Development of Explanatory Artificial Intelligence Techniques for Automatic Detection of Factors Affecting Different Population Levels of Almond Leaf Bee Cimbex quadrimaculata Müller, 1766 (Hymenoptera: Cimbicidae)

Abstract

In this study; data sets of factors affecting different population levels of almond leaf bee
Cimbex quadrimaculata (Hymenoptera: Cimbicidae) were taken from the almond fields of Nimri village
of Keban district of Elazığ province and Eğil district of Diyarbakır province in 2020, 2021 and 2022. In
the study; the independent variables affecting the population levels of the pest were altitude, temperature,
humidity, wind intensity, tree age, tree height, tree crown projection, and the dependent variables were
the larval level. The success classification values of Random Forest and Decision Tree algorithms were
used. Larval numbers were graded as slightly contaminated, moderately contaminated, contaminated and
highly contaminated with scale values of 0-3, 3-10, 10-30 and over 30 larvae per tree. These data were
then analysed using decision tree algorithms from machine learning classification models. The analyses
were interpreted according to Random Forest and Decision Tree algorithms in JASP 0.19.1 version
programme. According to the results of the study; it was determined that the most important factor
affecting the larval level in cultivated almonds in Eğil district was temperature, followed by humidity.
The least influencing factors were tree height and tree crown projection. In the wild almond trees of Eğil
District, humidity was the most important factor affecting the larval level, followed by temperature. In
the cultivated almond trees of Keban District of Elazığ Province, it was observed that the factor affecting
the larval level most was temperature, followed by humidity. In Wild Almond Trees of Keban District of
Elazığ Province, it was observed that the factor affecting the larval level most was temperature, followed
by humidity. In all parameters, the factors that least affect the larval population level were removed from
the application in the data analysis programme and the Random Forest algorithm was applied again. In
this application, the independent variables were altitude, temperature, humidity, wind intensity and tree
age. As a result of the re-application of the algorithm, it was observed that the accuracy value and other
values supporting this value increased. In the comparison of wild and cultivated almond classification
values between locations, it was determined that the factors affecting the larvae level most in the
application of success classification value determination of decision trees of wild almond trees in Eğil
district of Diyarbakır province and Keban district of Elazığ province were determined as temperature, humidity, altitude, tree age and wind intensity, respectively. In the application of determination of
success classification value of decision trees of cultivated almond trees in Eğil district of Diyarbakır
province and Keban district of Elazığ province, it was determined that the factors affecting the larval
level most were temperature, humidity, tree age, altitude and tree crown projection, respectively. In the
application of decision trees success classification value determination for cultivated almond trees in Eğil
district of Diyarbakır province and Keban district of Elazığ province, it was determined that the factors
affecting the larval level most were temperature, humidity, tree age, altitude and tree crown projection,
respectively. In the success classification of wild and cultivated almonds between locations (Keban/Eğil),
it was determined that the factors affecting the larvae level most in the decision tree success classification
value determination application were temperature, humidity, tree age, altitude and wind intensity,
respectively. As a result of all algorithm applications, temperature and humidity were found to be the
factors affecting the larval level the most. The factors affecting the larval level the least were found to be
tree height and tree crown projection.