The particle diffusion coefficient and convection velocity of electron cyclotron resonant heating plasmas have been evaluated using a density modulation experiment in the Large Helical Device. The diffusion coefficients and convection velocities were estimated by a newly developed analysis method. This method combines conventional quasi-Newton optimization and a genetic algorithm (GA). The GA finds the appropriate initial values. By giving the appropriate initial values for the quasi-Newton method, falling into an inappropriate solution which is not applicable for transport physics such as negative diffusion coefficients or extremely high values can be avoided. The diffusion coefficient obtained with the modulation experiment (Dmod) in the core region increases monotonically with the increase of the effective collisionality (). The experimental results of Dmod and the convection velocity (Vmod) can reconstruct the density profile at , but they do not reconstruct it at . The values of Dmod were from several times to three orders higher than those obtained by neoclassical prediction. On the other hand, Vmod was comparable with the neoclassical estimation. Ion-scale microturbulence was measured and compared with Dmod and Vmod. The comparison suggested possible linkage between the two.