The number of muons in the air shower induced by ultra-high energy cosmic rays (UHECRs) has been measured with surface detector (SD) arrays of various experiments. Monte Carlo (MC) prediction of the number of muons in air showers depends on hadronic interaction models and the primary cosmic ray composition. By comparing the measured number of muons with the MC prediction, hadronic interaction models can be tested. The Pierre Auger Observatory reported that the number of muons measured by water Cherenkov type SD is about 1.8 times larger than the MC prediction for proton with QGSJET II-03 model. The number of muons in the Auger data is also larger than the MC prediction for iron. The Telescope Array experiment adopts plastic scintillator type SD, which is sensitive to the electromagnetic component that is the major part of secondary particles in the air shower. To search for the high muon purity condition in air showers observed by the TA, we divided air shower events into subsets by the zenith angle θ, the azimuth angle ϕ relative to the shower arrival direction projected onto the ground, and the distance R from shower axis. As a result, we found subsets with the high muon purity ∼65%, and compared the charge density between observed data and MC. The typical ratios of the charge density of the data to that of the MC are 1.71 ± 0.10 at 1870 m < R < 2150 m and 3.24 ± 0.40 at 2850 m < R < 3280 m. The difference in the charge density between the data and the MC is larger at the higher muon purity. These results imply that the excess of the charge density in the data is partly explained by the muon excess.