Radon is a well-known precursor for geodynamic events such as earthquakes and volcanic tremors. Radon concentration variations in soil gas have been monitored worldwide, and extreme radon values have been identified as radon anomalies associated with geodynamic events. A radon time series contains many noise signals, primarily based on meteorological effects. Therefore, detecting anomalies from values outside the mean plus a few standard deviations or from values outside the average distribution threshold may not always yield good results. Instead of analyzing specific radon anomalies, an alternative method can be used to analyze the trend changes in the radon time series. This study examines locally estimated scatterplot smoothing (LOESS) to identify changes in the trend of the radon time series. During the two-year period of measurements, two separate groups with radon concentration anomalies and anomaly mechanisms were identified. In the first group, radon increases before the earthquake and decreases after the earthquake, while in the second group it shows the opposite behavior.