With the continuous development of genomics and precision medicine, targeted therapy and immunotherapy targeting biomarkers have ushered in a new era of anti-tumor therapy. However, due to the heterogeneity of tumor cells and the variability of tumor microenvironment, there are still significant differences in response to the same drug even in patient populations with the same biomarker enrichment. By combining omics data with drug sensitivity algorithms, the response of anti-tumor drugs can be predicted and transformed into personalized diagnosis and treatment strategies required for precision medicine, which is expected to improve the effectiveness of anti-tumor drugs in clinical treatment. Currently, machine learning is one of the commonly used modeling algorithms for predicting the response of anti-tumor drugs. However, due to differences in input data and algorithm construction methods, there is currently a lack of comprehensive literature review in this field. Therefore, this article provides a review of machine learning algorithms for predicting anti-tumor drug responses, summarizing publicly available cell genome characterization datasets, machine learning algorithms, and evaluation indicators in drug response prediction, as well as the current situation and challenges faced in clinical applications, in order to provide methodological references for the main research problems and potential solutions of machine learning algorithms in the field of drug response prediction.