In every region of the globe, agriculture is the most important activity, and it is highly dependent on rainfall. It was possible to forecast the rate of agricultural output by taking into account of some factors such as precipitation, wind, temperature, soil, fertilizer and solar radiation. Alterations in the weather play a significant part in crop yield. The accurate projection of crop yields at an early stage is beneficial for market pricing, arranging labour, transportation, and harvest organization. The primary objective of this research is to make an accurate prediction of crop yield. Introduction of deep learning models allows the perfect prediction of production rates. In this paper, an Improved Adaptive Moment Estimation Optimizer Function (IAOF-BiLSTM) is an Enhanced Optimizer Algorithm (EOA) is proposed to get an accurate prediction of yield along with the Bidirectional Long Short-Term Memory (BiLSTM) model. Data was collected from the open source repositories and government websites of Tamilnadu (India) from 1965 to 2022 for training and testing purpose. The proposed model is compared with other existing standard deep learning models and outcome of the proposed IAOF-BILSTM optimizer with Bi-LSTM algorithm predict very well and shows the less error rate in training data and also the algorithm effectively handles issues of underfitting and overfitting in predictions. The performance metrics such as MAE, RMSE and R2 were used to compare the loss value with existing models. The evaluated value is compared between the actual and predicted crop yield and the achieved results are measured using MAE, RMSE and R2.The result shows that the improved optimizer algorithm IAOF-BiLSTM model gives better accuracy and less error rate compared with existing models.