Background: Earthquake prediction plays an important role in preventing catastrophic damage. We present here a scheme for earthquake prediction using an improved version Tree Augmented Naïve Bayes (TAN). This approach can achieve an appropriate performance by extracting dependencies among seismicity features effectively. At first we use the simplest Discretization Method, Equal Interval Width, for discretization of six seismicity features (Time, Mean Magnitude, Energy, Slope, deviation, Magnitude deficit) drive from National Earthquake Information Center (NEIC), then in order to the magnitude of earthquake we utilize an improve version of Tree Augmented Naïve Bayes that is based on Decomposable Models.
Results: Finally, we test our method by using two schemes and compare it to Tree Augmented Naïve Bayes and Hidden naive Bayes classifier. F-measure metric is used for evaluation of our results.
Conclusion: Experimental results demonstrated proposed approach based on an improved version of Tree Augmented Naïve Bayes achieves a higher performance compared to two other methods.