The work is devoted to the development of the theoretical foundations of new universal complex chaos-dynamical approach to analysis and prediction of the atmospheric radon 222 Rn concentration changing from beta particles activity data of radon monitors (with pair of the Geiger-Müller counters). The approach presented consistently includes a number of new or improved methods of analysis (correlation integral, fractal analysis, algorithms of average mutual information, false nearest neighbors, Lyapunov exponents, surrogate data, non-linear prediction schemes, spectral methods, etc.) to solve problems quantitatively complete modeling and analysis of temporal evolution of the atmospheric radon 222 Rn concentration. There are firstly received data on topological and dynamical invariants for the time series of the 222 Rn concentration, discovered a deterministic chaos phenomenon using detailed data of measurements of the radon concentrations at SMEAR II station of the Finnish Meteorological Institute.