Land use change is a long-term process with certain complexity. The traditional transfer matrix method can only compare the two phases of land use data, but cannot summarize the long-term overall change law. The frequent itemsets method cannot derive the sequence trajectory. To overcome the shortcomings of these methods, this paper proposes a land use change model based on sequence model. According to the actual characteristics of land sequence data and the vertical format sequence pattern SPADE algorithm, taking Guangxi, China as an example, we calculated the sequence of 22 secondary land use types in 7 periods from 1970s to 2015. The results show that, in the 35 years, land use types changed in 6.58% of Guangxi, and the change areas were mainly concentrated in roads, towns and settlements; the top 3 types of land use change 1-sequence support degree in Guangxi are {wood land}, {other wood land} and {sparse forest land}, and the support degree values are 0.5109, 0.3810 and 0.2333, respectively. The top 3 types of the 2-sequence support degree are {wood land, other wood land}, {other wood land, wood land} and {sparse forest land, wood land}, and the support degree values are 0.2040, 0.0699 and 0.0640, respectively. The top 3 types of the 3-sequence support degree are {sparse forest land, other wood land, wood land}, {sparse forest land, wood land, other wood land} and {high coverage grassland, wood land, other wood land}, and the support degree values are 0.0065, 0.0044 and 0.0031, respectively. The land use change 3-sequence {with forest land, other woodland, and other construction sites} has a support degree of 0.0007. The land use change in the study area mainly occurs in forest land, and some forest land is converted into construction land types such as mines, quarries and traffic roads; urban land is mainly converted from dry land and paddy fields; there is no land unit converted into paddy fields during the study period. The land use change sequence analysis model proposed in this paper can calculate the sequence of multi-period land use change as a whole, and make up for the shortcomings of studying land use change from a long-term scale.