Comparative opinions widely exist in online reviews as a common way of expressing consumers’ ideas. Meanwhile, such online opinions are key proxies for detecting product competitiveness. Firstly, comparative opinion pairs are extracted in feature-levels through text mining and sentiment analysis. Then, based on the comparative opinion pairs, a single-link graph, a dichotomic-link graph and a multi-link graph are built respectively, where the weights of edges are determined by the sentimental strength. Next, a feature-level comparative network is calculated by employing sophisticated network algorithms, including PageRank and Hyperlink-Induced Topic Search. The proposed comparative network can identify the strengths and weaknesses of compared products. Experimental results show that the feature-level comparative networks are correlated with SalesRank significantly, thus enabling the prediction on sales volume.