Mainframe Data Element Mapping Using Graph Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for understanding the structure of data elements in mainframe program code are computationally expensive and complex, especially in parent-child impact analysis, due to the lack of explicit memory mapping and reliance on sequential parsing techniques.
Innovation Solution
A graph neural network-based model is employed to learn latent embeddings of data elements through parental and operation graphs, using message passing techniques, and is trained to predict memory mappings between variables, enabling efficient understanding of data element structures and interactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sequential parsing techniques are used to analyze parent-child variable relationships, then the analysis can be performed step-by-step, but the computational cost becomes extremely expensive and the process becomes complex when the structure goes deeper
Solution Approach 1:
The patent replaces the sequential mechanical parsing process with a neural network-based system. The parser first extracts variable relationships and builds a dependency graph, then a neural network model processes this graph to predict parent-child impact relationships in parallel, eliminating the sequential computational bottleneck while maintaining analysis accuracy.
Solution Approach 2:
The patent transforms the sequential one-dimensional analysis into a multi-dimensional approach by building a dependency graph that captures variable relationships in multiple dimensions simultaneously. This graph structure allows the neural network to process all relationships in parallel across different dimensions, dramatically improving computational efficiency.
2Loss of information
If parsing techniques are applied to capture data element relationships, then the structure can be understood, but the technique cannot solve the problem because memory mapping between variables is not explicitly available as physical statements in the code
Solution Approach 1:
The patent introduces an intermediary neural network model that acts as a bridge between the parsed code structure and the implicit memory mappings. The parser extracts explicit relationships and builds a dependency graph, then the neural network infers implicit memory mappings that are not directly visible in the code, completing the picture of data element relationships.
Solution Approach 2:
The patent performs preliminary parsing to extract all explicit variable relationships and build a comprehensive dependency graph before applying the neural network. This preliminary action captures all available structural information, which the neural network then enhances by inferring implicit relationships, ensuring no information is lost.
3Ease of operation
If the depth of parent-child analysis is solved one after the other from super-parent variable to elementary, then the sequential process can be managed, but the computational expense increases significantly
Solution Approach 1:
The patent transforms the static sequential process into a dynamic parallel process. The neural network model dynamically processes the entire dependency graph simultaneously, adapting to the structure's complexity without following a fixed sequential order, thereby maintaining ease of operation while dramatically improving computational speed.
Solution Approach 2:
The patent segments the analysis into two distinct phases: (1) parsing and dependency graph construction, and (2) neural network-based impact prediction. This segmentation allows the complex task to be divided into manageable components that can be processed efficiently, with the neural network handling the computationally intensive parent-child impact analysis in parallel.
Data Source
AI summary
Most of the existing production applications in different domains are still running on. Mainframe applications in production receive data from various resources and process these data within. Understanding the structure of input data and output data is extremely important. A method and system for machine learning based understanding of a plurality of data elements in a mainframe program code has been provided. The method discloses a machine learning model that understands the structure of data elements in a Mainframe program code. The model considered is a graph neural network based architecture model. The disclosed method replicates memory mapping happening in the application program environment. The method understands the structure of the data element and the impact created by each data element on other data elements in the application and interfacing applications. The disclosed solution serves as a building block in problems such as code translation, reverse engineering etc.