The human body is prone to numerous neurological abnormalities that deteriorate movement and coordination. One such condition is Parkinson's disease (PD), which primarily attacks the older masses and causes them discomfort in the latter phases of their lives. Freezing of Gait (FOG), being a special symptom of PD, deteriorates mobility and degenerates quality of life. This study analyzed three phenomena known as FOG events, namely, Start Hesitation, Turn, and Walking. The gaps in existing research highlighting the scarcity of understanding of the mechanisms and factors leading to FOG events have been addressed and eliminated. This research aims to predict which of the three movements will most likely be affected. A four-step sequential approach is proposed that is initiated by the exploration of patients’ accelerometer tests. The preprocessed, evenly sampled, and outlier-free data is then subjected to feature engineering. Following metadata integration, the feature-engineered dataset is fed to the Light Gradient Boosting Machine (LGBM) due to its ability to handle extensive datasets. The suggested model computes the probability of each of the three FOG events being impaired in a patient diagnosed with Parkinson's disease. Precision was evaluated in the presence and absence of outliers for a comparative analysis. In the absence of outliers, LGBM gave a superior precision of 0.7354, 0.8225, and 0.2689 for start hesitation, turning, and walking, respectively.