Dự đoán khả năng sạt lở đất ở Việt Nam bằng các thuật toán học máy

0 Người đánh giá. Xếp hạng trung bình 0

Tác giả: Trọng Huynh Phạm

Ngôn ngữ: Vie

Ký hiệu phân loại: 624 Civil engineering

Thông tin xuất bản: Khí tượng thủy văn 2023

Mô tả vật lý: 78-90

Bộ sưu tập: Metadata

ID: 396453

Thêm vào giỏ Liên kết toàn văn

Tóm tắt
Chủ đề

Việt Nam là quốc gia có địa hình đồi núi dốc và nằm trong vùng mưa nhiệt đới gió mùa, vì vậy hiện tượng sạt lở đất diễn ra khá phổ biến. Nghiên cứu này tập trung vào việc dự đoán khả năng sạt lở đất ở Việt Nam bằng các thuật toán hồi quy, Random Forest (RF), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbor regression (KNN), Linear Support Vector Regressor (SVR), và Linear Regression (LR). Các biến đặc trưng có liên quan đến sạt lở đất được sử dụng, bao gồm độ ẩm đất, địa chấn động đất, lượng mưa, độ cao và độ dốc. Các thuật toán được huấn luyện trên tập dữ liệu mẫu để đánh giá hiệu suất của chúng. Kết quả nghiên cứu cho thấy thuật toán Random Forest (RF) có thể dự đoán tốt khả năng sạt lở đất. Kết quả dự đoán từ tập huấn luyện và tập kiểm tra, với hệ số xác định R2 có giá trị cao nhất 0,85, thể hiện khả năng giải thích biến động dữ liệu tốt. Bên cạnh đó các giá trị (MSE) và (RMSE) thấp nhất lần lượt là 150,21 và 12,25. Các thuật toán khác cũng cho kết quả tương đối tốt, nhưng (RF) vượt trội hơn. Điều đó cho thấy cần kết hợp năm thuật toán này lại với nhau để xử lý một lượng lớn các dữ liệu có độ phức tạp cao, nhằm tạo ra một mô hình dự đoán sạt lở đất ở Việt Nam bằng các thuật toán Học máy có tính ổn định, chính xác., Tóm tắt tiếng anh, Vietnam is a country with hilly and sloping terrain, located in a region with tropical monsoon climate, making landslides quite common. This study focuses on predicting the likelihood of landslides in Vietnam using regression algorithms, namely Random Forest (RF), Extreme Gradient Boosting (XGBoost), K-Nearest Neighbor regression (KNN), Linear Support Vector Regressor (SVR), and Linear Regression (LR).Relevant landslide-related features are used, including soil moisture, seismic activity, rainfall, elevation, and slope. The algorithms are trained on a sample dataset to evaluate their performance. The research results show that the Random Forest (RF) algorithm can predict landslide susceptibility effectively. The prediction results from the training and testing datasets, with the highest coefficient of determination (R2) value of 0.85, demonstrate a good ability to explain data variations. Additionally, the lowest mean squared error (MSE) and root mean squared error (RMSE) values are 150.21 and 12.25, respectively.The other algorithms also yield relatively good results, but RF outperforms them. This indicates the need to combine these five algorithms to handle a large amount of complex data, to create a stable and accurate landslide prediction model in Vietnam using machine learning algorithms.

Tạo bộ sưu tập với mã QR