In the ancient day, earthquakes were attributed to supernatural power and interpreted as punishment by God for our sinful society [1,2]. Aristotle (384 − 322 B.C.) was the first person who explained earthquakes as a natural occurrence that is caused by changes in the underground structures of Earth. An earthquake is the trembling or shaking movement of the Earth’s surface, which occurs naturally or artificially. Artificial earthquakes occur due to the passing of a heavy vehicle over the road or an underground chemical or nuclear explosion. Natural earthquakes which are usually much stronger than artificial earthquakes are due to some internal changes within the Earth.
It is observed that the occurrence of massive earthquakes is usually a cyclic phenomenon- two consecutive seismic events are separated by a long aseismic period which may last for decades or even for centuries during which there are no seismic disturbances. During the seismic period which lasts only for a few seconds or a few minutes at the most, seismic waves of various types generated by earthquakes introduce considerable disturbances in the region, the free surface undergoes a movement that can be recorded by seismographs. On the other hand, it has been revealed by repeated geodetic surveys in seismically active regions of the earth, that there are slow, and aseismic surface movements involving horizontal and/or vertical displacements of the order of a few centimeters per year or less during the aseismic period.
The social-economic importance of earthquake prediction has been obvious for a long time. Powerful earthquakes are continually calamitous. An effective prediction of the occurrence of the next major seismic event may enable us to mitigate the loss of life and property [3,4]. So, predicting an earthquake is certainly a matter of high importance for research workers in seismology. Like any other physical science, a necessary pre-requisite for the development of an effective program of earthquake forecasting is a proper understanding of the dynamics of the earthquake processes in seismically active regions. In the middle of the 20th century, the theory of continental drift was practically rejected due to its limitations to explain properly the various causes of the earthquake, and this theory was replaced by the theory of plate tectonic [5]. Elastic rebound theory was introduced [6–7], according to that theory, major tectonic earthquakes are caused by the sudden release of elastic strain energy, stored in strained rock masses in some regions of the earth in the lithosphere, by dynamic fracture or fault movement, which occurs when the accumulated stresses become sufficiently large to overcome the local cohesive strength of the rock or the frictional forces which keep the faults in the locked state. In the 1990’s some great researchers like [8–10] after a considerable number of experiments upraise a quarry that a suitable method for earthquake prediction is to any extent possible or not. Though in 1975 the most eminent declaration of a successful prediction is reported for the Haicheng earthquake later states there exists no valid short-term prediction [11–16].
In seismology, Earthquake forecasting is the most popular branch concerned with seismic hazard, magnitude, and frequency of earthquakes [13,17]. The main objective is the faithful estimation of time, place, and magnitude was discussed [9] following [18–21]. According to [10,22] the predictions are evaluated by methods of statistical hypothesis testing with actual earthquakes. The occurrence of an earthquake is not simply homogeneous at all and for both space and time clustering occurs as described [23]. Earthquake prediction is tremendously tough and so many socio-economic issues are associated with this. A prediction is convenient only when it is perfect in both time and location. Certain precursory items have been pointed out that may have intense interconnection with the occurrence of a forthcoming earthquake and such precursory has been reviewed [24–25]. Abnormal behavior was observed in animals like cats, cattle, dogs, mice, rats, fish, birds, snakes, and so on before a massive earthquake. It has been described that they have come back to the P-wave discussed by [13, 26]. The precursor time may differ from a few minutes to various hours or even for several days, with escalated nervousness before an earthquake was studied by [26].
It has been observed, that during the seismically inactive period, gassy integrant of underground water and absorption levels of deliquescing minerals in seismically active regions remains almost constant. Accumulation of crustal strain observed by repeated geodetic surveys in seismically active regions may be identified as a precursory item before a major seismic event. Premonitory changes in the sea level have been reported just before the occurrence of the earthquake. Anomalous ground tilts a few months before a major shock has been reported particularly in China and Japan. The seismic “time gap” was elucidated as a signal of the forthcoming earthquake. In February 1975 the massive earthquake close to the city of Haicheng in Liaonping territory in China was efficiently forecasted following this seismic time gap demonstration discussed by [11,13,27] and many researchers. States that the next occurrence of stress release should be expected not in the segments in the seismic gap model where the unrelieved strain is the greatest [28–30].
A foremost precursory element to a forthcoming earthquake is changed in P-wave velocity (Vp), S-wave velocity (Vs) and Vp and Vs ratio discussed by [31]. Russian seismologist stated that the ratio Vp/Vs changed systematically at first it decreased by 5% then it returned to the normal values just before the earthquake [32]. A few earthquakes have been successfully forecast according to the analysis of foreshocks. In India Bhuj earthquake on January 2001 was also pursued by foreshocks in December 2000. In 2006, a study by [33] states essential records demonstrate the forecast was imperfect. Machine learning is one of the most robust consistent methodologies and is broadly used for earthquake predictions on account of their relevance to enhancement over time. With the extensive quantity of earthquake significant data, machine learning techniques are adequately proficient to enhance accuracy and efficiency in earthquake prediction. Different machine learning techniques along with, Support Vector Machine (SVM), Artificial Neural Network (ANN), Deep Neural Network (DNN), Naive Bayes (NB), Random Forest (RF), K-nearest neighbour (KNN) and Recurrent Neural Network (RNN) have been implemented for earthquake prediction. Temporal study on earthquake systematizes of Cyprus zone and evaluate sixty seismic indicators for the formation of short-term earthquake prediction applying ANN, SVM and RF were studied by [34].
In the Chinese Mainland to execute an estimation of the earthquakes [35] have used single multi-layer approach neural networks. A three-layer artificial neural network with Levenberg–Marquardt learning was introduced by [36] to represent the relationship between earthquake and radon. Using a machine learning method, thermal anomalies were observed before the occurrence of the earthquake in Imphal, India, in 2016, and analyzing different seismic certainty through satellite data for an earthquake was studied by [37]. [38] using R programming language applying different machine learning methods like SVM, KNN, RF, and NB algorithms for earthquake prediction. Earthquake signal identification using a neural network was introduced by [39] and the observing techniques used for earthquake prediction with particularized description was studied by [40]. A meticulous review of machine learning methods was introduced by [41–43] for earthquake prediction. Spatial and space-time prediction of strong earthquakes was studied by [44] making seismic hazards forecasts by using two different machine learning algorithms. [45] emphasize the challenges and scope introduced by big data for informed decision-making. Despite the significance of earthquake forecasting, it is even now a challenging problem for various reasons. Scientists are impelled to apply the progressively advanced approach to imitate earthquake processes for the complexity of earthquake forecast. The main purpose of this research is to find the range of the earthquake's magnitude.