Predictive process monitoring (PPM) aims to predict future events in business process execution; however, traditional approaches have challenges in forecasting both sequential event behavior and the temporal linkages among activities. This research presents a unique Fusion Model Architecture (FMA) that integrates Long Short-Term Memory (LSTM) networks with Convolutional Neural Networks (CNN) to tackle these difficulties. The FMA utilizes LSTM’s capacity to model long-term dependencies in event sequences, while CNN captures spatial characteristics and activity interrelations. The proposed model achieves an accuracy of 82.9% in forecasting the next occurrence in a sequence and exhibits a Mean Absolute Error (MAE) of 2.29 days in estimating event timestamps. Moreover, the FMA demonstrates remarkable training efficiency, necessitating under 2 min on standard processing units, making it appropriate for practical applications. The comprehensive analysis of the Helpdesk dataset reveals that FMA surpasses conventional LSTM-based models, delivering superior predictive accuracy and computational efficiency. The prospective uses of this model in process optimization and real-time monitoring are examined in conclusion.