Desenvolvimento de um sistema de apoio à decisão para a diagnose de doenças, pragas e distúrbios abióticos dos citros

Abstract

The Citrus average of yield is 22 ton/ha, which is considered low when compared to other producer countries. This is mainly due to the effects of plagues and diseases. A precise and liable diagnosis is therefore fundamental for the citrus disease control, besides avoiding unnecessary applications of pesticides, which one of the highest components in the total production cost. Computerized systems offer some solutions, is the use of a Decision Support System (DSS), a branch of the artificial intelligence, used to provide knowledge and reduce possible errors in diagnosis and management. In order to provide the knowledge and experience accumulated and disseminate it, in an efficient manner, the objectives in this research work were: i) Construct a solid knowledge basis to be used in the diagnosis of abiotic a biotic plant diseases; ii) Develop and implement a Decision Support System (DSS) and iii) Test and evaluate the system. The basic knowledge and experience was acquired by means of interviews with specialists from Fund for Citrus Plant Protection (FUNDECITRUS), and to complement the necessary information, specific literature about Citrus diseases were also consulted. For the development of the Citrus DSS, an interface in actual environment was constructed (Windows XP Home Editor), with photographs and a help command with technical terms in the specific area, to enable an easy access to the program. The DSS was made with the aid of the ´Borland Delphi´ tool version 5. A total of 562 questions were formulated, with 322 rules, besides including photographs to help in the diagnosis of the 34 diseases, 40 pests and 34 abiotics factors. The evaluation of DSS was divided in three distinct phases: verification and validation. In the verification phase, DSS was submitted to an analysis by means of a questionnaire submitted to citrus specialists. The validation phase of the program was conducted by four groups of people, with different knowledge levels (10 people/group), trying to provide the correct diagnosis. The results accounted for a 45,6% of correctness of the interviewed people and a 93,6% when using the DSS. The system was nominated Dr. Citrus.