Forecasting agricultural prices is exceptionally difficult because of the unpredictable global weather events, influential role of government policies, evolving consumer preferences and technology. This study focuses on the impact of price forecasting in the agricultural sector, specifically within the cotton industry. Accurate predictions of cotton prices are of most important to various stakeholders, including cotton farmers, textile mills and shippers. To enhance forecasting accuracy, this research employs advanced machine learning techniques such as Artificial Neural Networks (ANN), Recurrent Neural Networks (RNN), Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), as well as stacked LSTM models. These models are trained and evaluated using statistical metrics like RMSE, MAPE, SMAPE and MAE. Notably, the stacked LSTM model consistently outperforms other models, demonstrating superior predictive performance with minimal errors. This study also highlights the stacked LSTM's ability to effectively capture long-term data dependencies, leading to significantly improved prediction precision.