基于粒子群算法的地震预报方法研究

Earthquake Prediction Method Based on Particle Swarm Optimization

摘要:

针对地震预测中定量计算的困难性，利用地震前兆异常高维数据特征，研究一种基于粒子群聚类算法的地震预报模型。该模型输入为条带、空区、短水准等14项异常指标数据，输出为震级分类。模型设定聚类平均距离为粒子群算法的评价函数，发掘分析地震前兆数据与地震震级的关系。结果表明该模型能有效地根据地震前兆数据预测地震震级，与传统聚类k-means算法模型相比，稳定性强，预报准确性更高。历史地震数据实例研究表明，本文提出的模型充分利用了粒子群算法的高鲁棒性、高适应性和群体智能的协同策略，是改进地震预报效能的途径之一。

Abstract:

It is difficult to quantitatively calculate or predict earthquakes in advance；however，in areas of high density data regarding earth characteristics and monitoring prediction may be possible.This paper presents an earthquake prediction model that is based on the Particle Swarm Optimization Algorithm.The inputs of this model consist of 14 items，which are abnormal index data，they include banding，dead zone，short leveling，and so on，and the output is the classification of the earthquake magnitude.This model sets the average distance of cluster as the evaluation function of Particle Swarm Optimization Algorithm，explores and analyzes the relationships between pre earthquake precursor data and earthquake magnitude.The specific steps of the algorithm are stated as follows：Firstly，we normalize the original data of earthquake cases，which eliminates the dimensional effect；Secondly，we initialize the model parameters using reasonable values from earthquake cases；Thirdly，we pick up the speed through applying the Particle Swarm Optimization Algorithm and design the update strategy；and Finally，we design the evaluation function.If the algorithm satisfies the evaluation function，the algorithm needs to be stopped，and output the optimal solution；otherwise，it needs to turn to the third step.To verify and prove the correctness and efficiency of earthquake forecasting that is based on Particle Swarm Optimization Algorithm，an experiment in the environment of Matlab 2007a is conducted and a comparison with the classical k means Clustering Algorithm is made.The experimental data are divided into 3 categories，among which，Category 1 represents Magnitude 5~6 of earthquake，Category 2 represents Magnitude 6~7 of earthquake，and Category 3 represents Magnitude 7 and greater earthquakes.As for the accuracy rate，the overall forecast accuracy rate of k-means Algorithm is only 73.3 %；however，Particle Swarm Optimization Algorithm can increase the accuracy rate up to 83.3 %. To analyze the stability of the algorithm and the complexity of time，the results of five experiments were randomly selected to calculate the average distance among clusters and the processing time. Through analysis，we show that the processing time of the Particle Swarm Optimization Algorithm is slightly longer than the processing time for k means Algorithm.The average distance using Particle Swarm Optimization Algorithm to conduct cluster is smaller than that of k means Algorithm，this indicates that Particle Swarm Optimization Algorithm has an advantage of better stability than the classical k means Clustering Algorithm.The experimental results indicate that，this model can effectively predict the earthquake magnitude in accordance with the earthquake precursor data.Compared with the traditional cluster k means algorithm model，Particle Swarm Optimization Algorithm is much stronger，and the forecast accuracy is much higher.The research and analysis of the example of historical seismic data indicate that the model suggested in this paper makes full use of the highly robust and flexible Particle Swarm Optimization Algorithm as well as the coordination strategy of swarm intelligence.This represents a promising approach to improve the efficiency of earthquake forecast.