Recent studies have demonstrated that the autofluorescence technique is a very promising tool for early-stage cancer diagnosis in various tissues. Many researchers have applied the autofluorescence technique through fiber-optic cables to excite tissues and collect the fluorescent emission signal from the tissues for discrimination analysis. In this study, we developed a Monte Carlo-light-induced autofluorescence (MCS LIAF) simulation model to optimize the oblique angle in the excitation optical fiber and the spatial arrangements in the receiving optical fiber. Our aim was to discriminate cervical tissues at different dysplastic stages. The model combined the structure of multi-layered tissues, tissue optical scattering and absorption parameters, tissue fluorophore concentration, the characteristic fluorescent spectrum of fluorophores, and the excitation and receiving arrangement of the optical fibers. The results show that the optimal oblique angle of the excitation optical fiber is between 0° and 45° and that the optimal receiving optical fiber is positioned 200 um away from the origin. We also observed that changing the excitation angle is very useful in differentiating normal from cervical intraepithelial neoplasia (CIN) I or CIN II tissues. Also, using the fluorescence peak ratio of NADH/collagen can help discriminate CIN III from normal tissues and CIN I/II tissues.