Academia.eduAcademia.edu

Figure 6 – uploaded by abdullah al hasan

See full PDFdownloadDownload figure
Receiver Characteristics Curves (ROC) are commonly applied to determine the accuracy of a predic- tive model. Therefore, this model has been considered to measure the accuracy of the BRBES’s outputs. In this model, Area under Curve (AUC) is considered as one of the important metrics. When the value of AUC becomes one then it can be concluded that the accuracy of the prediction is 100% correct. The earthquake magnitude of 6.8 of the original data has been considered as the baseline data. When an earthquake with more than “6.8” is found then the benchmark value is considered as 1, otherwise it is considered as “0”. Column 13 of Table 6 shows the benchmark data, which has also been used to gener- ate ROC curves. SPSS 23 has been used to generate the ROC curves.

Table 6 Receiver Characteristics Curves (ROC) are commonly applied to determine the accuracy of a predic- tive model. Therefore, this model has been considered to measure the accuracy of the BRBES’s outputs. In this model, Area under Curve (AUC) is considered as one of the important metrics. When the value of AUC becomes one then it can be concluded that the accuracy of the prediction is 100% correct. The earthquake magnitude of 6.8 of the original data has been considered as the baseline data. When an earthquake with more than “6.8” is found then the benchmark value is considered as 1, otherwise it is considered as “0”. Column 13 of Table 6 shows the benchmark data, which has also been used to gener- ate ROC curves. SPSS 23 has been used to generate the ROC curves.

Related Figures (11)

Geo-science, although the identification of earthquake patterns and clusters has been investigated by the researchers of various countries for long time [7][8][9][10][11][12][13][14][15][16].
Geo-science, although the identification of earthquake patterns and clusters has been investigated by the researchers of various countries for long time [7][8][9][10][11][12][13][14][15][16].
Figure 1: The BRB framework for predicting earthquake.
Figure 1: The BRB framework for predicting earthquake.
in this case Eq. (5) will be applied, otherwise Eq. (4). Thus, by applying Eq. (5), the matching degrees for this input data (low) can be obtained for Low as 0.8 (Low = (50 — 10)/(S0 — 0)= 0.8) for ”>Medium“ as 0.2 (Medium = | — 0.8 = 0.2) and for High as ”0“, which are illustrated (see Table 2). When the referential categories are assigned with matching degrees then the rule is called packet antecedent and hence, it is considered as active. 2. Rule Activation Weight calculation The activation weight calculation of a rule comprises calculating the combined matching degree, which is obtained by using Eq. (6) [46] as well as by calculating activation weight, which is obtained by applying
in this case Eq. (5) will be applied, otherwise Eq. (4). Thus, by applying Eq. (5), the matching degrees for this input data (low) can be obtained for Low as 0.8 (Low = (50 — 10)/(S0 — 0)= 0.8) for ”>Medium“ as 0.2 (Medium = | — 0.8 = 0.2) and for High as ”0“, which are illustrated (see Table 2). When the referential categories are assigned with matching degrees then the rule is called packet antecedent and hence, it is considered as active. 2. Rule Activation Weight calculation The activation weight calculation of a rule comprises calculating the combined matching degree, which is obtained by using Eq. (6) [46] as well as by calculating activation weight, which is obtained by applying
value Snouid be 1n DeTween U tO Ll. From Table 3, it can be observed that rule “6” consists of three antecedent attributes with their indi- vidual matching degrees, which need to be combined, by applying Eq. (6). The importance of this rule to calculate the unusual behavior of the animal can be acquired by applying Eq. (7). The implication of this value is that this rule has an important impact in getting the result or it is highly sensitive. 3. Modified Belief Degree There could be a situation when input data for all the antecedent attributes of a BRB cannot be available and this phenomenon can be considered as ignorance. In this situation, the degree of belief of the original BRB needs to be modified, which can be obtained by using Eq. (8). where 6,; is the normalized antecedent attribute weight, obtained by dividing the individual an- tecedent attribute weight by the summation of all antecedent attribute weights of a rule. Hence, its value should be in between 0 to 1. EE op ek: ky RE NT ee ee ee ee Le a) ee ee ee ee ee es, Se: cee
value Snouid be 1n DeTween U tO Ll. From Table 3, it can be observed that rule “6” consists of three antecedent attributes with their indi- vidual matching degrees, which need to be combined, by applying Eq. (6). The importance of this rule to calculate the unusual behavior of the animal can be acquired by applying Eq. (7). The implication of this value is that this rule has an important impact in getting the result or it is highly sensitive. 3. Modified Belief Degree There could be a situation when input data for all the antecedent attributes of a BRB cannot be available and this phenomenon can be considered as ignorance. In this situation, the degree of belief of the original BRB needs to be modified, which can be obtained by using Eq. (8). where 6,; is the normalized antecedent attribute weight, obtained by dividing the individual an- tecedent attribute weight by the summation of all antecedent attribute weights of a rule. Hence, its value should be in between 0 to 1. EE op ek: ky RE NT ee ee ee ee Le a) ee ee ee ee ee es, Se: cee
Table 4: Belief Degrees Update From Table 4, it can be observed that the original belief degrees of rule no. 6 have been modifie since the input data of ” Vocal response” antecedent attribute is absent. The updated values of the belie degrees are obtained by applying Eq. (8). 4. Rule Aggregation The rules of BRB need to be aggregated to obtain output data in response to the input data. As an instance, the input data of X7BRB consists of [X1=Low, X2 = Medium, and X3 = High]. The output value, i.e. the value of X7 consequent attribute, needs to be calculated in response to these input data, which can be achieved by aggregating the rules associated with X7BRB. The ER (Evidential Reasoning) inference mechanism is applied to obtain this overall calculative value in terms of fuzzy values. There are two forms of ER, namely, recursive and analytical. The analytical ER is considered to reduce the computational complexity as shown in Eqs. (9) and (10).
Table 4: Belief Degrees Update From Table 4, it can be observed that the original belief degrees of rule no. 6 have been modifie since the input data of ” Vocal response” antecedent attribute is absent. The updated values of the belie degrees are obtained by applying Eq. (8). 4. Rule Aggregation The rules of BRB need to be aggregated to obtain output data in response to the input data. As an instance, the input data of X7BRB consists of [X1=Low, X2 = Medium, and X3 = High]. The output value, i.e. the value of X7 consequent attribute, needs to be calculated in response to these input data, which can be achieved by aggregating the rules associated with X7BRB. The ER (Evidential Reasoning) inference mechanism is applied to obtain this overall calculative value in terms of fuzzy values. There are two forms of ER, namely, recursive and analytical. The analytical ER is considered to reduce the computational complexity as shown in Eqs. (9) and (10).
The BRBES consists of a three-layer architecture, which comprises user interface, inference, ar knowledge-base layers as can be seen in Fig. 2. Since the system is web-based, various web programming tools such as PHP, CSS and HTML are considered to build the system interface. The inference layer consists of various inference procedures of BRBES as discussed in the previous section. This layer has been developed by using PHP, JavaScript, and JQuery. For simplicity and shorter development cycles, PHP has been considered. To make the client side behavior dynamic, JavaScript has been used, which maintains the link between the inference and interface layers. On the contrary, JQuery has been considered to maintain the link between the knowledge-base and inference layers. The BRBES’s knowledge-base is developed using MySQL be- cause of its flexibility. The initial BRB is also stored in MySQL. In addition, MySQL facilitates quick data access and provides necessary security.
The BRBES consists of a three-layer architecture, which comprises user interface, inference, ar knowledge-base layers as can be seen in Fig. 2. Since the system is web-based, various web programming tools such as PHP, CSS and HTML are considered to build the system interface. The inference layer consists of various inference procedures of BRBES as discussed in the previous section. This layer has been developed by using PHP, JavaScript, and JQuery. For simplicity and shorter development cycles, PHP has been considered. To make the client side behavior dynamic, JavaScript has been used, which maintains the link between the inference and interface layers. On the contrary, JQuery has been considered to maintain the link between the knowledge-base and inference layers. The BRBES’s knowledge-base is developed using MySQL be- cause of its flexibility. The initial BRB is also stored in MySQL. In addition, MySQL facilitates quick data access and provides necessary security.
Figure 3: BRB Interface.
Figure 3: BRB Interface.
Table 6: Earthquake prediction results generated by BRBES and FLBES along with original results an ANN based system was also developed. Its results are shown in column 12 of Table 6. The result of the same earthquake by using this system is found as 8.29, which is also far away from the original data.
Table 6: Earthquake prediction results generated by BRBES and FLBES along with original results an ANN based system was also developed. Its results are shown in column 12 of Table 6. The result of the same earthquake by using this system is found as 8.29, which is also far away from the original data.
Figure 4: Sanriku Earthquake, Japan, 1896. [Original Result: 8.5, BRBES Result: 8.494]
Figure 4: Sanriku Earthquake, Japan, 1896. [Original Result: 8.5, BRBES Result: 8.494]
possible if the input data can be acquired by deploying wireless sensor network technologies in a region [52][53][54].
possible if the input data can be acquired by deploying wireless sensor network technologies in a region [52][53][54].
Related topics:
Computer ScienceEarthquake PredictionBelief Rule Base
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved