This study aimed to fit regression models for harvest prediction in Prata type bananas. The experiment consisted of plants and bunches of bananas carried out in Guanambi, BA, with genotypes Dwarf Prata (AAB) and BRS Platina (AAAB), planted at a spacing of 3.0 × 2.5 m, with irrigation. Measurements of vegetative and yield characteristics were sampled at random. Models of simple and multiple linear regression were estimated, considering as independent variables with highest correlation

more »

... s with the masses of the bunch (MB) and hands (MH). The simple linear regression models allow prediction of the masses of the bunch and hands, according to the number of hands (NB), with better precision, for both genotypes and at least 120 days prior to harvest. For the' Dwarf Prata', the equations were: MCA = -10.05 + 3.08NH; r² = 0.99, e, MPE = -12.38 + 3.55NH; r² = 0.99. For the 'BRS Platina' equations were: MCA = -1.37 + 3.02NH; r² = 0.97 e MPE = -2.74 + 2.88NP; r² = 0.97. The determination coefficients for the adjusted models ensure consistency of the regressions for the estimation of Prata type banana production.