Ανάπτυξη στοχαστικών αλγορίθμων και μοντέλων μηχανικής μάθησης για την εξέλιξη πληθυσμών: Εφαρμογή στο δάκο της Ελιάς
Date Issued
February 15, 2022
Type
Διδακτορική Διατριβή
Abstract
The olive fruit fly is a parasitic insect that depends on the fruit of the olive tree for the perpetuation of its species, causing great damage, if its population is left unchecked, to olive products, such as olive oil and table olives. The presence of the insect has been recorded in all olive-producing countries and due to its dependence on climatic conditions (temperature, humidity) the total number of generations that result varies from region to region. In this context, with regard to stochastic algorithms, the software for simulating the spatiotemporal evolution of the fly population based on actual measurements in the field is initially presented. Then, the development and comparison of three models of dispersion of the olive fruit fly is described based on the fruiting of the olive grove and finally, the influence of the presence of microclimates on the evolution of the population of the olive fruit fly through simulation, using the aforementioned software, is studied. Continuing, with regard to the use of machine learning methods, initially, the use of environmental conditions to predict the next trap measurement is studied and the performance of four feature vectors is compared with the use of various machine learning algorithms. Also, the development of a model for the automatic identification of the olive fruit fly in images of the contents of the McPhail trap, is presented. Finally, the use of machine learning methods for the identification of microclimates is being studied, which would help in the prediction of outbreaks of infestation of the olive grove by the olive fruit fly depending on the time of year and the climatic characteristics of each microclimate. Finally, the conclusions are drawn on the performance of the aforementioned experiments and possible points that any future research could focus on.
Subjects