Abstract: Diesel is often used as an accelerant in arson cases, so identifying its source is crucial to the investigation. Stable isotope analysis is a method for detecting molecular elements at the element level. The carbon and hydrogen isotope information of diesel n-alkanes has geographical characteristics reflecting the source of raw materials. Previous studies have successfully established the relationships between field samples and control samples using cluster analysis and discriminant analysis algorithms. However, with increased accessibility, criminals can cross borders more easily, complicating the investigation process. Traditional analytical methods typically rely on known diesel sources to compare isotopic information between the sample and the control material. However, criminals often purchase diesel from different locations and ship it to crime scenes, thus separating control samples from crime scene evidence and complicating further investigation and evidence collection. Therefore, tracing and origin analysis of diesel oil is a necessary condition to improve the diesel oil tracing system based on isotope information. This method supports the investigation of cross-regional fire crimes. In this study, discriminant analysis and neuwere ral network algorithms were used to analyze the carbon and hydrogen isotope data of n-alkanes in diesel samples from Beijing, Gansu, Heilongjiang, Shandong. The corresponding analysis model was established, and the effectiveness of different statistical methods for diesel engine source inference was discussed. The carbon and hydrogen isotopes of n-C₁₂-n-C₁₉ in 44 diesel oil samples from Beijing, Gansu, Heilongjiang, and Shandong were analyzed by gas chromatography-isotope ratio mass spectrometry (GC-IRMS). The results were evaluated using discriminant analysis, multilayer perceptron (MLP) and radial basis function (RBF). The results showed significant regional differences are in the carbon and hydrogen isotope ratios of diesel n-alkanes. When the discriminant analysis algorithm was used to determine the diesel fuel source, the model based on carbon and hydrogen isotope data has a better classification effect than the model based on a single element. Using carbon and hydrogen isotope data as indexes, the source traceability of diesel oil was evaluated by MLP and RBF neural network models. The results showed that the classification accuracies of MLP and RBF models are 90.0% and 90.9%, respectively. Among the three models, the discriminant function model has the highest overall accuracy. However, when the samples from four regions were distinguished by the two neural network models, the classification accuracy of most regions is better than or close to the discriminant analysis model, and some accuracy is lower than the discriminant analysis model. The horizontal comparison of discriminant analysis results should be combined in application to achieve the best traceability results. This study effectively fills the gap in domestic stable isotope diesel source analysis, providing strong technical support for the investigation of diesel arson in the future.