Score Model Normalization via Directed Acyclic Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex score models with many characteristics and bins require inefficient evaluation processes, leading to redundant comparisons and increased computational complexity.
Innovation Solution
Normalization of score models into directed acyclic graphs (DAGs) followed by transformation into structured rules language (SRL) programs and then into concurrent, class-based, object-oriented programming languages like JAVA, C, or COBOL, allowing for efficient evaluation and elimination of redundant evaluations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a straightforward evaluator exhaustively evaluates all bins in each characteristic, then complete evaluation of the score model is achieved, but computational complexity increases to O(C*B)
Solution Approach 1:
The score model is segmented into a directed acyclic graph where characteristics are mapped to intermediate nodes and score values to leaf nodes. This segmentation allows the evaluation to be divided into independent paths, enabling efficient traversal without exhaustive evaluation of all bins.
Solution Approach 2:
The evaluation approach transitions from a flat exhaustive search across all bins to a hierarchical tree structure. By adding the dimension of node levels in the DAG, the system can prune branches and avoid redundant evaluations while maintaining complete evaluation of valid paths.
2Adaptability or versatility
If the score model is represented with many characteristics and bins, then detailed scoring capability is achieved, but evaluation efficiency decreases
Solution Approach 1:
The score model is pre-processed and normalized into a directed acyclic graph structure before evaluation. This preliminary transformation organizes the many characteristics and bins into a hierarchical format that enables efficient traversal during actual evaluation, maintaining scoring detail while improving productivity.
Solution Approach 2:
Intermediate nodes representing characteristics serve as mediators between the root and leaf nodes. These intermediaries organize the evaluation process by grouping related bins and score values, allowing efficient traversal through the hierarchical structure without evaluating all possibilities exhaustively.
3Stability of the object's composition
If bin ranges are not normalized and checked for gaps and overlaps, then original score model structure is preserved, but evaluation redundancy increases
Solution Approach 1:
The bin ranges are normalized and checked for gaps and overlaps as a preliminary action during model construction. This pre-processing step identifies and eliminates redundant evaluations by merging overlapping bins and filling gaps, reducing evaluation redundancy while preserving the essential model structure.
Solution Approach 2:
Paths of the directed acyclic graph terminating with equivalent leaf nodes are merged. This merging operation consolidates redundant evaluation paths, reducing the total number of evaluations required while maintaining the same scoring outcomes and model structure.
Data Source
AI summary
Data is received that characterizes a score model. Thereafter, the score model is normalized by transforming it into a directed acyclic graph. The directed acyclic graph is then transformed into a structured rules language program. The structured rules language program is then transformed into a program using a concurrent, class-based, object-oriented computer programming language (e.g., JAVA, C, COBOL, etc.). Related apparatus, systems, techniques and articles are also described.