Thesis Type: Postgraduate
Institution Of The Thesis: Eskisehir Osmangazi University, FEN BİLİMLERİ ENSTİTÜSÜ, Fen Bil.Enst.Md.Lüğü, Turkey
Approval Date: 2023
Thesis Language: Turkish
Student: BEYZA NUR DOĞAN
Principal Supervisor (For Co-Supervisor Theses): Coşkun Güçlü
Co-Supervisor: Yasin Altay
Abstract:
Tropinota (Epicometis) hirta (Poda) fruit trees, strawberries, roses, wheatgrass feed
on the flowers of plants such as wheatgrass, eat male and female organs. Sometimes they
feed on shoots, leaves or even fruit. Since their ability to fly is high, they migrate to different
plants and continue their damage. As a result, damaged flowers cannot set fruit. They are
important pests related to agriculture. This study aimed to develop a morphometric
characterization method to determine the gender of the beetle species T. hirta using treebased data mining algorithms based on certain taxonomic characteristics. A total of 516
female and 504 male individuals were used in the study, and their morphometric features
were used as independent variables. The dependent variable for the data mining algorithms
was the gender of the individuals, defined as a binary response variable. The performance of
the algorithms- Chi-Square Automatic Interaction Detector (CHAID), Exhaustive CHAID,
Classification and Regression Tree (CART), and Quick Unbiased, Efficient Statistical Tree
(QUEST)- in differentiating the sexes was evaluated using several metrics such as the area
under the ROC curve, Matthews correlation coefficient, and Cohen's Kappa coefficient. The
results showed that all algorithms were able to successfully distinguish the sexes of the
individuals with high accuracy. The first branching of the tree structure generated by the
algorithms was based on the Filagellum Length (FU) of the individuals, with specific cutoff points identified for each algorithm. The findings of this study demonstrated the potential
of using morphometric characteristics and tree-based data mining algorithms to determine
the sexes of T. hirta individuals, and suggested that such methods can be integrated into
taxonomic studies to improve the accuracy of species identification.