In online programming education, if teachers can determine any difficulties their students are experiencing and provide support, it would significantly improve the outcome of their teaching. This paper describes an attempt to build a time prediction model on the demand for personalized affective support based on a modified version of the Synthetic Minority Over-sampling Technique. We designed and conducted a data collection experiment based on the specific features of the affective support. Meanwhile, the modified oversampling algorithm can ascertain the time for providing such support for learners, which solves the problem of a class imbalance distribution. In addition, we obtained a sorting algorithm of the time prediction regarding the demand for personalized affective support in programming learning and constructed a time prediction model on the demand for affective support. Meanwhile, we conducted experiments on both public data and our own collected data to verify the effectiveness of the constructed model. The results show that the model is able to judge whether learners need affective support during the writing code process.