Objective longitudinal monitoring of symptoms related motor fluctuations can provide valuable information for the clinical management of patients with Parkinson's disease. Current methods for long-term monitoring of motor fluctuations, such as patient diaries, are ineffective due to their time consuming and subjective nature. Researchers have shown that wearable sensors such as accelerometers can be used to gather objective information about a patient's motor symptoms. In this paper, we present preliminary results from our analysis on wearable sensor data gathered during longitudinal monitoring of 5 patients with PD. Our results indicate that it is possible to track longitudinal changes in motor symptoms by training a regression model based on Random Forests.