A deep learning approach for forecasting daily prayer times in Nigerian geo-political zones

Abstract

Accurate determination of prayer times is essential for the daily worship practices of Muslim communities. However, many existing online and offline prayer time prediction systems are limited by regional variations, weather conditions, and complex geographical factors. Most current approaches rely on conventional algorithms or manual adjustments, which reduce accuracy across diverse locations. This study addresses these limitations by applying deep learning techniques to predict the five daily Islamic prayer times across major cities in Nigeria, including Kaduna, Borno, Abuja, Lagos, Enugu, and Rivers. Historical prayer time data were obtained from the Adhan API and subjected to preprocessing and feature engineering. Two models; Long Short-Term Memory (LSTM) and Artificial Neural Networks (ANN) were developed and evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and the coefficient of determination (R^²). Results show that both models achieved high predictive performance, with near-perfect R^² values. However, ANN consistently produced lower MAE, RMSE, and MAPE for Fajr, Dhuhr, Asr, and Maghrib, indicating superior accuracy, while LSTM slightly outperformed ANN for Isha. The best-performing model for each prayer was used to forecast daily prayer times for 2025. The study demonstrated that deep learning models can significantly improve the accuracy of prayer time prediction and provide a reliable solution for faith-based timekeeping.