We present an analysis of new spectroscopic observations of the bright Be star kappa Dra obtained at the Ondrejov observatory during 1992-2003 and UBV photometric observations secured at several observatories. General characteristics and a line identification of the spectrum of ? Dra are obtained in the regions 3730-5650 Angstrom and 5850-7800 Angstrom by a comparison with the theoretical spectrum. The fundamental stellar parameters have been obtained from a comparison with a grid of NLTE model atmospheres. The best fit was found for T-eff = 14 000 K, log g = 3.5, and v sin i = 170 km s(-1). These values together with a Hipparcos parallax lead to a stellar mass M = 4.8 +/- 0.8 M. and radius R = 6.4 +/- 0.5 R.. It is encouraging to see that these values agree well with the expected evolutionary mass and radius for the effective temperature we derived. Long-term variations of ? Dra were analysed using measurements of equivalent widths, central intensities, peak intensities of emission lines and emission peak velocity differences for Halpha, Hbeta, Hgamma, Hdelta, and some helium, silicon, and iron lines. The previously reported period of 23 years in the variation of the emission strength is probably a cyclic, not a strictly periodic phenomenon. An attempt to find out a period from all available records of the Hbeta emission strength led to a value of (8044 +/- 167) days (22.0 years) but the phase plots show that each cycle has a different shape and length. The maximum strength of the emission lags behind the brightness maximum. This is a behaviour usually observed for long-term changes of Be stars with a positive correlation between the brightness and emission strength. Since there are obviously no published speckle observations of the star, we suggest these should be carried out. They could help to deny or confirm the possibility that the emission episodes are triggered by a periastron passage of a putative binary companion moving in an eccentric orbit with a 8044-d period, as it seems to be the case for some Be binaries. For the moment, the nature and origin of the disk around kappa Dra remains unknown. From the comparison of the electronic spectra obtained at different phases of the long-term cycle and synthetic spectra it appears that there are no detectable changes in the photospheric part of the Balmer lines related to variations in the Balmer emission strength which could be attributed to an extended photosphere corresponding to inner parts of the disk, optically thick in continuum.