Subsurface well logging is a cornerstone in the oil and gas industry, providing critical insights into reservoir properties and aiding in reservoir characterisation. Neutron Porosity Index (NPHI) is a key parameter derived from well logging data, offering valuable information about reservoir porosity. Accurate prediction of NPHI is essential for reservoir evaluation and production forecasting. Traditional methods often rely on empirical relationships or physics-based models, which may be limited in their ability to capture the complex relationships present in the data. With the advancement of machine learning techniques, there has been a growing interest in developing data-driven models for well logs prediction. In this study, we proposed a novel approach for predicting NPHI in subsurface formations via multivariate time series analysis using Long Short-Term Memory (LSTM). To demonstrate the performance advantage of the proposed method, we also conduct a comprehensive comparative analysis of the normal Artificial Neural Networks (ANN). Both LSTM and ANN models are rigorously evaluated on the dataset from Volve oil field well 15/9-F-1A. LSTM has exceeded the aforementioned ANN model during forecasting, thereby offering valuable insights for reservoir characterisation and hydrocarbon exploration in the oil and gas industry.