2 دانشیار گروه مهندسی کامپیوتر، دانشکده فنی و مهندسی، دانشگاه گیلان، رشت، ایران
عنوان مقاله [English]
The crisis in the event of an incident or accident occurs suddenly and unexpectedly that urgent attention is needed for proper decision making. Despite technological advances, the suffering caused by natural disaster such as earthquake, flood, avalanche, hurricane, volcano, fire and there is abnormal. Mining activities are always risks with the risks of mine called, has been associated. Seismic hazard in underground mines is one of the threats to human life. Seismic hazard identification much more difficult to identify the natural hazards to an earthquake. Using advanced seismic earthquake and Seismoacoustic predict. The accuracy of the information generated is not optimal. The complexity of seismic processes and the disproportion between the low-energy seismic events and a number of high-energy phenomena (eg greater than 10 ^ 4 J) makes statistical techniques to predict the seismic hazard is not enough. So the search for better opportunities for risk prediction has become imperative. Clustering techniques for seismic hazard data and artificial neural networks were used to predict. In most applications, the results obtained by the methods listed in the situation "dangerous" and "safe" have been reported. Unbalanced distribution of positive samples (dangerous situation) and negative (safe condition), is a serious problem in seismic hazard forecasting. Currently, the methods used can not be expected to take appropriate and high sensitivity. A number of factors related to the risk of earthquakes, was proposed. Among other factors, the shake with energy greater than 10 ^ 4 J mentioned earthquake prediction can be defined with different methods. But the main goal is for all methods of seismic hazard assessment. In some cases, the data on the time and date, an increase in seismic activity that could lead to the destruction of rocks using pressure (Rockburst), the predicted.
You can not focus on one parameter, occurrence or non-occurrence of earthquakes in the area or within the time limit specified. Should study several parameters at the same time be an acceptable technique for predicting earthquake found that one way to do this is to use expert system. The news systems provided that these systems have the ability to use more, better and smarter them. For example, in hydrology analysis only considered changes in water ions other parameters such as fault behavior, changes in sea level, etc., are not considered to predict earthquakes. On the other hand all expert systems to predict earthquakes from Precursor not use to predict. Some expert systems by the time information, location and depth of previous earthquakes, earthquakes predict the future. Studies have shown that large earthquakes decision tree and artificial neural network down accurately predicted. In this article by support vector machines based on particle swarm big earthquakes with higher accuracy than was some other expert systems.
Expert systems commonly used in earthquake prediction. In these expert systems, various parameters are used such as fault behavior, the concentration of radon, energy, pulse and the number of bumps. The occurrence of earthquakes can be measured by checking these parameters. The accuracy of earthquake prediction by expert systems is relatively higher than non-expert system methods. In this paper, the accuracy and the kinds of data used in different expert systems in the field of earthquake prediction were studied. In addition, different algorithms to predict earthquake with an expert system based on support vector machines, decision trees, neural networks, support vector machines based on particle swarm optimization, Bayesian and MLP network was implemented in Rapidminer. The results show that support vector machine-based expert system that is optimized by particle swarm algorithm in comparison to neural network-based expert systems, support vector machines, Bayesian, decision tree and network MLP has a better prediction accuracy.