In developed electricity markets, the deregulation boosted competition among companies participating in the electricity market. Therefore, the enhanced reliability and availability of gas turbine systems is an industry obligation. Not only providing the available power with minimum operation and maintenance costs, but also guaranteeing high efficiency are additional requisites and efficiency loss of the power plants leads to a loss of money for the electricity generation companies. Multivariate Adaptive Regression Spline (MARS) is a modern methodology of statistical learning, data mining and estimation theory that is significant in both regression and classification is a form of flexible non-parametric regression analysis capable of modeling complex data. In this study, single shaft, 6MW class industrial gas turbines located at various sites have been monitored. The performance monitoring of a gas turbine consisted of hourly measurements of various input variables over an extended period of time. Using such measurements, predictive models for gas turbine heat rate and the gas turbine axial compressor discharge pressure values have been generated. The measured values have been compared with the values obtained as a result of the MARS models. The MARS-based models are obtained with the combination of gas turbine performance input and target variables and the complementary meteorological data. The results are presented, discussed, and conclusions are drawn for modern energy and cost efficient gas turbine and power plant maintenance management as the outcomes of this study.