Four machine learning algorithms were used to predict the solid solution phases of high-entropy alloys (HEAs). To improve the model accuracy, the K-fold cross validation was adopted. Results show that the K-nearest neighbor algorithm can effectively distinguish body-centered cubic (bcc) phase, face-centered cubic (fcc) phase, and mixed (fcc+bcc) phase, and the accuracy rate is approximately 93%. Thereafter, CoCrFeNi2Alx (x=0, 0.1, 0.3, 1.0) HEAs were prepared and characterized by X-ray diffractometer and energy disperse spectrometer. It is found that their phases are transformed from fcc phase to fcc+bcc phase, which is consistent with the prediction results of machine learning. Furthermore, the influence of Al content on the microstructure and tribological properties of CoCrFeNi2Alx (x=0, 0.1, 0.3, 1.0) HEAs was evaluated. Results reveal that with the increase in Al content, the nanohardness and microhardness increase by approximately 45% and 75%, respectively. The elastic limit parameter H/Er increases from 0.0216 to 0.030, whereas the plastic deformation resistance parameter H3/Er2 increases from 0.0014 to 0.0045, which demonstrates an improvement in nanohardness with the increase in Al addition amount. In addition, the wear rate decreases by 35% with the increase in Al addition amount. This research provides a new idea with energy-saving and time-reduction characteristics to prepare HEAs.