Understanding the impacts of aerosol chemical composition and mixing state on cloud condensation nuclei (CCN) activity in polluted area is crucial for determining CCN number concentrations (N<sub>CCN</sub>) accurately. In this study, we predict CCN number concentrations (N<sub>CCN</sub>) by applying κ-Köhler theory under five assumed schemes of aerosol chemical composition and mixing state based on field measurement in Beijing during the winter of 2016. Our results show that the EIS scheme (with an assumption that sulfate, nitrate, and secondary organic aerosols are internally mixed and that primary organic aerosols, POA, and black carbon, BC, are externally mixed; and the chemical composition is size dependent) achieves the best closure to predict N<sub>CCN</sub> with ratios of predicted-to-measured N<sub>CCN</sub> (R<sub>CCN_p/m</sub>) of 0.90–1.12 under both clean and polluted conditions over the campaign. Also, IB scheme (with an assumption of internal mixture and bulk chemical composition for particles) shows good closure with R<sub>CCN_p/m</sub> of 1.01–1.19 under clean conditions, implying that the IB assumption is sufficient for CCN prediction in continental clean regions. On polluted days, IS scheme (assuming particles with internal mixture and chemical composition is size-resolved) achieve better closure than the IB scheme due to the heterogeneity and variations in particle composition at different sizes. The improved closure achieved using EIS and IS assumptions highlights the importance of measuring size-resolved chemical composition for CCN predictions in polluted regions. NCCN is significantly underestimated (with R<sub>CCN_p/m</sub> of 0.6–0.8) by using the schemes of external mixture with bulk (EB) or size-resolved composition (ES), implying that the primary particles experience rapid aging and physical mixing processes in urban area. However, our results show that the mixing state of particles plays a minor role on CCN prediction when the κ<sub>org</sub> exceeds 0.1.