<p>Water is considered a valued asset that plays a dominant role among all living beings. The quality of water is generally affected by pollution that creates an impact on human health, water species and its related ecosystem. In order to control water pollution and chemical contaminations, water quality prediction research is essential as it helps ensure the survival rate of nautical life. Existing predictive methods for water quality face various challenges, including poor generalizability, poor prediction accuracy, and high complexity. Some of the methods draw on data containing large amounts of unknown or undefined input data, which in turn leads to time-consuming processes. To overcome these challenges, we propose a deep learning model Bidirectional- Long Short-Term Memory (Bi-LSTM) to predict the quality of water. Here, the optimal features are chosen using designed Adaptive Spider Monkey Optimization (ASMO) for water quality prediction. Initially, data normalization is performed by Min–Max normalization and the dimension of the data is transformed into a feature extraction phase that has the (Convolutional Neural Network (CNN) feature. Later, the feature conversion performs the conversion of data features into vector form and the ASMO approach is designed for selecting the relevant features from the resultant converted features. Finally, the Bi-LSTM with selected features is used for predicting the water quality. Moreover, ASMO + Bi-LSTM maintained high-performance metrics with a True Negative Rate (TNR), True Positive Rate (TPR), and accuracy of 94.68%, 96.55%, and 92.57%.</p>