The spectral diagnostic bands and their ranges of styrene in different soils are extracted under indoor conditions and used as the basis for the identification and content prediction of styrene in soil. The soil spectral reflectance is processed by the differential processing method and the spectral data conversion method to increase the difference in spectral change among samples. The stepwise multiple linear regression (SMLR), partial least squares regression (PLSR), and support vector machine regression (SVMR) methods are used to model and predict the styrene content in different soils. The results show that the spectral characteristics of different soils contaminated by styrene are located near 1800, 2200, and 2400 nm, respectively. Under the influence of its physical and chemical properties and styrene content, the decrease rate of the soil spectral reflectance increases first and then decreases until the styrene is saturated in soils, and the change in reflectivity tends to be stable. The PLSR model has the best prediction effect on the styrene content in soils, followed by the SMLR model, and the SVMR model has the worst effect. The determination coefficient of the PLSR model is 0.982--0.998, indicating that the model is stable, and the difference between the corrected and predicted standard deviations is 0.004--0.016, indicating that the model has high prediction accuracy.