In this study, we proposed a combined approach, which amalgamates machine learning and lexicon- based approaches for multiple-domain sentiment classification that supports Cantonese-based social media analysis. Our study contributes to the existing literature not only by investigating the effectiveness of the proposed combined approach for supporting social media analysis in the Cantonese context but also by verifying that the proposed method outperforms the baseline approaches, which are commonly used in the literature. We demonstrated that social media network-based classifiers can be general classifiers that support multiple-domain sentiment classification.