Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide. The five-year survival rate of HNSCC has not improved even with major technological advancements in surgery and chemotherapy. Currently, docetaxel, cisplatin, and 5-fluoruracil (TPF) treatment has been the most popular chemotherapy method for HNSCC; but only a small percentage of HNSCC patients exhibit a good response to TPF treatment. Unfortunately, at present, no reasonably effective prediction model exists to assist clinicians with patient treatment. For this reason, patients have no other alternative but to risk neoadjuvant chemotherapy in order to determine their response to TPF. In this study, we analyzed the gene expression profile in TPF-sensitive and non-sensitive patient samples. We identified a gene expression signature between these two groups. We further chose 10 genes and trained a support vector machine (SVM) model. This model has 88.3% sensitivity and 88.9% specificity to predict the response to TPF treatment in our patients. In addition, four more TPF responsive and four more TPF non-sensitive patient samples were used for further validation. This SVM model has been proven to achieve approximately 75.0% sensitivity and 100% specificity to predict TPF response in new patients. This suggests that our 10-genes SVM prediction model has the potential to assist clinicians to personalize treatment for HNSCC patients.