The apparatus for recommending an optimum training is provided with a standard model database which stores standard value information of standard skill corresponding to a standard talent image, and a skill/training correspondence database which stores training information corresponding to the standard value information. Furthermore, a questionnaire response section and a test response section collect skill information of a trainee in a question and answer format, via on-line. Then, a skill analysis section objectively analyzes the skill and talent image of the trainee, based on the comparison result of the standard value information and the skill information. Finally, a training selection section selects training information that suits to the trainee from the skill/training correspondence database as the optimum training recommendation information, based on the result of analysis by the skill analysis section.