Özet:

Artificial intelligence-based studies have a great interest in creating decision support systems in business sectors, producing effective outputs, increasing system efficiency and providing cost-effective solutions. Especially in the development of the innovation process, the acceleration of the innovation process and its evolution into the target area, innovations are provided with artificial intelligence-based studies. In the realization of these innovations, data has a critical meaning for artificial intelligence-based studies. It plays an important role in the functionalization of models trained through algorithms by computers or special machines. However, insufficient data access, legal regulations, ethical rules, confidentiality procedures, privacy, data sharing limitations and cost; are major obstacles to revealing the potential of data. To overcome these obstacles, the synthetic data generation approach is preferred. But, since there is no standard framework for the synthetic data generation approach, research on the development of new and current approaches continues. In this study, a new synthetic data generation approach based on a genetic algorithm is proposed. In this direction, data diversity has been increased and a new generation has been obtained by using the crossover and mutation genetic operators adapted to the target dataset to produce artificial data in the dynamics of the original dataset. Then, the category definition of the artificial samples in this generation was done using the cost function component of the genetic algorithm. In the last stage, 6 different machine learning classifiers were used to measure the success of the similarity of the artificial data produced to the original data. The maximum sensitivity criterion was obtained as 100% with the Support Vector Machine classifier on the enriched dataset. This indicates that educational success tends to be in the positive direction in proportion to the increasing number of data.

Anahtar Kelimeler: