Barium stars are chemically peculiar stars that exhibit enhancement of s-process elements. Chemical abundance analysis of barium stars can provide crucial clues for the study of the chemical evolution of the Galaxy. The Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST) has released more than 6 million low-resolution spectra of FGK-type stars by Data Release 9, which can significantly increase the sample size of barium stars. In this paper, we used machine-learning algorithms to search for barium stars from low-resolution spectra of LAMOST. We have applied the Light Gradient Boosting Machine (LGBM) algorithm to build classifiers of barium stars based on different features, and build predictors for determining [Ba/Fe] and [Sr/Fe] of barium candidates. The classification with features in the whole spectrum performs best: for the sample with strontium enhancement, Precision = 97.81% and Recall = 96.05%; for the sample with barium enhancement, Precision = 96.03% and Recall = 97.70%. In prediction, [Ba/Fe] estimated from Ba ii line at 4554 Å has smaller dispersion than that from Ba ii line at 4934 Å: MAE4554 Å = 0.07, σ 4554 Å = 0.12. [Sr/Fe] estimated from Sr ii line at 4077 Å performs better than that from Sr ii line at 4215 Å: MAE4077 Å = 0.09, σ 4077 Å = 0.16. A comparison of the LGBM and other popular algorithms shows that LGBM is accurate and efficient in classifying barium stars. This work demonstrated that machine learning can be used as an effective means to identify chemically peculiar stars and determine their elemental abundance.