<![CDATA[Fertilization accounts for 40-60% of total oil palm maintenance costs, yet smallholder farmers who manage roughly 41% of Indonesia's oil palm area still determine dosages from habit and visual estimation rather than from measured soil condition or plant age, a gap linked in field surveys to dosage errors of 25-35% relative to PPKS-recommended values. Prior decision-support tools, including web-based lookup tables and static recommendation charts, lacked dynamic input handling, did not model the inherent uncertainty in soil nutrient data, and were validated only internally without independent field verification. To close this gap, this study designed, implemented, and empirically validated SawIT PalmExpert, a Mamdani Fuzzy Logic-based expert system that recommends macro-fertilizer dosages (Urea, TSP/SP-36, KCl/MOP, Kieserite/Dolomite) from five inputs soil pH, Nitrogen (N), Phosphorus (P), and Potassium (K) obtained from accredited laboratory analysis (pH H2O 1:5; Kjeldahl N; Bray-1 P; NH4OAc K), together with plant age. The specific contribution of this research is threefold: (1) the first documented web-based expert system to integrate soil chemical status and plant age simultaneously as continuous fuzzy inputs for oil palm fertilizer dosing, rather than treating them as independent lookup criteria; (2) a 24-rule Mamdani knowledge base explicitly calibrated against PPKS agronomic standards through structured expert interviews and three rounds of rule validation, rather than derived from generic literature; and (3) a multi-dimensional empirical validation combining functional, accuracy, usability, and field simulation testing on a single deployed system. Concretely, black-box functionality testing across 10 scenarios all passed; fuzzy output accuracy testing on 15 cases against PPKS references yielded 100% conformity within a +/-10% tolerance (maximum deviation 2.9%, confined to fuzzy-set transition zones); usability testing with 10 purposively sampled respondents (6 smallholder farmers, 4 extension workers), using a validated instrument (content validity index 0.88; Cronbach's alpha 0.84), produced an overall score of 4.24/5.0 ("Very Good") across five dimensions ease of use, information clarity, response speed, recommendation relevance, and overall satisfaction; and field simulation testing on three representative soil scenarios received full agronomist approval. These results demonstrate that a laboratory-calibrated Mamdani fuzzy expert system can deliver both algorithmic accuracy and practical usability, positioning SawIT PalmExpert as a scientifically grounded, field-ready decision-support alternative to experience-based fertilization for smallholder oil palm farmers.]]>
Copyrights © 2026
