The result of a study on the application of an improved statistical prediction method for estimating the intermediate-term (months) and long-term (years) behavior of solar flux (R¿10.7) is discussed. The study indicates that better predictions, in a chi square sense, are possible by selecting sets of the solar flux data such that each set (cycle) starts and ends at the maxima (for minima) for the data base and initialization point of the procedure. Then one applied a Lagrangian least-square statistical technique. Evidence is also presented to support the existence of an aperiodic variation in the periods as well as the amplitudes.