Objective The research in this paper aims to fill the gap of the current lack of bus driving cycles in line with the characteristics of the urban terrain in Chongqing. Methods An improved K-mean clustering algorithm combined with a Markov chain was proposed to construct driving cycles. The effective driving data of bus No. 805 in urban areas of Chongqing were smoothed using a filtering algorithm from which 1 721 kinetic segments were extracted. The feature parameter matrices of each kinetic segment were dimensionally reduced using kernel principal component analysis suitable initial cluster centers were selected with the particle swarm algorithm for K-means and the kinetic segments were classified and labeled using the improved K-means clustering algorithm. Finally reasonable kinetic segments were selected through the Markov chain to construct a bus driving cycle in urban areas of Chongqing with a duration of 1 310 seconds. Results The constructed driving cycle was compared with the data collected from real vehicles and some typical cycles at home and abroad. The results showed that the constructed cycle exhibited consistent trends in the joint distribution of speed and acceleration with the original data and the errors of the characteristic values of the constructed driving cycle compared with the data collected from real vehicles were all less than 7%. The constructed cycle highly matched the characteristics of the original data and can well reflect the high idle time ratio frequent acceleration and deceleration and other actual traffic conditions in urban areas of Chongqing. Conclusion The constructed cycle can provide a basic standard for bus route planning emission testing and fuel consumption testing in Chongqing and the research methods involved can serve as a reference for energy management of new energy vehicles control strategies as well as city driving cycle construction and optimization of city bus routes.