Terahertz time-domain spectroscopy (THz-TDS) has been applied to medical applications as THz waves are sensitive to biomolecules. In this paper, terahertz time-domain attenuated total reflection spectroscopy (THz-ATR) was used to identify human colon cancer cell line. The refractive indices of cancerous and normal cell lines with six different concentrations were analyzed in the frequency domain from 0.1-1.2THz. t-Distributed Stochastic Neighbor Embedding (t-SNE) and Principal Components Analysis (PCA) methods were adopted to extract the key features from the THz spectrum. Experimental results showed that the features extracted by t-SNE had better recognition accuracy owing to enhanced clustering efficacy. This work may be helpful for detecting human colon cancer cells by using terahertz techniques.