Self-Adaptive Genetic Programming for Manufacturing Big Data Analysis

Abstract

While black-box-based machine learning algorithms have high analytical consistency in manufacturing big data analysis, those algorithms experience difficulties in interpreting the results based on the manufacturing process principle. To overcome this limitation, we present a Self-Adaptive Genetic Programming (SAGP) for manufacturing big data analysis. In Genetic Programming (GP), the solution is expressed as a relationship between variables using mathematical symbols, and the solution with the highest explanatory power is finally selected. These advantages enable intuitive interpretation on manufacturing mechanisms and derive manufacturing principles based on the variables represented by formulas. However, GP occasionally has trouble adjusting the balance between high accuracy and detailed interpretation due to an incommensurable symmetry of the solutions. In order to effectively handle this drawback, we apply the self-adaptive mechanism into GP for managing crossover and mutation probabilities regarding the complexity of tree structure solutions in each generation. Our proposed algorithm showed equal or superior performance compared to other machine learning algorithms. We believe our proposed method can be applied in diverse manufacturing big data analytics in the future.