluation through the multi-objective joint model. The target conflict prediction unit is configured to extract the weight coefficient of any feature at the current time, analyze the expected change of each feature, and perform risk warning on the expected conflict.

[0107] It is apparent to those skilled in the art that the present application is not limited to the details of the foregoing exemplary embodiments, and that the present application can be implemented in other concrete forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all respects as illustrative and not restrictive, and the scope of the present application is defined by the appended claims rather than the foregoing description, and it is intended to encompass all changes falling within the meaning and scope of equivalents of the claims. Any reference signs in the claims should not be considered as limiting the claims involved.

Claims

1. A method for managing application development data based on an AI large model, the method comprising: The management method comprises the following steps: Step S100: multi-dimensional data collection is performed on the application development process, corresponding application development records are generated, and a high-dimensional feature set of the application development records is obtained; and conflicts between features are identified by mining multi-dimensional data in any application development record; Step S200: based on the conflicts between different features, the association between different features is analyzed; and a weight coefficient is set for each feature according to the association between different features; Step S300: the corresponding data of any feature in each application development record is analyzed to obtain the conflict proportion of any feature; and a multi-objective joint model is constructed by combining the corresponding weight coefficient to evaluate the development of any application development record; Step S400: a high-dimensional feature set of real-time development records at the current time is obtained, preliminary evaluation is performed by the multi-objective joint model; the weight coefficient of any feature at the current time is extracted, the expected change of each feature is analyzed, and a risk warning is given for the expected conflicts; The step S300 comprises the following steps: Step S301: arbitrarily select an i-th conflict feature in the application development record, obtain a development data set of the dimension where the i-th conflict feature is located, if the development data set is structured data and there is a numerical range corresponding to the data field, then preset a conflict numerical range (d min ,d max ) for the i-th conflict feature; obtain the numerical range (d1 i ,d2 i ) of the i-th conflict feature in the selected application development record, if d1 i <d min <d2 i <d max , obtain a first conflict proportion η1 = (d2 i -d min ) / (d max -d min ), if d min <d1 i <d2 i <d max , obtain a first conflict proportion η1 = (d2 i -d1 i ) / (d max -d min ), if d min <d1 i <d max <d2 i , obtain a first conflict proportion η1 = (d max -d1 i ) / (d max -d min ); Step S302: If the development data set does not exist a numerical range, obtaining the number of the same features v existing in the development data set, and setting the number of the features v in the development data set total , obtaining the second conflict ratio η2=v / v total ; Step S303: Set the feature label of the i-th conflicting feature as Flag. i If the i-th conflict feature has a numerical range, then Flag i =1, if the i-th conflicting feature does not have a numerical range, then Flag =1. i =0; Construct a multi-objective joint model to evaluate any application development record: wherein η1 i is the first conflict proportion of the i-th conflict feature, η2 i is the second conflict proportion of the i-th conflict feature; and the evaluation value Z of the selected application development record is obtained by inputting each conflict feature in the selected application development record into the multi-objective joint model. 2.The AI large model-based application development data management method of claim 1, wherein: The step S100 comprises the following steps: Step S101: a plurality of data collection methods are preset to collect data of the application development process to obtain a plurality of development data sets of different dimensions, wherein one data collection method corresponds to one development data set of one dimension; and the development data sets of the plurality of dimensions are summarized to generate one application development record; Step S102: any development data set of any dimension in any application development record is selected, if the development data in the development data set is structured data, the data type and data field of the development data are extracted to obtain a feature group, if the development data of the selected dimension is unstructured data, the development data is semantically analyzed by using an AI large model to extract a plurality of features to obtain a feature group; and the feature groups of each dimension are summarized to obtain a high-dimensional feature set of the selected application development record; Step S103: a conflict type database is constructed in advance, the conflict type database comprises a plurality of conflict types, and each conflict type is assigned a corresponding conflict feature set; a feature group is selected from the high-dimensional feature set, and a conflict type is selected; the selected feature group and the conflict feature set of the selected conflict type are compared, if there is a same feature, the selected feature group is set as the same feature group; Step S104: count the number of the same feature groups as n1, set the number of conflict features of the conflict feature set as N, calculate the proportion of the number of the same feature groups as a = n1 / N, and preset a number proportion threshold a th If a ≥ a th , the conflict type corresponding to the conflict feature set is selected as one of the conflict types of the application development record. 3.The AI large model-based application development data management method of claim 2, wherein: The step S200 comprises the following steps: Step S201: a conflict type in any application development record is selected, two conflict features are selected from the selected conflict type, and are set as the ith conflict feature and the jth conflict feature, respectively, to establish an association set R(i,j) between the two conflict features; Step S202: Obtain the number of conflict types containing the associated set R(i,j) in each conflict type, which is m1 (i,j) , set the total number of conflict types as M1, and calculate the proportion of the number of conflicts of the associated set R(i,j) as β1=m1 (i,j) / M1; count the number of records containing the associated set R(i,j) in each application development record as m2 (i,j) , set the number of application development records as M2, and calculate the conflict occurrence frequency of the associated set R(i,j) as β2=m2 (i,j) / M2; according to the formula: G(i,j)=β1×β2; The association value G(i,j) between the ith conflict feature and the jth conflict feature is calculated; Step S203: the ith conflict feature is selected, the association values between the ith conflict feature and the remaining conflict features are obtained, and the expected change of each feature is analyzed according to the formula: Wherein e is the total number of conflict characteristics; the feature coefficient P of the ith conflict characteristic is calculated i ; the feature coefficient of each conflict characteristic is obtained, and the formula is: The weight coefficient Q of the ith conflict feature is calculated i . 4.The AI large model-based application development data management method of claim 3, wherein: The step S400 comprises the following steps: Step S401: collecting development data under each dimension in the real-time development record, and respectively obtaining a feature group of any dimension, generating a high-dimensional feature set of the real-time development record, comparing the high-dimensional feature set with a conflict feature set of each conflict type to obtain a same feature set; randomly selecting a u-th feature from the same feature set, and obtaining a weight coefficient of the u-th feature at a current time point as (Q u ) now ; Step S402: obtain all development data corresponding to the u-th feature, if the development data of the u-th feature has a numerical range, then obtain the first conflict proportion η1 of the u-th feature u , if there is no numerical range, then obtain the second conflict proportion η2 u ; input all features into the multi-objective joint model to obtain the real-time evaluation value Z of the real-time development record now ; Step S403: preset an abnormality evaluation threshold Z th , if Z now ≥ Z th , an abnormality prompt is given to the real-time development record, if Z now < Z th , the number of features contained in the same feature set is n now , according to the formula: wherein Select() is a selection function, if the development data of the u-th feature exists a value range, then Select(u)=η1 u , otherwise, Select(u)=η2 u ; the expected weight coefficient (Q u ) now of the u-th feature is calculated; if (Q u ) now < (Q u ) now , then the u-th feature is set as an abnormal feature; Step S404: count the number of abnormal features in the same feature set as n ’ , and obtain the abnormal proportion τ = n ’ / n now ; if τ ≥ τ th , a risk warning is given to the real-time development record, wherein τ th is an abnormal proportion threshold.

5. An application development data management system for performing an AI large model-based application development data management method according to any one of claims 1-4, characterized by: The management system comprises a multi-dimensional feature analysis module, a feature correlation analysis module, a model weight analysis module and a real-time anomaly identification module; The multi-dimensional feature analysis module is configured to collect multi-dimensional data of the application development process, generate corresponding application development records, and obtain a high-dimensional feature set of the application development records; and mine the multi-dimensional data in any application development record to identify conflicts between features. The feature correlation analysis module is configured to analyze correlations between different features based on the conflicts between the different features; and set weight coefficients for the features according to the correlations between the different features. The model weight analysis module is configured to analyze corresponding data of any feature in each application development record to obtain a conflict proportion of the feature; and construct a multi-objective joint model to evaluate the development of any application development record by combining the weight coefficients. The real-time anomaly identification module is configured to obtain a high-dimensional feature set of a real-time development record at a current time, preliminarily evaluate the real-time development record by the multi-objective joint model, extract weight coefficients of any feature at the current time, analyze expected changes of the features, and perform risk warning on expected conflicts.

6. The application development data management system of claim 5, wherein: The multi-dimensional feature analysis module comprises a high-dimensional feature generation unit and a feature conflict analysis unit. The high-dimensional feature generation unit is configured to collect multi-dimensional data of the application development process, generate corresponding application development records, and obtain a high-dimensional feature set of the application development records. The feature conflict analysis unit is configured to mine the multi-dimensional data in any application development record to identify conflicts between features.

7. The application development data management system of claim 5, wherein: The feature correlation analysis module comprises a feature correlation analysis unit and a weight coefficient setting unit. The feature correlation analysis unit is configured to analyze correlations between different features based on the conflicts between the different features; and the weight coefficient setting unit is configured to set weight coefficients for the features according to the correlations between the different features.

8. The application development data management system of claim 5, wherein: The model weight analysis module comprises a conflict proportion analysis unit and a joint model construction unit. The conflict proportion analysis unit is configured to analyze corresponding data of any feature in each application development record to obtain a conflict proportion of the feature; and the joint model construction unit is configured to construct a multi-objective joint model to evaluate the development of any application development record by combining the weight coefficients.

9. The application development data management system of claim 5, wherein: The real-time anomaly identification module comprises a preliminary evaluation analysis unit and a target conflict prediction unit. The preliminary evaluation analysis unit is configured to obtain a high-dimensional feature set of a real-time development record at a current time, preliminarily evaluate the real-time development record by the multi-objective joint model; and the target conflict prediction unit is configured to extract weight coefficients of any feature at the current time, analyze expected changes of the features, and perform risk warning on expected conflicts.

Images