Achieving precise prediction of TOC content in hydrocarbon source rocks across three-dimensional space holds significant value for determining basin hydrocarbon generation potential and identifying favorable exploration targets. To achieve this objective, this study first employs a Support Vector Machine (SVM) algorithm for TOC logging characterization. By establishing a nonlinear relationship between core TOC measurements and logging curves, a Genetic Algorithm-optimized Support Vector Machine (GA-SVM) model is utilized to compute TOC logging curves. The GA-SVM algorithm demonstrated excellent agreement with core TOC measurements, validating its predictive accuracy. Comparisons with traditional methods and multiple linear regression algorithms revealed superior prediction precision, highlighting the advanced nature of the GA-SVM approach. Second, Cascade Random Forest (CRF) algorithm was employed for three-dimensional quantitative prediction of hydrocarbon source rock TOC content. Multiple seismic attributes were selected as input parameters, while the TOC curve predicted by the GA-SVM algorithm—after resampling and filtering—served as the output parameter. Three-dimensional quantitative TOC predictions were obtained through feature enhancement classification using the cascaded forest. The CRF algorithm predictions, GA-SVM predictions, and measured core TOC data showed good agreement at well locations, indicating the random forest algorithm"s high accuracy and stability for three-dimensional TOC content quantification. By combining GA-SVM and CRF machine learning algorithms, this study achieved high-precision quantitative prediction of hydrocarbon source rock TOC content from one to three dimensions.