Fitness value curves prediction in the evolutionary process of genetic algorithms

The evolutionary process of a Genetic Algorithm (GA) depends on several factors, the initialization parameters are some of them [4]. One way to inspect the evolutionary process of a GA is to analyze the maximum, average, and minimum fitness of each generation. This article focused on the use of a machine learning model to predict the maximum, average, and minimum fitness values during the evolutionary process of a GA, and this, only with the knowledge of its initialization parameters. In order to accomplish this goal, a Random Forest model was trained with data from different GA executions for a given problem. The prediction process was performed with a very promising performance, where the challenge of predicting the evolutionary process of a GA was fulfilled with low error rates. This approach opens up several opportunities for advances in the segment, and in a way, contributes to the investigation and improvement of GAs, as well as demonstrating the importance of monitoring and storing the information generated during their evolutionary process.