A new arrival scheduling method based on classification and optimization is proposed to balance the efficiency of arrival operation and working experiences of air traffic controllers （ATCOs）. The arrival scheduling problem is transformed into a binary classification problem based on analyzing the operational characteristics of arrival control. A classifier based on the random forest algorithm is built to predict the landing sequence of arrival aircraft. The dynamic sequencing for arrival aircraft is achieved by using a composite scoring method and implementing a rolling time window mechanism. Furthermore， the landing times are optimized by concerning the landing sequence and optimization model. Finally， three groups of actual arrival operation data of Changsha Huanghua International Airport in rush hours are used to verify the feasibility of the proposed method. The results indicate that the random forest classifier’s prediction results are close to the real landing sequence results with an accuracy of more than 99.00%. Compared with the traditional first-come-first-served heuristic， the proposed method reduces the average delay， average flight time， maximum flight time， maximum delay， and variation of the landing sequence. Compared with the conventional optimization method， the proposed method reduces the number of landing position shifts from 12 to two while not significantly decreasing arrival scheduling performance.