The prediction of first-day box office has been a hot research topic. However, existing research on movie prediction models lacks analysis and exploration of the holiday factor. In order to uncover the holiday factor contributing to the fluctuation of movie box office, we propose a prediction method for the first-day box office considering the Holiday Factor. Initially, we present a SARIMABox-Cox model for time series analysis. The experimental results reveal significant differences in movie box office during weekends and weekdays, as well as during summer vacation and non-summer vacation periods. Consequently, we introduce two variables, weekend dummy variable and summer vacation dummy variable, as holiday factors. Subsequently, to further investigate the impact of the holiday factor on the accuracy of first-day box office prediction, we integrate the holiday factor with multiple linear regression (MLR), random forest (RF), support vector machine regression (SVR), deep neural network (DNN), and extreme gradient boosting (XGBoost) to establish several first-day box office prediction models. Experimental results on the publicly available dataset Daily Earnings Data for 3400 Movies indicate a substantial improvement in the performance of models considering the holiday factor, with the highest enhancement reaching 16.42%.