Internal combustion engines play a dominant (and steadily increasing) role in both on-road and off-road applications. With emission regulations becoming more and more stringent and the requirement to reduce the greenhouse gas emission, engine development efforts, nowadays, are more focused towards the effective utilization of fuel (and hence lower CO2 emission) and reduction of regulated emissions. The spark advance, for example, has a considerable influence on performance, combustion and emission characteristics of a spark-ignition engine, and therefore, it is important to determine the optimum spark timing for effective utilization of engine (or for improved fuel economy, power output and emissions). At optimum spark timing, peak cylinder pressure usually occurs around 15 degrees after top dead center (ATDC), and 50% of mass fraction burned point generally occurs around 9 degrees ATDC. However, when engines operate at lighter loads, cycle-to-cycle combustion variability also becomes important to define the optimum spark timing. Thus, in the present work, a stationary 661 cm3, single-cylinder, water-cooled spark-ignition engine (modified from a compression-ignition engine with arrangements to vary the compression ratio and spark timing) is used to determine the optimum spark timing at different loads and compression ratios. An artificial neural network (ANN) model is also developed to predict the optimum spark timing, particularly at lighter loads, by considering coefficient of variation of indicated mean effective pressure along with location of peak pressure for optimization purpose. Experimental data obtained at different operating loads and compression ratios with gasoline as a candidate fuel were utilized for training, testing, and validating the ANN model.