Predicting Natural Hazards and Disaster Risks using Improved Particle Swarm Optimization based Support Vector Machine

Abstract

Predicting natural hazards before its occurrence is crucial to reduce the impact it causes economically to the cities/areas. Therefore, advanced Machine Learning (ML) models can be used to manage disasters by performing an early detection of hazards susceptible locations. The geographical data on natural hazards is collected pre-processed to eliminate the null values. The data is applied with the normalization technique and performed data scaling to transform all the variables to the same range. To obtain complex hierarchical features, Residual Network (ResNet) is used and the extracted features are selected using Attention Based Isolated Forest (ABIF) to solve the standard linear or quadratic optimization problem. To obtain the optimal solution in predicting the natural hazards the parameters of Support Vector Machine (SVM) are optimised using Improved Particle Swarm Optimization (IPSO). This ensures the model’s generalization and predictive ability and to handle high dimension data. The results showed that the proposed IPSO-SVM achieved better prediction accuracy with 98.59%, and precision of 97.74% when compared to the conventional ML methods such as Natural Gradient Boosting Machines (NGBoost) and AdaBag- Nearest Shrunken Centroids (AB-NSC).

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 4 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.