Extending upon joint modeling of response and response times (e.g., van der Linden, 2007; van der Linden, Klein Entink, & Fox, 2010), joint modeling of responses and response times has been proposed for cognitive diagnosis (Minchen & de la Torre, 2016; Zhan, Jiao, & Liao, 2017). Further, answer change data could be informative for cognitive diagnosis of test-takers’ strengths and weaknesses. The joint modeling of responses, response time, and answer change behaviors for cognitive diagnosis (Jiao, Ding, & Yin, 2020) has been proposed. The estimation accuracy in person-related parameters increased. However, due to the increase of data sources, the increase in the number of model parameters increased the computational burden in joint modeling. This study explores to use some machine learning algorithms including both supervised and unsupervised algorithms to analyze responses, response time, and answer changes for cognitive diagnosis. The results will be compared with the joint modeling approaches of the three types of data. Both simulation study and real data analysis will be conducted to investigate cognitive classification decision under different study conditions.