Recommendation Engine Integrating DEA and Machine Learning for DMU Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current B2B analysis tools separately evaluate Decision-Making Units (DMUs) using customer key performance indices (KPIs) and benchmarks, lacking a meaningful connection between the two methods, which hinders providing data-driven, prescriptive recommendations for improving efficiency and savings.
Innovation Solution
A recommendation engine that integrates machine learning algorithms with data envelopment analysis (DEA) to identify key KPIs, allowing for continuous monitoring and near real-time recommendations, enabling DMUs to improve efficiency by moving underperforming units towards an efficient frontier based on peer performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If separate evaluation methods (KPIs and benchmarks) are used for DMUs, then evaluation coverage is comprehensive, but meaningful connection and integrated recommendations are lacking
Solution Approach 1:
The patent combines separate KPI evaluation and benchmark evaluation into a unified DEA framework. The system integrates multiple input parameters (e.g., procurement spend, number of employees) and output parameters (e.g., savings achieved) to create a comprehensive efficiency score that simultaneously considers both KPI performance and benchmark comparisons, eliminating the need for separate evaluation methods
Solution Approach 2:
The DEA model serves multiple functions simultaneously: it evaluates individual DMU performance against KPIs, compares DMUs against each other through benchmarking, identifies efficient frontier units, and generates prescriptive recommendations. This single unified system replaces multiple separate evaluation tools while maintaining comprehensive assessment capabilities
2Ease of manufacture
If traditional separate evaluation tools are used, then implementation is simple, but prescriptive recommendations for improvement are not provided
Solution Approach 1:
The system provides continuous feedback by calculating efficiency scores for each DMU, comparing them against the efficient frontier, and generating prescriptive recommendations. The DEA model identifies specific input parameters that should be adjusted and recommends target values based on performance of efficient peer DMUs, enabling actionable efficiency improvements
Solution Approach 2:
The system pre-calculates efficient frontier boundaries and identifies optimal target states for underperforming DMUs before implementation. By determining the efficient frontier in advance and pre-computing recommended adjustments to input parameters, the system provides ready-to-implement prescriptive recommendations that guide DMUs toward optimal performance
3Device complexity
If static evaluation methods are used, then computational complexity is low, but near real-time monitoring and recommendations are not achieved
Solution Approach 1:
The system transitions from static to dynamic evaluation by continuously updating efficiency scores as new data becomes available. The DEA model can be re-executed with updated input parameters to reflect current DMU performance, enabling near real-time monitoring. The efficient frontier and peer comparisons are dynamically recalculated to provide current prescriptive recommendations
Data Source
AI summary
In an example embodiment, a recommendation engine provides recommendations as to how decision-making units (DMUs) can improve efficiency, or savings can utilize machine learning algorithms and data envelopment analysis (DEA). DEA is a linear programming methodology, and is used in the example embodiment to identify one or more key performance indices (KPIs) that are most important to a DMU.