How multi‐objective genetic algorithms handle lack of data, sparse data and excess data: evaluation of some recent case studies in the materials domain

Abstract

An overview of multi-objective optimization and the associated concept of Pareto-optimality are elucidated in detail, keeping the biologically inspired genetic algorithms in perspective. The effective role of the genetic algorithms in handling three different kinds of data driven models where the decision has to be made from (i) no data (ii) excess data or (iii) sparse data are elaborated through three materials engineering applications, where other strategies like inverse modeling, neural network and data mining have worked in tandem with the multi-objective genetic algorithms. 

Modelling Power Consumption in Ball-End Milling Operations

“…Using a multiobjective genetic algorithm and a Pareto frontier, the most suitable neural network is identified through an Akaike approach to avoid the problems of overfitting and underfitting and obtain the best forecasting capabilities. This technique is applied in several different industrial fields [20][21][22][23][24] and would have been a good alternative to the neural network model described in this article, if it had presented overfitting or underfitting problems. Furthermore, in [20], Pettersson et al combined genetic algorithms, data mining, and neural networks in tandem to construct generic predictive models.…”

“…However, it would have been an interesting option if the number of experiments carried out was not enough or the data collected led to less accurate power predictions. Further information on multiobjective genetic algorithms for handling situations where data is lacking, sparse, or excessive can be found in [21]. In this article, overtraining problems were avoided by applying the early stopping method as described in the previous section.…”