Earthquakes are a natural disaster which causes loss of life and property. For this reason, earthquake professionals show great effort for reliable earthquake predictions. Earthquake attempts to guess with techniques rely on different modeling, approach, and certain precursors for seismic activity. These predictions in current technology can also improve by using various machine learning techniques. Because machine learning techniques such as association rule mining facilitate the interpretation of many complex problems, particularly for large databases. Furthermore, the techniques can develop more reliable approaches by combining previously acquired information. The aim of this study is to predict the key conditions influencing soil amplification from the data of the 8400 ground motion records for 100 different soil profiles. This is achieved by applying association rule mining on 31 different parameters related to soil profiles and the ground motion records with respect to amplification levels. The rules generated for predicting soil amplification are mathematically validated. Results from the proposed rule-based predictions for soil amplification show the most effective parameters (conditions) related to ground motion and soil such as frequency content, period and intensity. Additionally, probabilities of soil amplification and damping respect to soil type and peak ground acceleration were determined. These findings may provide valuable insights for future research on soil amplification.