Coffee is considered as a functional food due to its being rich in bioactive compounds, mainly chlorogenic acid (CGA). CGA concentration in coffee aqueous solution was investigated based on visible/near-infrared (Vis/NIR) spectroscopy in this research. To enhance the spectral difference among different samples and increase the signal to noise ratio, Lorentz function curve fitting was applied to fit raw Vis/NIR spectra of samples. Then, the fitting parameters were used to correct raw full spectra. Partial least squares (PLS) regression method was used to develop calibration models of CGA concentration. Full-spectrum models were built with raw and fitting parameter-corrected spectra, respectively. Further, wavelength selection methods, such as genetic algorithms (GAs) and success projection algorithms (SPAs), were applied to eliminate redundancy information and identify relevant information from full spectra. Calibration models based on the effective wavelengths selected by GA and SPA methods were developed. The overall results showed that LFPs a/b-corrected spectra had a better performance compared with other processing methods. Performance of the selected wavelength model was better than that of the full-spectrum model. Final results indicated that the SPAs-PLS method provided a more precise prediction model of CGA concentration with R (c) of 0.913 and R (cv) of 0.795.