Early characterization and grading of liver fibrosis are important because this condition can progress into cirrhosis, which is irreversible unless discovered timely and effectively treated. In this study, we aim to extract new features regarding the dynamics of the time series of ultrasound (US) radio frequency (RF) data and examine their effectiveness for noninvasive, cost-effective, and rapid grading of early liver fibrosis. We propose the combination of spectral and fractal features with the time-domain features of US RF time series, which were averaged over a region of interest. Experiments on early liver fibrosis staging were conducted using two classifiers, namely, support vector machines (SVM) and random forests. Experimental results showed that the proposed method achieved the highest classification accuracy of 96.67% and an average classification accuracy of 77.33% for differentiating the stages of liver fibrosis by using random forest. Hence, RF time series can be used for in vivo tissue characterization of liver fibrosis. This study describes a promising tool for non-invasive early detection and grading of liver fibrosis.