CN-122020403-A - Dynamic modeling and self-learning optimization method in intelligent AI data insight analysis

CN122020403ACN 122020403 ACN122020403 ACN 122020403ACN-122020403-A

Abstract

The invention provides a dynamic modeling and self-learning optimization method in intelligent AI data insight analysis, which comprises the steps of obtaining multi-source heterogeneous data, preprocessing the multi-source heterogeneous data, identifying the multi-source heterogeneous data through an anomaly detection model to obtain an anomaly identification result, carrying out dynamic modeling and self-learning optimization on the anomaly detection model, wherein the dynamic modeling is carried out through so-called nerve architecture search, the self-learning is carried out through deep learning and reinforcement learning, and carrying out anomaly repair according to the anomaly identification result. Through the capabilities of dynamic modeling, self-learning optimization and the like, intelligent insight analysis of service data and operation and maintenance logs is realized, and the running stability of the system and the accuracy of the service data are ensured.

Inventors

ZHAO XIANMING
XIANG YANG
LIN YUN

Assignees

北京红山信息科技研究院有限公司

Dates

Publication Date: 20260512
Application Date: 20260120

Claims (8)

1. A dynamic modeling and self-learning optimization method in intelligent AI data insight analysis is characterized by comprising the following steps: Acquiring multi-source heterogeneous data, and preprocessing the multi-source heterogeneous data; Identifying the multi-source heterogeneous data through an anomaly detection model to obtain an anomaly identification result; The anomaly detection model carries out dynamic modeling and self-learning optimization, wherein the dynamic modeling carries out modeling through so-called neural architecture searching, and the self-learning is realized through deep learning and reinforcement learning; And carrying out abnormality repair according to the abnormality identification result.
2. The method of claim 1, wherein the step of determining the position of the substrate comprises, The multi-source heterogeneous data comprises service data, log data and system data transmitted by different service platforms.
3. The method of claim 1, wherein the step of determining the position of the substrate comprises, The method comprises the steps of preprocessing multi-source heterogeneous data through an ETL engine and a standardized tool, wherein the preprocessing comprises missing value filling, repeated elimination, normalization, feature coding and dimension reduction.
4. The method of claim 1, wherein the step of determining the position of the substrate comprises, The anomaly detection model is constructed by adopting one or a combination of a plurality of clustering algorithms, classification algorithms, similarity calculation and deep neural networks.
5. The method of claim 1, wherein the step of determining the position of the substrate comprises, The self-learning optimization process of the anomaly detection model comprises the following steps: Automatically acquiring feedback information, generating training samples according to the feedback information, automatically self-learning the abnormal detection models according to the number of the training samples or the performance of the abnormal detection models to obtain new models, performing isolated operation on the new models, comparing the performance between the new models and the abnormal detection models, and replacing the abnormal detection models according to the new models.
6. The method of claim 1, wherein the step of determining the position of the substrate comprises, The automatic modeling process of the anomaly detection model comprises the following steps: constructing related constraint conditions aiming at a deep neural network in the anomaly detection model, wherein the constraint conditions comprise constraints of network depth, parameter quantity and operation type; Constructing a super network containing candidate operations with replaceable architecture, wherein in the super network, each candidate operation is provided with corresponding architecture weight, wherein the super network is in a computational graph structure, wherein the computational graph contains feature graphs generated by different candidate operations as nodes, the candidate operations are taken as edges, and different replacement architectures in the candidate operations are provided with corresponding network weights; and sequentially or alternately updating the network weights and the architecture weights of the super network, and selecting corresponding candidate operations and corresponding network weights as final architecture and corresponding network weights according to the final architecture weights.
7. The method of claim 1, wherein the step of determining the position of the substrate comprises, And performing self-learning optimization circulation of the anomaly detection model by deploying a simulation environment.
8. A dynamic modeling and self-learning optimization system in intelligent AI data insight analysis, characterized by performing the method of any of the preceding claims 1-7.

Description

Dynamic modeling and self-learning optimization method in intelligent AI data insight analysis Technical Field The invention relates to the technical field of data analysis, in particular to a dynamic modeling and self-learning optimization method in intelligent AI data insight analysis. Background With the deep digital transformation of enterprises, the data size and complexity are greatly increased, and the traditional data analysis method faces serious challenges in terms of dynamics, intellectualization and automation. The current system has the problems of rigidifying data processing flow, single abnormality detection means, lagging model updating, dependence on manpower and the like, and is difficult to meet the requirements of enterprises on real-time, accurate and self-adaptive data insight. Therefore, an intelligent analysis method for intelligent restoration comprehensive AI data, which can integrate dynamic modeling, self-learning optimization and intelligent restoration, is urgently needed to cope with complex and changeable business environments, including system and data stabilization. Disclosure of Invention In view of this, the present invention proposes a dynamic modeling and self-learning optimization method in intelligent AI data insight analysis to solve the problems existing in the prior art. In order to achieve the above purpose, the present invention provides a dynamic modeling and self-learning optimization method in intelligent AI data insight analysis, comprising: Acquiring multi-source heterogeneous data, and preprocessing the multi-source heterogeneous data; Identifying the multi-source heterogeneous data through an anomaly detection model to obtain an anomaly identification result; The anomaly detection model carries out dynamic modeling and self-learning optimization, wherein the dynamic modeling carries out modeling through so-called neural architecture searching, and the self-learning is realized through deep learning and reinforcement learning; And carrying out abnormality repair according to the abnormality identification result. Optionally, the heterogeneous multi-source data includes service data, log data and system data transmitted by different service platforms. Optionally, preprocessing is performed on the multi-source heterogeneous data through an ETL engine and a normalization tool, wherein the preprocessing comprises missing value filling, repeated elimination, normalization, feature coding and dimension reduction processing. Optionally, the anomaly detection model is constructed by adopting one or a combination of a plurality of clustering algorithms, classification algorithms, similarity calculation and deep neural networks. Optionally, the self-learning optimization process of the anomaly detection model includes: Automatically acquiring feedback information, generating training samples according to the feedback information, automatically self-learning the abnormal detection models according to the number of the training samples or the performance of the abnormal detection models to obtain new models, performing isolated operation on the new models, comparing the performance between the new models and the abnormal detection models, and replacing the abnormal detection models according to the new models. Optionally, the automatic modeling process of the anomaly detection model includes: constructing related constraint conditions aiming at a deep neural network in the anomaly detection model, wherein the constraint conditions comprise constraints of network depth, parameter quantity and operation type; Constructing a super network containing candidate operations with replaceable architecture, wherein in the super network, each candidate operation is provided with corresponding architecture weight, wherein the super network is in a computational graph structure, wherein the computational graph contains feature graphs generated by different candidate operations as nodes, the candidate operations are taken as edges, and different replacement architectures in the candidate operations are provided with corresponding network weights; and sequentially or alternately updating the network weights and the architecture weights of the super network, and selecting corresponding candidate operations and corresponding network weights as final architecture and corresponding network weights according to the final architecture weights. Optionally, the self-learning optimization cycle of the anomaly detection model is performed by deploying the simulation environment. On the other hand, the invention provides a dynamic modeling and self-learning optimization system in intelligent AI data insight analysis, which is used for executing the method. Compared with the prior art, the invention has the beneficial effects that, The deep AI data insight analysis service is based on artificial intelligence, big data analysis and intelligent operation and maintenance technology, and aims to sol