Abstract: For promoting the efficiency and reducing the labor intensity in scheduling cigarette product delivery, an intelligent scheduling algorithm based on machine learning was designed. First, the sales orders of cigarette products were clustered with the agglomerative cluster. Second, a search tree was constructed by using an expert strategy function and a machine strategy function. Finally, an optimized schedule evaluation function was established and the schedule decision of cigarette product delivery was made. Accurate algorithm, genetic algorithm and the machine learning-based algorithm were comparatively tested by taking part of the sales orders of cigarette products of HongyunHonghe Group in April 2020 as an example. The results showed that the machine learning-based algorithm had better decision making performance and global convergence; for the same sales orders, it used 13.28% and 10.14% less vehicles, reduced transport mileage by 5% and 7.2%, reduced time cost by 68% and 45%, and the number of vehicles serving multiple points increased by 11.75 and 8.07 percentage points, respectively compared with accurate algorithm and genetic algorithm. This method provides a support for improving the delivery efficiency of cigarette products.