An evaluation of ecological data with machine learning techniques

Ecology is a scientific discipline that aims to allow natural scientists working on biotic and abiotic systems to study, understand, and inference complex relationships related to these systems. Machine learning techniques could be more profitable than statistical methods at the point where these complex relationships can be understood and prospective predictions can be made. With Machine Learning, the datasets which are nonlinear, high volume, have high amount attributes or can contain a large number of missing data that belong to ecological systems can be passed through the artificial learning process by using training algorithms and predictions can be made. In this thesis, two datasets are evaluated in order to understand how Artificial Neural Networks, K-NN, K-Means, Naïve Bayes, and Decision Tree Classification (ID3) methods which are widely used in Machine Learning can be used in ecological data. It is also aimed to have a better understanding of these methods. Five methods were tested in a designed and developed experiment set. In the testing processes, it has been tried to obtain the best performance values through a selected validation method by means of different parameters. In result, it was observed that the performance values were similar to those of other algorithms commonly used in the literature.