The detrended fluctuation analysis (DFA) method is used to analyze the human on-line activities in e-commerce. We comprehensively investigate the scale laws of browse and purchase behaviors, which have received little attention before. The time series of browse and purchase behaviors each obviously show a periodical character, and their probability density distributions each have a significant bimodal form. Based on the Fourier transform method, the power spectra of time series indicate that each of them obeys a stochastic process with a long-range self-similar feature (i.e., deviation far from the Poisson process). After identifying and filtering the influence of periodic trend based on power spectra, the detrended fluctuation analysis is used to study the scaling law of time series. Several interesting results can be found that their scaling behaviors on small and large scales show similar values that confirm the long-range correlations rooting in the time series of human on-line activities, and their average scaling exponent approximately equaling 1 suggests that the human online activity may be associated with a self-organized criticality. Although the empirical results are only the observed phenomena like those found in the Internet traffic and stock price fluctuation of financial market, we still think that they may provide an important insight to deeply understand the mechanism of human dynamic behaviors in e-commerce and predict their fluctuation trend for the potential business application.