Côte d'Ivoire and Ghana are the main largest producers of cocoa in the world, however, the cultivation of this crop has led to the loss of vast tracts of forest areas in both countries. The efficient and accurate methods for remotely identifying cocoa farms are essential for the implementation of sustainable cocoa practices and the periodic and effective monitoring of forests. In this study, a multi-feature Random Forest (RF) algorithm was developed to map cocoa farms from other classes. Normalized difference vegetation index (NDVI) and second-order texture features were input variables for the RF model to discriminate cocoa farms in both countries. The estimated area for cocoa in Cote d'Ivoire was 4.8Mha and 2.3Mha for Ghana. The Produce Accuracy (PA) and User Accuracy (UA) of the RF model were 95.08% and 83.69% respectively. The results demonstrate that a combination of the RF model and multi-feature classification can accurately discriminate cocoa plantations, effectively reduce feature dimensions and improve classification efficiency.