A fall on the same level is a leading causes of non-fatal injuries among construction workers. Previous research reveals that such incidents are associated with slip, trip and loss of balance (STL) events often caused by unsafe site conditions (e.g., slippery floors, obstacles on the path and uneven surfaces). Consequently, detecting STL events enable site management to identify these hazards and employ suitable risk mitigation "control" measures. This research examined foot plantar pressure distribution for automated detection and classification of STL events using wearable insole pressure sensors. Three volunteers participated in a laboratory controlled simulated experiment that examined different types of STL events, while the corresponding foot plantar pressure data were collected from wearable insole pressure sensors. Diverse features (e.g., time- And frequency-domains, and spatial-temporal features) were extracted from the foot plantar pressure distribution data, which was used to associate different pressure patterns with each type of STL event. Four machine learning classifiers [i.e., artificial neural network (ANN), decision tree (DT), K-nearest neighbor (KNN), and support vector machine (SVM)] were evaluated to select the best classifier. Cross validation results revealed that at approximately 85% of classification accuracy, the KNN classifier achieved the most accurate result using 0.64s window size, indicating a great potential to use the proposed approach to automate fall risk detection. Overall, this method would allow construction managers to understand how workers react to unsafe conditions associated with STL events, so as to minimize the fundamental causes of STL events and thus to reduce non-fatal fall injuries in construction.