The intention forecasting of the human driver plays an essential role in improving the safety and driver-machine coordination of the co-piloting vehicle, which is the decision-making foundation for driver assistance system (DAS). However, the quantitative prediction of the lane change decision, which determines the occasion and shape of the lane change path, has drawn little attention by the prior works. In this paper, a novel convolutional neural network (CNN)-based predictor that forecasts the driver’s lane change path is proposed. By including the driver’s eye-gazing and head-heading information, an earlier and more accurate prediction of driver’s time to lane change (TLC) and lane change duration (LCD), and we use the positive TLC and negative TLC to represent the classification results LC and RC, respectively. The classification and regression results validated that the inclusion of driver’s gazing information achieved an earlier and more precise detection of driver’s lane change intention. Furthermore, owning to the hierarchical structure of the CNN model and the parallel computing devices, the execution time of the proposed model is shorter and exhibits its powerful real-time computing performance. The estimated TLC and LCD are adopted to predict the driver’s target lane change trajectory with satisfactory accuracy. Therefore, the proposed model enables the DAS to generate a human-like lane change decision to minimize the driver-machine conflicts during lane change.