In marine or stationary engines, consistent engine performance must be guaranteed for long-haul operations. A dual-fuel combustion strategy was used to reduce the emissions of particulates and nitrogen oxides in marine engines. However, in this case, the combustion stability was highly affected by environmental factors. To ensure consistent engine performance, the in-cylinder pressure measured by piezoelectric pressure sensors is generally measured to analyze combustion characteristics. However, the vulnerability to thermal drift and breakage of sensors leads to additional maintenance costs. Therefore, an indirect measurement via a reconstruction model of the in-cylinder pressure from engine block vibrations was developed. The in-cylinder pressure variation is directly related to the block vibration; however, numerous noise sources exist (such as, valve impact, piston slap, and air flowage). A deep neural network (DNN) model is among the most feasible ways to reconstruct the in-cylinder pressure from engine block vibrations, minimizing the effect of noise or uncertainties. Vibration and in-cylinder pressure signals were measured on a natural gas (NG)/diesel dual-fuel engine using an accelerometer and a piezoelectric sensor. A parametric study was conducted to investigate the effects of the composition of the training conditions on the accuracy of the model. The main parameters, engine load, engine speed, diesel injection timing, and intake air pressure, were varied to investigate their effect on the model prediction. The results showed that the prediction accuracy was directly related to the amount of training data that had similar trends of in-cylinder pressure with the test conditions, irrespective of the total amount and types of engine parameters of the training data. With these investigations, we can identify the relationships between the training dataset and model performance that can be used in the development of a cylinder pressure reconstruction model for the real-time verification of performance in dual-fuel engines.