Earthquake prediction with meteorological data by particle filter-based support vector regression

Figures & data

Table 1. Comparison of several system (Mehdidoust & Shahbahrami, 2016).

Source	Earthquake magnitude	Probability of an earthquake	Used features
Ikram & Qamar (2014)	M ≥ 3.6	Accuracy ≥95%, 15 hours before occurrence	Magnitude, Intensity, Latitude and Longitude, Depth
Ikram & Qamar (2015)	M ≥ 3.6	Accuracy ≥95%, 12 hours before occurrence	Magnitude, Intensity, Latitude and Longitude, Depth
Moustra, Avraamides, & Christodoulou (2011)	All	80.55%	Magnitude, Electrical shock signal
Moustra et al. (2011)	M ≥ 5.2	84.01%	Magnitude, Electrical shock signal
Dehbozorgi & Farokhi (2008)	–	82.85%	Magnitude, Latitude and Longitude, Depth, Time
Sikder & Munakata (2009)	M ≥ 1.5	88.39%	Magnitude, Radon concentration in soil, Weather features
Aminzadeh, Katz, & Aki (1994)	M ≥ 5	73%	Magnitude
Aminzadeh et al. (1994)	All	67%	Magnitude

Figure 1. C value.

Figure 2. -tube (Insom et al., 2015)

Figure. 3. Data flow related to the proposed method

Table 2. Parameters set in proposed method.

Parameter	Amount	Meaning
VarLow	1	Lower limits of unknown parameters
VarHigh	100	Upper limit of unknown parameters
F	0	Exit Flag
E	0.01	Noise & terminate condition
T	20	Time steps
nPart	5	Number of Particles
Kernel Function	“Gaussian”	Kernel function
Epsilon	Unknown	Loss Function Parameter
Kernel Scale	Unknown	Kernel width(sigma)
C	Unknown	Tradeoff between educational error and the complexity of the model

Figure 4. The earthquakes recorded within the studied area (www.irsc.ut.ac.ir)

Table 3. Collected data details.

Row	Feature name	Minimum	Maximum	Mean	Unit	Used as
1	Average monthly temperature	−6.5	28.8	13.263	C	Input
2	Maximum monthly temperature	7.4	41	25.7157	C	Input
3	Minimum monthly temperature	−18	18.2	1.9417	C	Input
4	Average monthly wind speed	2.1	5.6	3.6037	Knot	Input
5	Monthly rainfall	0	83.2	20.4222	mm/month	Input
6	Number of earthquakes per month	0	477	12.8241	Count	Output
7	Average magnitude of earthquakes in month	0	4.4	2.8067	Richter Scale	Output

Figure 5. The comparison of actual and estimated values for mean magnitude.

Figure 6. Coincidence of estimated mean magnitude of earthquakes (R index).

Figure 7. The comparison of actual and estimated number of earthquakes.

Figure 8. The coincidence of estimated number of earthquakes (R index).

Table 4. The comparison of regression results for the mean magnitude of earthquakes.

Modeling tools	R	MSE	RMSE
Multilayered Perceptron Neural Network	0.88	0.00048	0.0219
Support vector regression (SVR)	0.2	0.002	0.0447
SVR-optimized by particle filter	0.96	0.0015	0.0387

Table 5. The comparison of regression results for the number of earthquakes.

Modeling tools	R	MSE	RMSE
Multilayered Perceptron Neural Network	0.5	0.00018	0.0134
Support vector regression (SVR)	0.12	0.00027	0.0164
SVR-optimized by particle filter	0.78	0.017	0.1303

Mehdidoust, J. Z., & Shahbahrami, A. (2016). Study Of Expert Systems In Predicting Earthquake.

 Google Scholar

Ikram, A., & Qamar, U. (2014). A rule-based expert system for earthquake prediction. Journal of Intelligent Information Systems, 43(2), 205–230. doi: 10.1007/s10844-014-0316-5

 Web of Science ®Google Scholar

Ikram, A., & Qamar, U. (2015). Developing an expert system based on association rules and predicate logic for earthquake prediction. Knowledge-Based Systems, 75, 87–103. doi: 10.1016/j.knosys.2014.11.024

 Web of Science ®Google Scholar

Moustra, M., Avraamides, M., & Christodoulou, C. (2011). Artificial neural networks for earthquake prediction using time series magnitude data or seismic electric signals. Expert Systems with Applications, 38(12), 15032–15039. doi: 10.1016/j.eswa.2011.05.043

 Web of Science ®Google Scholar

Dehbozorgi, L., & Farokhi, F. (2008). Effective feature selection for short-term earthquake prediction using Neuro-Fuzzy classifier –, Centran Tehran Branch, Sci. Assoc. of Electr. & Electron. Eng., Islamic Azad Univ.

 Google Scholar

Sikder, I. U., & Munakata, T. (2009). Application of rough set and decision tree for characterization of premonitory factors of low seismic activity. Expert Systems with Applications, 36(1), 102–110. doi: 10.1016/j.eswa.2007.09.032

 Web of Science ®Google Scholar

Aminzadeh, F., Katz, S., & Aki, K. (1994). Adaptive neural nets for generation of artificial earthquake precursors. IEEE Transactions on Geoscience and Remote Sensing, 32(6), 1139–1143. doi: 10.1109/36.338361

 Web of Science ®Google Scholar

Insom, P., Cao, C., Boonsrimuang, P., Liu, D., Saokarn, A., Yomwan, P., & Xu, Y. (2015). A support vector machine-based particle filter method for improved flooding classification. IEEE Geoscience and Remote Sensing Letters, 12(9), 1943–1947. doi: 10.1109/LGRS.2015.2439575

 Web of Science ®Google Scholar