Flour from numerous varieties of yams, representing the hardness diversity, were analyzed using Near Infrared Spectroscopy (NIRS). A total of 78 data were analyzed using the Chemflow software. The data pretreatment was standard normal variate and Savitzky Golay algorithm. Calibrations were developed using NIPALS Partial Least Square regression (PLSR), using the standard Wold's algorithm with 4 blocs cross validation. Principal component analysis (PCA) was performed before PLS regression. The coefficient of determination in cross validation (R2CV), the standard error in calibration (SEC) and in cross validation (SECV), the standard error in prediction (SEP) and the coefficient of correlation (R) were calculated. The objective was to evaluate yam flour samples for hardness parameter prediction, using NIRS.