Generative modeling has gained importance in music generation, offering new possibilities for understanding and creating music. In the case of Indian classical music (ICM), where Ragas form the melodic framework, developing generative models is challenging due to the specific rules and emotional depth associated with each Raga. This study focuses on Raga Yaman, known for its distinctive melodic structure to develop a hybrid generative model that integrates n-gram probabilities with expert insights. The proposed methodology involves constructing a dataset of compositions in Raga Yaman, including bandishes (short, composed songs), aalaps (slow improvisations), and taans (fast improvisations). N-gram probabilities are extracted from the dataset to identify statistical patterns governing note progressions, while expert musicians contribute insights into the rules of Raga and stylistic nuances. These elements are synthesized into a generative grammar, ensuring the production of authentic and musically meaningful phrases. Preliminary evaluations by music experts indicate that the model effectively captures the structural constraints and emotional essence of Raga Yaman, with positive feedback on its coherence and expressiveness. This hybrid approach bridges data-driven techniques and domain-specific expertise. Beyond its application to Raga Yaman, the framework holds potential for extension to other Ragas, contributing to the exploration, innovation, and preservation of ICM traditions.