Table 1 Summary of recent machine learning–based landslide susceptibility studies.

Azarafza et al. (2021)

Isfahan province

CNN-DNN deep learning with several benchmark classifiers

The CNN–DNN model for landslide susceptibility mapping was found to predict more accurately than the benchmark algorithms, with an AUC = 90.9%, IRs = 84.8%, MSE = 0.17, RMSE = 0.40, and MAPE = 0.42. The map provided by the CNN–DNN clearly revealed a high-susceptibility area in the west and southwest, related to the main Zagros trend in the province.

¹⁹

Nikoobakht et al. (2022)

Gorzineh-khil region, N of Iran

CNN with several benchmark classifiers

An evaluation of the results of the susceptibility assessment revealed that the CNN led the other classes in terms of 79.0% accuracy, 73.0% precision, 75.0% recall, and 77.0% f1-score, and, hence, provided better accuracy and the least computational error when compared to the other models (e.g., SVM, k-NN, DT, ).

¹⁸

Nanehkaran et al. (2022)

Alborz province

MLP, SVM, DT, RF

According to the susceptibility results, the north part of the studied region has high risks for rockfall failures. As machine learning-based performance analysis, the MLP-based model by 0.82 accuracy/0.85 precision reached the highest rank more than other classifiers such asSVM (accuracy = 0.78/precision = 0.78), DT (accuracy = 0.70/precision = 0.73), and RF (accuracy = 0.68/precision = 0.70). Also, overall accuracy obtained from ROC, the main model MLP (AUC = 0.811) is higher than SVM (AUC = 0.780), DT (AUC = 0.740), and RF (AUC = 0.500).

¹⁴

Ghasemian et al. (2022)

Kurdistan Province

robust deep-learning (DP) with several benchmark classifiers

Results based on the testing dataset revealed that the DP model had the highest accuracy (0.926) of the compared algorithms, followed by NBTree (0.917), REPTree (0.903), and SVM (0.894). The landslide susceptibility maps prepared from the DP model with AUC = 0.870 performed the best. We consider the DP model a suitable tool for landslide susceptibility mapping

²⁵

Hamedi et al. (2022)

Ardabil province

CNN-LSTM

The AUC values for CNN and LSTM models were 0.821 and 0.832, respectively. Therefore, it can be concluded that LSTM performance is slightly better than CNN; but in general, both models have close performance and the accuracy of both models is acceptable.

²⁶

Shen et al. (2023)

western Azerbaijan province

MLP, SVM,

Results indicate that machine learning algorithms are more effective than other methods for evaluating areas’ sensitivity to landslide hazards. The Simple SVM and Kernel Sigmoid algorithms performed well, with a performance score of one, indicating high accuracy in predicting landslide-prone areas.

²⁷

Shahabi et al. (2023)

Kurdistan Province

SVM, RF, DT,

Results indicate that DT with 0.94, RF with 0.82 and SVM with 0.75 accuracy is provide proper landslide susceptibility analysis and maps.

²⁸

Janizadeh et al. (2023)

National level

XGBoost, RF,

The RF model produced the best results (AUC = 0.95). The XGBoost model was not as robust (AUC = 0.93).

²⁹

Yousefi et al. (2024)

Kermanshah Province

ANFIS, SVM, LBO, EEFO

The results showed higher accuracy from the stacking ensemble technique with EEFO and LBO algorithms with AUC-ROC values of 94.81% and 94.84% and RMSE values of 0.3146 and 0.3142, respectively. The proposed approach can help managers and planners prepare accurate and reliable LSMs and, as a result, reduce the human and financial losses associated with landslide events.

³⁰

Hosseinzadeh et al. (2024)

Semnan and Kashmar Plains

SVM, RF, DT, LR, BLR, MLR

The results suggest that the BRT method is not significantly affected by data set size, but increasing the number of training set data points in MLR results in a decreased error rate.

³¹

Mao et al. (2024)

Lake Urmia Basin

SVM, RF, DT, LR, Fuzzy logic, TOPSIS

The outcomes of the landslide susceptibility assessment reveal that the main high susceptible zones for landslide occurrence are concentrated in the northwestern, northern, northeastern, and some southern and southeastern areas of the region. Moreover, when considering the implementation of predictions using different algorithms, it became evident that SVM exhibited superior performance regardingboth accuracy (0.89) and precision (0.89), followed by RF, with and accuracy of 0.83 and a precision of 0.83. However, it is noteworthy that TOPSIS yielded the lowest accuracy value among the algorithms assessed.

³²

Jahanbani et al. (2024)

Mazandaran province

AdaBoost, DT,

The ROC curve surpassing 0.96, accompanied by an accuracy of 0.93%, a sensitivity of 0.95%, and a specificity of 0.92%, this amalgamation substantiates its prowess.

³³

Rastkhadiv et al. (2025)

Sarvabad

FR, SVM, RF, WoE

As results, among the FR-RF, WoE-RF, FR-SVM, and WoE-SVM models, the FR-RF model accounted for the most high performance. In the end, it can be concluded that obtaining an accurate and reasonable spatial prediction map can help managers and urban planners identify zones susceptible to landslide occurrence so that they can manage the potential crises of landslide-prone zones.

³⁴

Feng et al. (2025)

Bakhtegan watershed

CNN with several benchmark classifiers

This study demonstrates that deep learning, particularly CNNs, offers a powerful and scalable solution for landslide susceptibility assessment. The findings provide valuable insights for urban planners, engineers, and policymakers to implement effective risk reduction strategies and enhance resilience in landslide-prone regions.

³⁵