Sunspot observations since 1610 A.D. show that the solar magnetic activity displays long-term changes, from Maunder Minimum-like low-activity states to Modern Maximum-like high-activity episodes, as well as short-term variations, such as the pronounced 11-year periodicity. Information on changes in solar activity levels before 1610 relies on proxy records of solar activity stored in natural archives, such as 10Be in ice cores and 14C in tree rings. These cosmogenic radionuclides are produced by the interaction between Galactic cosmic rays (GCRs) and atoms in the Earth's atmosphere; their production rates are anti-correlated with the solar magnetic activity. The GCR intensity displays a distinct 11-year periodicity due to solar modulation of the GCRs in the heliosphere, which is inversely proportional to, but out of phase with, the 11-year solar cycle. This implies a time lag between the actual solar cycles and the GCR intensity, which is known as the hysteresis effect. In this study, we use the North Greenland Ice Core Project (NGRIP) records of the 10Be flux to reconstruct the solar modulation strength (Φ), which describes the modulation of GCRs throughout the heliosphere, to reconstruct both long-term and subdecadal changes in sunspot numbers (SSNs). We compare three different approaches for reconstructing subdecadal-scale changes in SSNs, including a linear approach and two approaches based on the hysteresis effect, i.e. models with ellipse-linear and ellipse relationships between Φ and SSNs. We find that the ellipse approach provides an amplitude-sensitive reconstruction and the highest cross-correlation coefficients in comparison with the ellipse-linear and linear approaches. The long-term trend in the reconstructed SSNs is computed using a physics-based model and agrees well with the other group SSN reconstructions. The new empirical approach, combining a physics-based model with ellipse-modeling of the 11-year cycle, therefore provides a method for reconstructing SSNs during individual solar cycles based on 10Be in ice cores. This, in turn, represents a new window for studying short-term changes in solar activity on unprecedented timescales, which may help improve our understanding of the solar dynamo.