Solar radiation is a key factor influencing sustainable building engineering, in terms of both optical and thermal properties of building envelopes. Solar irradiance data in a conventional weather data file are broadband, representing the total of ultraviolet (UV), visible light (VIS), and near-infrared radiation (NIR), three components of the solar spectrum; however, these three components play different roles in sustainable building design and engineering. For instance, solar VIS always provides benefits to indoor building energy savings (e.g., electrical lighting), while solar NIR is beneficial to building energy savings in winter but undesirable in summer. As a consequence, there is a need for reliable separate analyses focusing on individual solar radiation components. In this work, we explore and test classification-based modeling methods for decomposing hourly broadband global horizontal solar irradiance data in conventional weather files into hourly global horizontal solar NIR component. This model can then be conveniently implemented for sustainable building design and engineering purposes.