Description
The olive fruit fly is a pest that attacks olives in olive tree plantations that results in lower quality olive oil that is produced from such olives. The presence of olive fruit fly is highly dependent on meteorological factors. Knowing in advance when the olive fruit fly will attack the olive tree plantation can enable olive oil producers (plantation owners) to take appropriate actions.
The goal of this study was to develop a (mathematical) model that will help olive oil producers (plantation owners) to predict the presence/absence and amount of the olive fruit fly in their plantation using past meteorological and infestation data (no. of olive fruit flies) for the years 2006 to 2012 and to integrate this model in a web-based application.
For the development of the model we used the classical data mining approach consisting of six phases: problem understanding, data understanding, data pre-processing, modeling, evaluation and interpretation of the models, deployment. For the first two phases and the evaluation and interpretation phase we got help from biologists, experts in the olive fruit fly.
Several machine learning algorithms were tried in the modeling phase and decision trees turned out to be the best performing algorithm with prediction accuracy 73.08% where individual prediction rules reached more than 90% accuracy. The most informative attribute was “maximal leaf wetness in the last 15 days”.
Whereas, a relatively accurate prediction model was built from the meteorological data at hand, still much work has to be done in automating the data gathering process. There is also room for improvement of the prediction model in terms of both prediction accuracy and size.
Primary authors
Dr
Branko Kavšek
(University of Primorska)
Mr
Damjan Jurič
(University of Primorska)
Dr
Dunja Bandelj
(University of Primorska)
Dr
Maja Podgornik
(University of Primorska)
