Based on the multiple-vantage observations of STEREO, SOHO, wind and other spacecraft, the fast and wide coronal mass ejections (CME) during the 24th solar cycle from January 2010 to September 2014 are selected in this paper. Using the outputs of Richardson’s (2014) empirical model of solar energetic particle (SEP) intensity under different conditions, the effects of its associations such as CME, pre-CME, and type II radio bursts, on SEP intensity are analyzed, and the relationship between SEP event and these characteristics is also discussed. The main conclusions are as follows. 1) The presence or absence of pre-CME within 13 h before fast CME significantly improves the model prediction effect and has a significant influence on whether fast CME produces SEP event. Compared with the events without pre-CMEs, the events with pre-CMEs have a low proportion of false alarms (FR: 47.7% vs.70%). However, the number of pre-CMEs does not improve the model output. 2) CMEs with type-II radio bursts have significantly lower FR to generate SEP events than fast CMEs without type-II radio bursts (42% vs.68%). And selecting type-II radio bursts as a constraint will filter out some small/weak SEP events, the relationship between model predictions and observations especially for large SEP events (e.g. I _p≥ 0.01 pfu/MeV) will stand out. Moreover, if the type-II radio enhancement is taken into account, FR can be further reduced to 29.4%, and the proportion of hits can be further increased (HR: 48.5%), and the model prediction is significantly improved. 3) The larger the start frequency of type II radio bursts, the smaller the end frequency is, and FR decreases slightly, but at the same time, a large number of SEP events are excluded by this condition, and the results show that the constraints on the start/end frequency of type-II radio bursts do not improve the model predictions distinctly. 4) If the sub-classification of type-II radio bursts is considered as the model constraint, the CMEs associated with multi-band type-II radio bursts have better model predictions than those with single-band events. For example, m-DH-km type-II radio bursts have lower FR (35.4%) and higher HR (48%), and the accuracy of empirical model is higher. In summary, we find that in addition to the velocity and angular width of CME, the associations of pre-CME, type II radio bursts and their enhancement, and multi-band sub-classification are the favorable conditions for CME to generate SEP events. The SEP intensities obtained by the empirical model have better consistency with the observations, and better predictions can be obtained. This investigation indicates that SEP events are more likely generated by fast and wide CMEs accompanied by pre-CMEs, multi-band type II radio bursts and their enhancements, which seem to serve as discriminative signal for SEP-rich and SEP-poor CMEs.