Interpretable deep learning framework for monthly rainfall prediction using attention and channel sparsity transformer model

Over the past 50 years (1973–2022), South Korea has experienced a modest increase in average precipitation. Notably, the maximum hourly rainfall has significantly risen (Kim et al., 2022). There has been an observed intensification in the intensity and frequency of extreme precipitation events, leading to substantial socioeconomic damages, flash floods, and urban flooding with severe consequences (Dave et al., 2021). Addressing locally occurring extreme precipitation in densely populated areas requires accurate urban-scale rainfall predictions. Given the anticipated increase in future precipitation variability on the Korean Peninsula, detailed spatiotemporal prediction technologies are crucial for effectively managing and responding to extreme precipitation events.We developed a rainfall prediction system through deep learning approach. In this study, we crafted a rainfall prediction system based on U-Net, a deep-learning architecture widely employed as a foundational model in previous rainfall prediction studies (Badrinath et al., 2023; Han et al., 2023; Lyu et al., 2023). The Advantage of U-Net lies in its end-to-end usability, minimizing the need for manual feature extraction, even with limited training data in the weather domain. To predict rainfall patterns over time, we adopted a recursive approach, drawing inspiration from prior research (Ayzel et al., 2020). In the case of the concentrated rainfall event in Osong in South Korea, July 2023, comparing QPE (Quantitative Precipitation Estimation) with the proposed rainfall prediction technology showed superior performance in terms of spatial patterns and rainfall intensity for the 10-minute lead time. Conversely, numerical weather prediction (Korea Local Analysis and Prediction System) failed to capture the rainfall pattern. For the 180-minute lead time, numerical prediction successfully detected rainfall, while the proposed prediction technology did not capture the rainfall pattern.Despite being in the early stages of development, case studies validate that our proposed system effectively simulates rainfall patterns that traditional nowcasting or numerical methods may not accurately replicate. However, limitations emerged in predicting localized rainfall intensity as the prediction time lengthened, revealing a tendency for spatial patterns of rainfall to be smoothed. As a follow-up study, our objective is to explore the applicability of deep learning across various aspects of the rainfall prediction process. This includes investigating super-resolution and blending of data produced by existing rainfall prediction methods, conducting empirical studies of deep learning models for domestic heavy rainfall cases, and optimizing existing numerical models. Our goal is to assess the feasibility of deep learning and enhance the accuracy of continuous prediction technology.

Accurate rainfall prediction is essential not only for water resource management but also for forecasting and mitigating the impacts of climate change-driven weather events such as floods and droughts. Due to the high spatiotemporal variability of complex meteorological phenomena like rainfall, effective prediction necessitates in high-quality data collection, model application, and uncertainty analysis. Unlike existing studies that focus primarily on developing deep learning models to improve rainfall prediction accuracy, this study evaluates the uncertainty of rainfall predictions using pre-existing deep learning models, U-Net and ConvLSTM, with artificially generated elliptical rainfall data. Artificial rainfall data were designed with four temporal patterns: constant, gradually increasing, gradually decreasing, and time-varying. These patterns were applied in horizontal, vertical, and diagonal movements to evaluate the models' ability to handle spatiotemporal complexity. The results indicate that both deep learning models exhibited spatial smoothing issues on rainfall predictions over time. However, the U-Net model demonstrated superior spatiotemporal performance compared to ConvLSTM. While this study focuses solely on deep learning models for rainfall prediction, future research will consider factors such as data complexity and loss functions to conduct a comprehensive evaluation of prediction uncertainty. This work is expected to contribute to the development of methodologies for rainfall modeling using deep learning approaches. Funding: This research was supported by Disaster-Safety Platform Technology Development Program of the National Research Foundation of Korea(NRF) funded by the Ministry of Science and ICT. (No. 2022M3D7A1090338)

With an increase in the intensity and frequency of extreme rainfall events, there is a pressing need for accurate rainfall nowcasting applications. In recent years, precipitable water vapor (PWV) data obtained from GNSS observations have been widely used in rainfall prediction. Unlike previous studies mainly focusing on rainfall occurrences, this study proposes a transformer-based model for hourly rainfall prediction, integrating the GNSS PWV and ERA5 meteorological data. The proposed model employs the ProbSparse self-attention to efficiently capture long-range dependencies in time series data, crucial for correlating historical PWV variations with rainfall events. Additionally, the adoption of the DILATE loss function better captures the structural and timing aspects of rainfall prediction. Furthermore, traditional rainfall prediction models are typically trained on datasets specific to one region, which limits their generalization ability due to regional meteorological differences and the scarcity of data in certain areas. Therefore, we adopt a pre-training and fine-tuning strategy using global datasets to mitigate data scarcity in newly deployed GNSS stations, enhancing model adaptability to local conditions. The evaluation results demonstrate satisfactory performance over other methods, with the fine-tuned model achieving an MSE = 3.954, DTW = 0.232, and TDI = 0.101. This approach shows great potential for real-time rainfall nowcasting in a local area, especially with limited data.

Prediction of rainfall is needed by every farmer to determine the planting period or for an institution, eg agriculture ministry in the form of plant calendars. BMKG is one of the national agency in Indonesia that doing research in the field of meteorology, climatology, and geophysics in Indonesia using several methods in predicting rainfall. However, the accuracy of predicted results from BMKG methods is still less than optimal, causing the accuracy of the planting calendar to only reach 50% for the entire territory of Indonesia. The reason is because of the dynamics of atmospheric patterns (such as sea-level temperatures and tropical cyclones) in Indonesia are uncertain and there are weaknesses in each method used by BMKG. Another popular method used for rainfall prediction is the Deep Learning (DL) and Extreme Learning Machine (ELM) included in the Neural Network (NN). ELM has a simpler structure, and non-linear approach capability and better convergence speed from Back Propagation (BP). Unfortunately, Deep Learning method is very complex, if not using the process of simplification, and can be said more complex than the BP. In this study, the prediction system was made using ELM-based Simplified Deep Learning to determine the exact regression equation model according to the number of layers in the hidden node. It is expected that the results of this study will be able to form optimal prediction model.Keywords: prediction, rainfall, ELM, simplified deep learning

RfGanNet: An efficient rainfall prediction method for India and its clustered regions using RfGan and deep convolutional neural networks

Background: Rainfall prediction is a critical task in meteorology and environmental science, with far-reaching implications for disaster preparedness, agriculture, and water resource management. Rainfall prediction can benefit greatly from the application of deep learning techniques like Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) networks, which have demonstrated great promise in time series forecasting. Objective: To use LSTM and GRU to forecast rainfall in the Kathmandu metropolitan area using information gathered from the Department of Hydrology and Meteorology, Babarmahal, Kathmandu, Nepal. Materials and Methods: Historical meteorological data was collected from Department of Hydrology and Meteorology, Babarmahal, Kathmandu, Nepal and preprocessed to create a suitable dataset. With this preprocessed dataset containing variables such as temperature, humidity, atmospheric pressure, wind speed, and direction, two deep learning methods, LSTM and GRU, were trained. To assess the performance, various evaluation metrics, including Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and R2 were used. Results: The daily rainfall has been predicted using LSTM and GRU using 0.0001 learning rate, 50 epochs and 8 batch size. RMSE, MAE and R2 values of LSTM are 2.51, 1.79 and 0.81 respectively. Similarly, RMSE, MAE and R2 values of GRU are 2.31, 1.51 and 0.95 respectively. Conclusion: Test results show that the GRU model's predictions are generally near to the actual recorded rainfall amounts, as evidenced by the fact that the model's test RMSE and MAE are fewer than those of the LSTM. A higher R2 value of GRU suggests a better fit in the rainfall data, as more of the variance in the outcome is explained by the predictors.

Recent sensing advances have enabled the use of knee acoustic emissions (KAEs)—sounds generated during flexion and extension—as non-invasive biomarkers for detecting knee osteoarthritis (OA). Most existing KAE-based OA classifiers use hand-crafted features with conventional machine learning, which can work on small datasets but often generalize poorly across devices and recording conditions and provide limited insight into the acoustic patterns driving predictions. We propose a robust and interpretable deep transfer learning framework that classifies OA directly from raw KAE signals. The method learns discriminative time-frequency representations, leverages transfer learning to mitigate data sparsity, and incorporates explainable artificial intelligence (XAI) to confirm that decisions rely on physiologically plausible acoustic components. Using a clinically representative KAE dataset that, for the first time, includes a substantial group of healthy participants with high body mass index, we systematically compare the proposed approach with several benchmark algorithms. Our method consistently attains an average accuracy of about 89% for distinguishing OA from healthy knees across multiple initializations and dataset splits, indicating strong performance and stability. XAI visualizations further highlight the key time-frequency regions that influence the model’s predictions. This work demonstrates the promise of deep transfer learning for accurate, interpretable, and scalable OA assessment and home monitoring.

Rainfall plays a prominent role in managing of water resources. The accurate prediction of rainfall is the greatest challenge in the field of hydrologic studies. The prediction of rainfall is necessary to overcome natural disasters like flood and drought. The inaccurate prediction of rainfall causes either dry or overflow in water storage structures. In this study different types of Machine Learning (ML) and deep learning techniques are adopted to predict rainfall pattern of Aiyar river basin, in Tiruchirappalli district. The comparative study of these ML models is done to identify the best ML model for the study area. The comparison was done for different scenarios and time intervals. The rainfall data from years 1987 to 2023 is used for predicting the daily rainfall in the basin. The rainfall data from years 1987 to 2007 is used for testing and the remaining years data is used for training the data set. The Theisen polygon method is used to average and weighted the rainfall data in the basin. The ML models and deep learning techniques used in this study are Linear model, Support Vector Machine (SVM) and Long Short-Term Memory (LSTM) models. The rainfall was predicted for different time scenario by using different ML algorithms like Autocorrelation method. The accuracy of the predicted model results was tested with RMSE, MASE and R square values. The result shows coefficient between 0.5 to 0.9 within the limit from the daily rainfall values. From the overall model comparison, it is observed that the SVM model accuracy is high compared to the other models involved in this study. It is concluded that two different methods ML and deep learning methods have been applied with same data in which SVM ML techniques gives better results in this study area. In future the predicted rainfall data of this study can be used for accurate flood forecasting and modelling of Aiyar basin.

Rainfall prediction is the active area of research as it enables the farmers to move with the effective decision-making regarding agriculture in both cultivation and irrigation. The existing prediction models are scary as the prediction of rainfall depended on three major factors including the humidity, rainfall and rainfall recorded in the previous years, which resulted in huge time consumption and leveraged huge computational efforts associated with the analysis. Thus, this paper introduces the rainfall prediction model based on the deep learning network, convolutional long short-term memory (convLSTM) system, which promises a prediction based on the spatial-temporal patterns. The weights of the convLSTM are tuned optimally using the proposed Salp-stochastic gradient descent algorithm (S-SGD), which is the integration of Salp swarm algorithm (SSA) in the stochastic gradient descent (SGD) algorithm in order to facilitate the global optimal tuning of the weights and to assure a better prediction accuracy. On the other hand, the proposed deep learning framework is built in the MapReduce framework that enables the effective handling of the big data. The analysis using the rainfall prediction database reveals that the proposed model acquired the minimal mean square error (MSE) and percentage root mean square difference (PRD) of 0.001 and 0.0021.

Lung adenocarcinoma remains a leading cause of cancer-related mortality, and the diagnostic performance of computed tomography (CT) is limited when dependent solely on human interpretation. This study aimed to develop and evaluate an interpretable deep learning framework using an attention-enhanced Squeeze-and-Excitation Residual Network (SE-ResNet) to improve automated classification of lung adenocarcinoma from thoracic CT images. Furthermore, Gradient-weighted Class Activation Mapping (Grad-CAM) was applied to enhance model interpretability and assist in the visual localization of tumor regions. A total of 3800 chest CT axial slices were collected from 380 subjects (190 patients with lung adenocarcinoma and 190 controls, with 10 slices extracted from each case). This dataset was used to train and evaluate the baseline ResNet50 model as well as the proposed SE-ResNet50 model. Performance was compared using accuracy, Area Under the Curve (AUC), precision, recall, and F1-score. Grad-CAM visualizations were generated to assess the alignment between the model's attention and radiologically confirmed tumor locations. The SE-ResNet model achieved a classification accuracy of 94% and an AUC of 0.941, significantly outperforming the baseline ResNet50, which had an 85% accuracy and an AUC of 0.854. Grad-CAM heatmaps produced from the SE-ResNet demonstrated superior localization of tumor-relevant regions, confirming the enhanced focus provided by the attention mechanism. The proposed SE-ResNet framework delivers high accuracy and interpretability in classifying lung adenocarcinoma from CT images. It shows considerable potential as a decision-support tool to assist radiologists in diagnosis and may serve as a valuable clinical tool with further validation.

Weather is an external factor that has major influences on planning, work and maintenance in the field of civil engineering. One of the factors that affects the weather is rainfall which often causes flooding in Indonesia especially in coastal cities. Many civil works are greatly hampered during the rainy season. To improve rainfall prediction, various methods have been developed using Computer Vision, Deep Learning, and Big Data. By using this technologies, groundbased cloud recognition is carried out to classify cloud types according to the classification of World Meteorological Organization which divides clouds into 10 types of clouds. Cloud classification can be identify by cloud characteristics like cloud thickness, cloud area and cloud color that the possibility of rain can be estimated. Through cloud observations using ground-based cameras and automatic weather measuring data as verification of rain data and using data processing using Deep Learning and Big Data, a rainfall prediction model can be made with a higher level of accuracy based on Computer Vision and Deep Learning. This model will be very helpful in predicting rainfall, especially in coastal cities in Java which often experience flooding. This research is part of a proposal from a dissertation. The stages in the research are still preparing an experimental method and preparing a weather reading tool in the form of a ground-based camera and an automatic weather meters.

Since the previous decade, weather has changed rapidly, raising concerns about unpredictable rains. Humidity, pressure, wind, and temperature affect rainfall. Thus, studying such characteristics to predict rainfall is intriguing. Machine learning and deep learning have simplified rainfall prediction, which is still under development. This study implements machine learning and deep learning models trained on 145460 rows and 25 characteristics. These models are tested on a test dataset using measures like accuracy and PRF score to determine the best model for rainfall prediction. Tuning each model's hyper-parameters improves results. A comparative study examined how different factors affect rainfall. This study covers preprocessing, feature engineering, model selection, and implementation. This work aims to offer an acceptable and accurate model for predicting rainfall in a web-app. This study compares machine learning and deep learning methods and provides easy access to them.

Rainfall prediction is critical in agriculture, water resource management, transportation, and disaster planning. As a critical component of the hydrological cycle, precisely forecasting daily rainfall helps to lessen the effects of floods, droughts, and other extreme weather events. However, due to the nonlinear and complicated interaction of meteorological factors, daily rainfall forecast remains a grim problem. The proposed work adopts a systematic deep learning-based framework, integrating Long Short-Term Memory (LSTM) and Bidirectional LSTM (BiLSTM) models and Transformers to predict the rainfall of the next day. Besides deep models, a few machine learning classifiers, such as the Logistic Regression (LR), Support Vector Machines (SVM), and the K-Nearest Neighbor (KNN) will be applied to benchmark the performance.One of the main insights in this study is the systematic comparison of a wide variety of activation functions, such as Sigmoid, ReLU, Tanh, Swish, Leaky ReLU, and ELU to measure their effects on model accuracy, convergence, and generalization performance. An experiment was employed on real-life rainfall of the USA dataset, which proved the efficiency of using the BiLSTM model and ReLU and Leaky ReLU activation functions to achieve up to 99% accuracy, whereas Transformer model was also efficient as it obtain up to 98%, but it was not stable and showed a tendency towards occasional instability during the training process. On the contrary, traditional machine learning models displayed an intermediate predictive accuracy, with average accuracies of approximately 87%. Significantly, the results were validated through multi-seed experiments with confidence intervals and baseline benchmarks, demonstrating that the near-perfect scores were not incidental but consistently stable across different initializations. This study provides a new theoretical standpoint in terms of activation functions in deep sequence models as it studies beneficiaries and shortcomings through a mathematical point of view. The results highlight that activation function selection can have an important impact on the accuracy of forecasts and can be significant in building more accurate, scalable, and general learning subroutines in weather predictions.

Data-driven multi-step prediction and analysis of monthly rainfall using explainable deep learning

Natural disasters and floods brought on by heavy rainfall pose serious threats to human health and lives every year on a global scale. The intricacy of meteorological data makes it difficult to provide accurate rainfall predictions, despite their critical importance in nations like India where agriculture is the primary occupation. Rainfall forecasting has recently benefited from Artificial Intelligence (AI) developments such as Deep Learning (DL) and Machine Learning (ML) techniques. This article provides a comprehensive survey of recent studies that use AI techniques for rainfall prediction, analyzing them based on the ML algorithms and DL methods used, organized by publication year. The findings show that DL approaches are more effective than traditional ML methods and shallow neural network models. This research is important as it has significant impacts on agriculture, disaster preparedness, and water resource management. Finally, it outlines future research directions for further advancements in rainfall prediction through AI methodologies.