{"id":185337,"date":"2025-08-29T19:14:16","date_gmt":"2025-08-29T19:14:16","guid":{"rendered":"https:\/\/www.europesays.com\/us\/185337\/"},"modified":"2025-08-29T19:14:16","modified_gmt":"2025-08-29T19:14:16","slug":"assessment-of-university-students-earthquake-coping-strategies-using-artificial-intelligence-methods","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/us\/185337\/","title":{"rendered":"Assessment of university students\u2019 earthquake coping strategies using artificial intelligence methods"},"content":{"rendered":"

In this section; information about Coping with Earthquake Stress Strategy Dataset, Confusion Matrix and Performance Metrics, Matthews Correlation Coefficient, Chi-Square Test, Correlation Matrix, Cross Validation, Artificial Intelligence Models, Logistic Regression, Bagging and Random Forest are given. Flow chart of the study is shown in Fig. 1<\/a>. All methods were performed in accordance with the relevant guidelines and regulations, and the study was approved by Ethics Committee for Social and Behavioral Sciences at Necmettin Erbakan University, in line with the Declaration of Helsinki.<\/p>\n

Coping with earthquake stress strategy dataset (CESS dataset)<\/p>\n

In the study, the Scale of Coping Strategies with Earthquake Stress developed by Yondem and Eren was used19<\/a>. Earthquake Stress Coping Strategies Scale was applied to 858 people. The data were collected electronically and made available for analysis. The created data set consists of 24 variables in total. These extracted features are transferred to the dataset. Using the cross-validation method, the trained data are directed to classification models. The trained algorithms provide output according to the determined classes.<\/p>\n

In the data set creation stages, demographic questions were first asked and then a scale consisting of 16 questions was used. Informed consent was obtained from all participants prior to their involvement in the study. Data were collected from 858 university students between 19.01.2024 and 07.03.2024. In total, a data set consisting of 24 variables was obtained. The 24 variables are shown in Table 1<\/a>. The dataset in Table\u00a01<\/a> includes both categorical variables (e.g., gender, residence type) and ordinal Likert-scale items. Categorical variables were processed using label encoding to convert them into numerical values. The Likert-scale items were treated as ordinal data to preserve their inherent order. No normalization or scaling was applied to these variables. Class labels were determined based on the average scores participants received from the 16-item CESS scale, using threshold values as follows: 1.00\u20132.50\u2009=\u2009Disagree; 2.51\u20133.00\u2009=\u2009Agree; 3.01\u20135.00\u2009=\u2009Strongly Agree. While setting these thresholds, the frequency distribution of the mean CESS scores was examined. It was observed that the frequencies clustered within these defined intervals. This distribution is illustrated in Fig.\u00a02<\/a>. Therefore, the data were categorized into three distinct classes for classification purposes. This classification approach reflects the ordinal nature of Likert-type data. Standard classifiers were selected over ordinal-specific ones as the clustered distributions enabled effective nominal separation, as evidenced by high MCC values. The results obtained from these variables consist of 3 classes. These are:<\/p>\n

Table 1 Coping with earthquake stress strategy dataset Variables.<\/b>Fig. 1<\/b>\"figure<\/a>Fig. 2<\/b>\"figure<\/a><\/p>\n

Threshold-based frequency distributions.<\/p>\n

Table 2 Sample confusion matrix notation and Explanations.<\/b><\/p>\n