Machine learning (ML) approach have been widely used to guide the discovery and design of materials. In this work, four ML algorithms were utilized to predict high-entropy alloys (HEAs) for solid solution phases. To improve the accuracy of the model, the K-fold cross validation was adopted. The results showed that the K-Nearest Neighbor can effectively distinguish BCC phase, FCC phase and mixed FCC + BCC phase, with an accuracy of 93%. Thereafter, the CoCrFeNi2Alx (x = 0, 0.1, 0.3 and 1) system alloys were prepared and characterized by X-ray diffraction (XRD) and Energy Disperse Spectroscopy (EDS), whose phase transformed from single FCC to FCC plus BCC, which is well consistent with the ML prediction. Furthermore, the influence of Al content on mechanical behavior and wear resistance of CoCrFeNi2Alx (x = 0, 0.1, 0.3 and 1) HEAs were both evaluated in micro- and macro scale. It was shown that with the increasing of Al content, the nano-hardness and micro-hardness increased by ~45% and ~75%, respectively. The H/Er increased from 0.0216 to 0.030 while H3/Er2 increased from 0.0014 to 0.0045, which demonstrated an improvement in nano-hardness with increasing Al addition. In addition, the wear rate decreased by 35% with the increasing Al content. This study will provide a new thought to design HEAs via an energy- and time saving method.