In the field of machine learning (ML), data scarcity is a serious difficulty because data gathering can be costly, time-consuming, and challenging, especially in poor nations. The complexity of using datasets to forecast livestock diseases enabling intervention and surveillance is overstated. This work provides a data synthesis method that has been applied to produce fresh data samples with accuracy from limited real-world data, in order to meet this difficulty. The ML models are trained on a large amount of data, which eliminates overfitting. We propose the synthesis of categorical data for training machine learning models for the prediction of Peste des Petits Ruminants (PPR) disease using Generative Adversarial Networks, namely the Conditional Tabular Generative Adversarial Network. The training score increased to 0.89 and the cross-validation score to 0.87 when the Random Forest algorithm was applied to the synthesized data, according to the results. The Peste des Petits Ruminants (PPR) disease can be predicted and monitored with the help of the generated dataset. With more data available, the suggested approach can be used in any area with categorical data for data-driven models and may enhance the performance of machine learning models.