The froth flotation process is extensively used for the selective separation of valuable base metal sulfides from uneconomic associated minerals. However, in this complex multiphase process, various parameters need to be optimized to ensure separation selectivity and peak performance. In this study, two machine learning (ML) models, artificial neural network (ANN) and random forests (RF), were used to predict the efficiency of in-house synthesized chitosan-polyacrylamide copolymers (C-PAMs) in the depression of iron sulfide minerals (i.e., pyrite) while valuable base metal sulfides (i.e., galena and chalcopyrite) were floated using nine flotation variables as inputs to the models. The prediction performance of the models was rigorously evaluated based on the coefficient of determination (R2) and the root-mean-square error (RMSE). The results showed that the RF model was able to produce high-fidelity predictions of the depression of pyrite once thoroughly trained as compared to ANN. With the RF model, the overall R2 and RMSE values were 0.88 and 4.38 for the training phase, respectively, and R2 of 0.90 and RMSE of 3.78 for the testing phase. As for the ANN, during the training phase, the overall R2 and RMSE were 0.76 and 4.75, respectively, and during the testing phase, the R2 and RMSE were 0.65 and 5.42, respectively. Additionally, fundamental investigations on the surface chemistry of C-PAMs at the mineral–water interface were conducted to give fundamental insights into the behavior of different metal sulfides during the flotation process. C-PAM was found to strongly adsorb on pyrite as compared to galena and chalcopyrite through zeta potential, X-ray photoelectron spectroscopy (XPS), and adsorption density measurements. XPS tests suggested that the adsorption mechanism of C-PAM on pyrite was through chemisorption of the amine and amide groups of the polymer.