The invention discloses a soil heavy
metal traceability and treatment method based on
machine learning regression, and belongs to the technical field of environmental
geochemistry and
artificial intelligence crossing. Comprising the following steps: acquiring geological
background information of a research area, selecting regression indexes capable of reflecting heavy
metal properties, and preprocessing; taking the preprocessed
data set as a
training set, constructing an element specificity
machine learning regression model for each heavy
metal element, establishing a nonlinear mapping model between the soil chemical components and the heavy metal elements, and training and optimizing the
machine learning regression model; and collecting chemical component data of each sampling point in the target area, inputting the chemical component data into the trained
machine learning regression model, outputting a heavy metal background predicted value of each sampling point, and determining a natural background value and human input content of the heavy metal. The problem of misjudgment of soil heavy metal heterogeneity is solved, a natural and human source quantitative
separation method is established to support accurate
traceability treatment, and an element specificity prediction framework is constructed to improve the adaptability of heavy metal elements.