Harumanis Mango (Mangifera indica) is known as one of the best table tropical fruit, due to its aroma and sweetness. Harumanis mango cultivar is included in the national agenda as a specialty fruit from Perlis, Malaysia for the world. Despite its overwhelming local demand in Malaysia and also internationally, the fruit supply never meets the demand. Mango flowering prediction is important as one of the factors to predict mango yield in order to implement effective forward marketing. Forward marketing is a contract that is signed between supplier and client based on the amount of delivery and the price of delivery in future, based on the predicted yield. Harumanis mango is a species that only bear fruit once a year. The biotic and environmental factors are reported in the literature as the factors that influence the mango trees flowering and fruit-bearing. The pre-processing and analysis done shows that the biotic and abiotic factors have non-linear relation with the yield. It is essential to develop, train and test the Harumanis mango tree flowering prediction model through machine learning approaches such as K-Nearest Neighbors (k-NN), Naive Bayes, Support Vector Machine (SVM), Classification Trees (CAT) and Random Forests (RF). Harumanis flowering predictive model on biotic and abiotic factors developed, trained and tested through the data accumulated from Harumanis trees in the greenhouse. The biotic factors are lysimeter, Length of First Whorl (LFW), Length of Second Whorl (LSW), Length of Third Whorl (LTW) and Diameter of the Whorl (DW). The Harumanis flowering prediction on biotic factors indicates that SVM technique prediction accuracy is at 79.9% as compared to k-NN, Naïve Bayes, CAT and RF at 66.5%, 74%. 71.3% and 72%, respectively. The SVM predictive model further tested on several kernels which are linear, polynomial, radial basis and sigmoid. The radial basis kernel accuracy is at 79.9% compared to linear, polynomial, and sigmoid at 65%, 65.7% and 59.8% respectively. The environmental data from Perlis Meteorology Department and the yield from Bukit Bintang Orchard were analyzed to identify the significant abiotic factors in predicting the Harumanis mango yield. Later, the abiotic factors which are average minimum temperature and average soil moisture of the 10 days from Harumanis greenhouse are calculated and utilized in developing the Harumanis flowering predictive model. The Harumanis mango tree flowering prediction on abiotic factors shows that SVM technique is at 90.6% accuracy rate compared to k-NN, Naive Bayes, CAT and RF at 76.6%, 75%, 82.1% and 72% respectively. Concluded that the prediction model using SVM on radial basis kernel through biotic and abiotic factors displays the highest prediction accuracy at 79.9 % and 90.6% accordingly. The SVM with radial basis kernel model able to perform flowering prediction although using a limited data due to the nature of the agricultural domain where the data collection and observation require a longer period of time.
