Liberica coffee is a popular one in the global trade market beside Arabica and Robusta. Moisture content (MC) is one of the most important quality parameters of green coffee beans production and trading. NIR spectroscopy is one of the alternative methods for the rapid determination of chemical compounds in the products non-destructively. The purposes of this study are to determine the best calibration model, data transformation, and data pretreatment for moisture content of Liberica coffee. The ground coffee samples were measured by FT-NIRS in the wavelength of 1000-2500 nm, properly the moisture contents of the same samples were determined by the oven method. The obtained spectrums were transformed into absorbance (Log 1/R) and Kubelka-Munk (K/S) units. Data pretreatment, such as standard normal variance (SNV), second derivative (dg2), and multivariate calibration method such as a partial least square (PLS), were carried out to develop the best calibration model. Good accuracy for the prediction of moisture content of Liberica coffee green bean was obtained from the spectral data pretreated with dg2 and Kubelka-Munk(K/S) data transformation with the statistical evaluation values of r (0.87), RPD (2.05), consistency (99%) and CV (5.76 %)