The invention provides a soil
pollution accumulation prediction method based on nesting of multiple self-correlation scales, and the method comprises the steps: analyzing a multi-scale nested space structure of
soil pollutants through a semi-variation function, determining a block gold effect, an offset
abutment value and a range parameter corresponding to each scale, and calculating the contribution degree of the block gold effect, the offset
abutment value and the range parameter to overall variation; in combination with original sampling data, identifying main driving factors of
pollutant accumulation under different scales, and comprehensively evaluating and sequencing the importance of the factors by adopting various quantitative analysis methods; constructing a cross-scale prediction
model set, taking each scale dominant driving factor as an input feature, carrying out modeling by using multiple basic models, screening and optimizing the models through clustering and a multi-objective optimization
algorithm, and forming an integrated learning framework of dynamic
weight distribution; and carrying out weighted fusion on prediction results according to contribution degrees of all scales, generating a soil
pollutant spatial distribution prediction map reflecting multi-scale nested features, and outputting the prediction map through a
visual tool, thereby improving prediction precision and spatial interpretation capability.