digilib@itb.ac.id +62 812 2508 8800

ABSTRAK NNAMDI KINGSLEY UZOUKWU
PUBLIC Dwi Ary Fuziastuti

The main objective of this thesis is to describe the application of fuzzy logic-based machine learning for predictive modeling in agriculture. Fuzzy logic is a methodology that employs imprecision in the mapping of inputs to outputs. It is effective in using human-readable language to encode control rules for multivariate non-linear systems. In this study, the heat index for a growth room is modeled with temperature and relative humidity using fuzzy rules extracted from sensor data collected over a 19-day period with Arduino IoT infrastructure. Further, exploratory data analysis is performed to uncover the prevailing weather conditions in the growth room for the interval of study. When evaluated on a test set, the developed model obtained R2 of 0.965, and RMSE of 0.1, performing better than the multiple linear regressor and the lasso, and worse than the random forest regressor on both metrics. The exploratory data analysis shows a temperature range of 23oC – 28oC and a mean reading of 24.87oC. The relative humidity range is 63% - 85% with a mean reading of 75.82%. The fuzzy logic model proved to be an effective learning technique for the agricultural system to which it was employed.