Abstract: The construction of rural human settlements has become an important task in the implementation of rural revitalization strategy. The research on the internal mechanism and key driving forces can provide scientific basis for the governance and construction of rural human settlements. The measurement model of rural human settlements quality, canonical correlation analysis, and ridge regression model were used to realize the measurement of rural human settlements quality in Gansu Province. and the key driving factors were identified. The results show that:(1) The average value of rural human settlements quality in Gansu Province was 0.238, and the proportion of counties above the mean level was 45.98%. The living conditions, the industrial and economic development, and the governance of ecological environment in Gansu Province show a decreasing trend from northwest to southeast. The dimension of public service facilities is high in the middle and low at both ends. The low-level and lower-level evaluation units in infrastructure dimension are relatively scattered and widely distributed. (2) In the group of natural geographic variables, typical load of irrigated arable land was -0.798, and it has positive effect on the target variables such as infrastructure, living conditions and industrial economy. All the explanatory indicators in the regional economic development group showed positive correlation with the typical variables, and positively promoted the quality of human settlements. In the social control variable group, the canonical loadings of two explanatory variables, i.e. the proportion of residents with primary school education level-lower, and the proportion of minority villages, were -0.976 and -0.494, respectively, which negatively affected rural infrastructure construction and industrial economic development. (3) The ridge regression coefficient of the proportion of irrigated arable land was 0.143 (P<0.05), which is the key driving factor to improve the quality of rural human settlements, while the proportion of villages in mountainous and hill areas was -0.134 (P < 0.05) is the key constraint factor. The proportion of villages and towns near railway stations or expressway intersections (P<0.01) and labor force in non-agricultural industries (P<0.05) played positive driving role. In the social and cultural environment, the regression coefficient of the proportion index of residents below primary school education level was negative (P<0.01). The low education level of residents has become the key obstacle to the improvement of rural living environment quality.

Key words: rural human settlements, key driving factors, canonical correlation analysis, ridge regression model, Gansu Province