In this study, the thermodynamic performance and emission behavior of the EF7 engine were comprehensively evaluated both experimentally and numerically when fueled with blended fuels, including pure gasoline, bioethanol (at 5%, 10%, and 15% by volume), and toluene (at 5%, 10%, and 15% by volume). The primary objective was to quantitatively analyze the effects of varying ratios of oxygenated and aromatic additives on key combustion parameters, including power, torque, and exhaust emissions, as well as to develop a predictive framework for emissions using an artificial neural network (ANN). Experimental tests were conducted under full-load conditions at constant engine speeds of 2000, 2500, and 3000 rpm. To analyze the combustion reactions, stoichiometrically balanced equations and reaction enthalpies were calculated for all fuel blends. The results indicated that E10 and T15 significantly influenced the optimization of environmental and performance parameters, respectively. Bioethanol notably reduced carbon monoxide (CO) and hydrocarbon (HC) emissions, while toluene, through increasing the octane number, improved power and torque, although a marked increase in nitrogen oxides (NOₓ) was observed for toluene-containing fuels. Furthermore, a multilayer ANN model was developed in the TensorFlow environment. This model, with inputs including fuel type, power, torque, and engine speed, predicted emissions with a mean absolute error below 5%. The close agreement between model predictions and experimental data demonstrated that machine learning can provide an effective, reliable, and scalable approach for simulating multiparameter combustion systems and for analyzing and predicting emissions in spark-ignition engines.