Sales prediction is a major task in every business. A potential prediction may significantly impact the revenue loss, out of stock and excessive stock. Much existing research has been applied to predict dairy sales prediction. However, some limitations are required for an advanced forecasting model that can efficiently capture the difficult relationships between ecological factors, biological data and sales predictions. This is critical because accurate forecasting allows dairy companies to optimize inventory, reduce revenue loss and better recognize demand patterns. Thus, the current research aims to address this by developing a hybrid deep learning Multi-Perspective Fusion Bi-LSTM with Universal Scale CNN that is capable of capturing both local and long-term dependencies in sales data while also accepting changing environmental conditions and real-time data sources. The proposed research utilize a dairy supply chain dataset to evaluate the proposed model CNN (Convolutional Neural Network). The proposed model utilizes a combination of Universal Scale CNN (1D-CNN) for feature extraction and Multi-Perspective Fusion Bi-LSTM for time-series forecasting. The 1D-CNN efficiently captures local patterns in the data, while the Bi-LSTM, with its bidirectional structure, learns long-term dependencies, improving the model's prediction abilities. This combination facilitates the model to capture both local patterns and long-term dependencies in the data. This proposed combination of 1D-CNN and Multi-Perspective Fusion Bi-LSTM addresses the limitations of traditional models by effectually extracting features and capturing both short-term and long-term patterns in the data. Thus, the numerical values attained by the proposed model are MSE as 36011.60, R2 as 0.9524, MAE as 82.004 and RMSE as189.767. This approach outperforms existing methods in terms of accuracy and robustness in predicting dairy sales. The performance metrics offer a broad analysis in terms of accurate prediction of the proposed model.