Analyzing data from the Israeli-Palestinian conflict is a critical task. Accurate fatality trend analysis is essential to understand the true human cost of the violence. Hybrid deep learning models offer a powerful method to learn complex patterns within such data. This paper outlines a direct comparative study of two specific architectures. Deep Neural Decision Trees and Deep Neural Decision Forests are trained for analyzing fatality trends in the Israeli-Palestinian conflict. Experiments were conducted on a public conflict dataset to assess the performance of the trained hybrid deep learning models. The experimental results clearly demonstrated the DNDF model’s superiority. It achieved a higher accuracy of 99.33% versus DNDT’s 98.03%. More importantly, its AUC score was 0.271, significantly outperforming DNDT’s 0.245. This indicates a much stronger ability to correctly distinguish between classes. The study concludes that the DNDF architecture is a more robust and effective model for this specific analytical task.