Human motion has been widely analyzed for different purposes such as human motion prediction, human identification. In this study, we propose a method for human style prediction based on the 3D marker coordinate. There are 28 subjects performed walking with non-weight condition and with tote bag of 5% of body weight. The 3D motion of each subjects was recorded using the motion capture system, which captured the trajectories of 43 reflective markers attached to the body of a patient. From the reflective markers, 17 joint coordinates were extracted using Vicon Nexus software. After that, two gait cycles of each walker were extracted and normalized in time and space. The Principal Component Analysis (PCA) method was applied to reduce the dimensionality and decompose the variance of the walking with different weight conditions of eigenvector with respective weight matrix. The linear transformation then was utilized to predict motion between normal walking and tote bag walking style. The mean error of the prediction for all markers was 39.6±18.1 mm, which was estimated using Euclidean distance of two corresponding markers between the measured data and the predicted data. This result shows a potential application for motion prediction in the future.