Transfer and semi-supervised learning-based spatial estimation method for air quality index of non-city region

Abstract

The invention discloses a transfer and semi-supervised learning-based spatial estimation method for an air quality index of a non-city region. The method comprises the following steps of (1) based onterrain distribution and air quality monitoring station deployment situations, looking for an auxiliary region around a target region, and constructing an auxiliary sample set; (2) based on a transferlearning technology, and in combination with a tagged sample set of the target region and the auxiliary sample set, training multiple regression models; and (3) based on a semi-supervised learning technology, and by utilizing an untagged sample set of the target region, enhancing and fusing the multiple regression models to obtain a final air quality index spatial estimation model. The method isbased on the transfer and semi-supervised learning technologies, and effectively utilizes tagged data generated by air quality monitoring stations of the non-city region of the auxiliary region and untagged data of the non-city region of the target region, thereby solving the problem that the target region lacks of the air quality monitoring stations deployed in the non-city region.

Description

technical field [0001] The invention relates to machine learning technology, in particular to a spatial estimation method for air quality index. Background technique [0002] Air pollution is a very serious problem at present, and monitoring air quality index is of great significance for pollution assessment and environmental governance. At present, the air quality index is monitored in real time through air quality monitoring stations. However, due to the high construction cost of air quality monitoring stations, the number of them is very limited, which makes the air quality index data extremely sparse in space. [0003] The spatial estimation of air quality index is to estimate the air quality index of any location without air quality monitoring stations. To solve this problem, traditional methods mainly consider the spatial distance between the current location and surrounding air quality monitoring stations, and use linear interpolation (such as inverse distance inter...

AI Technical Summary

Problems solved by technology

However, there are still big problems in the implementation of this method. The reasons are as follows: most of the air quality monitoring stations in China are deployed in urban areas, and there are very few air quality monitoring stations deployed in non-urban areas. Air quality monitoring stations in urban areas make the labeled data generated by existing air quality monitoring stations in these areas unable to cover all types of non-urban areas, so directly using these labeled data cannot be trained to distinguish between urban areas and non-urban areas. A Model of Air Quality Variation Law

Images