Abstract: The quantitative analysis models of 9 elements, including chromium, nickel, copper, silicon, manganese, vanadium, carbon, titanium, and aluminum in scrap steel were established based on laser induced breakdown spectroscopy with Stacking integrated algorithm model. 12 alloy steel standard samples were collected using a portable laser induced breakdown spectroscopy instrument. After the spectral data were subjected to error removal, averaging, and baseline correction, the spectral lines of each element and the matrix element (iron element) were screened based on the spectral line database of the National Institute of Standards and Technology (NIST). The spectral lines of each element and the normalization lines were optimally matched using the degree of correlation, and the optimal normalized spectral line pairs for each element were obtained. The spectral line data normalized by the optimal spectral line pairs were taken as the input for each element model. The outputs of the models, namely the Lasso regression model, the ridge regression model, and the quadratic linear regression model, were combined and used as the input for the meta-learner. The certified values of the elements were used as the output of the meta-learner. The meta-learner was selected as the logistic regression model for training and modeling. Finally, the Stacking integrated algorithm models for each element were obtained. The results showed that the correlation coefficients of determination of the models of 9 elements were in the range of 0.985 6-0.999 7,with the root mean square errors in the range of 0.008 1-0.046 8, and the average absolute errors in the range of 0.006 0-0.034 5. The RSDs (n=5) of the elemental determination values were less than 7.0%. The model was used to predict the alloy steel standard sample, and the absolute values of the relative error between the determined values and the certified values were less than 10%.