Abstract: In view of the research on the impact of spatial scale on urban ecological environment and human activities, this paper takes Nanchang City as the research area and divides three grid scales of 300×300, 500×500 and 700×700. Using the remote sensing ecological index RSEI (target vector) to quantify the urban ecological environment quality, combined with POI points, Weibo check-in points and road network data (feature vectors), the random forest regression model is used to analyze the fitting effect between the two scales. The results show that: ①RSEI and POI points, Weibo check-in points and road network have strong negative correlation at three scales; and the negative correlation is RSEI and Weibo check-in point data, and the worst is RSEI and road network data. ②The random forest regression model at 300×300 scale has the best fitting effect. As the scale becomes larger, the effect of the fit will get worse. ③Regardless of the scale change, the standardized residuals fitted by random forest regression show a normal distribution; and as the scale becomes larger, the randomness of the standardized residuals space distribution also increases. The random forest regression model provides an effective way for the relationship between urban ecological environment and human activities, and provides a scientific basis and reference for urban ecological civilization construction.

Key words: spatial scale, RSEI, urban ecological environment, random forest regression, Nanchang