Mapping Evaluation Research Articles

Enhanced forecasting of emergency department patient arrivals using feature engineering approach and machine learning

BackgroundEmergency department (ED) overcrowding is an important problem in many countries. Accurate predictions of ED patient arrivals can help management to better allocate staff and medical resources. In this study, we investigate the use of calendar and meteorological predictors, as well as feature-engineered variables, to predict daily patient arrivals using datasets from eleven different EDs across three countries.MethodsSix machine learning (ML) algorithms were tested on forecasting horizons of 7 and 45 days. Three of them – Light Gradient Boosting Machine (LightGBM), Support Vector Machine with Radial Basis Function (SVM-RBF), and Neural Network Autoregression (NNAR) – were never before reported for predicting ED patient arrivals. Algorithms’ hyperparameters were tuned through a grid-search with cross-validation. Prediction performance was assessed using fivefold cross-validation and four performance metrics.ResultsThe eXtreme Gradient Boosting (XGBoost) was the best-performing model on both prediction horizons, also outperforming results reported in past studies on ED arrival prediction. XGBoost and NNAR achieved the best performance in nine out of the eleven analyzed datasets, with MAPE values ranging from 5.03% to 14.1%. Feature engineering (FE) improved the performance of the ML algorithms.ConclusionAccuracy in predicting ED arrivals, achieved through the FE approach, is key for managing human and material resources, as well as reducing patient waiting times and lengths of stay.

Application of autoregressive integrated moving average and vector autoregression prediction models for stock and price stabilization at perum bulog

This study implements the ARIMA (Autoregressive Integrated Moving Average) method to predict rice stock levels at Perum Bulog in Lhokseumawe City. The objective of this research is to forecast rice stock levels for the upcoming year, 2024, and provide Perum Bulog Lhokseumawe with a more accurate reference for future stock planning. The benefit of this research lies in its potential to enable preventive measures to minimize rice stock shortages and support Perum Bulog in formulating policies to reduce the risk of future shortages. Error testing for rice prices with the ARIMA model shows MAPE values of ARIMA (1,0,0) = 21,256,076,432,276.10%, ARIMA (0,0,2) = 3.54%, and ARIMA (1,0,2) = 21,305,440,935,436.50%. Rice stock testing with MAPE error values yielded ARIMA (2,0,0) = 13.49%, ARIMA (0,0,1) = 13.65%, and ARIMA (2,0,1) = 10.27%. The error measurement results for prediction using MAPE are 1.5% for rice prices and 58.9% for rice stock. Based on these error measurements, the VAR model appears sufficiently accurate for predicting rice prices, as the MAPE value is at a low level.

The MAPE Analysis of Arima (p,d,q) on LQ45 Stock Price to Determine Training Data Period

Most of the research using the Arima (p,d,q) focused on the accuracy of prediction results. Unlike other research, this work examines the training data period suitable for modeling ARIMA (p,d,q) in stock prices. Due to the volatile movement of stocks, the number of training data is assumed to affect the LQ45 prediction results. This research used five kinds of training data, including daily data for up to 5 years. With these five types of data series, the Arima (p,d,q) was made for LQ45 stocks. The prediction was conducted for two months after obtaining the model 5 data series of LQ45 stocks. Two months of data were used for January and February 2021 prediction test data. The results of this prediction were compared with the test data to produce the MAPE value. Based on the observations and calculation results, the most suitable stock to use the Arima (p,d,q) was ASII. In 5 years, the stocks produced the lowest MAPE value of 0.05%. Relatively stable LQ45 stocks with no change in the Arima (p,d,q) using four consecutive data series were ACES, CTRA, INTP, MIKA, and TLKM. Based on the MAPE value analysis performed in this study, we concluded that the best period to use the Arima (p,d,q) for LQ45 stocks is two years, with a median error rate of only 6.0091%.

Data Decomposition Modeling Based on Improved Dung Beetle Optimization Algorithm for Wind Power Prediction

Accurate wind power forecasting is essential for maintaining the stability of a power system and enhancing scheduling efficiency in the power sector. To enhance prediction accuracy, this paper presents a hybrid wind power prediction model that integrates the improved complementary ensemble empirical mode decomposition (ICEEMDAN), the RIME optimization algorithm (RIME), sample entropy (SE), the improved dung beetle optimization (IDBO) algorithm, the bidirectional long short-term memory (BiLSTM) network, and multi-head attention (MHA). In this model, RIME is utilized to improve the parameters of ICEEMDAN, reducing data decomposition complexity and effectively capturing the original data information. The IDBO algorithm is then utilized to improve the hyperparameters of the MHA-BiLSTM model. The proposed RIME-ICEEMDAN-IDBO-MHA-BiLSTM model is contrasted with ten others in ablation experiments to validate its performance. The experimental findings prove that the proposed model achieves MAPE values of 5.2%, 6.3%, 8.3%, and 5.8% across four datasets, confirming its superior predictive performance and higher accuracy.

Development of multistage crop yield estimation model using machine learning and deep learning techniques.

In this research paper, machine learning techniques were applied to a multivariate meteorological time series data for estimating the wheat yield of five districts of Punjab. Wheat yield data and weather parameters over 34years were collected from the study area and the model was developed using stepwise multi-linear regression (SMLR), artificial neural network (ANN), support vector regression (SVR), random forest (RF) and deep neural network (DNN) techniques. Wheat yield estimation was done at the tillering, flowering, and grain-filling stage of the crop by considering weather variables from 46 to 4th, 46 to 8th, and 46 to 11th standard meteorological week. Weighted and unweighted Meteorological variables and yield data were used to train, test, and validate the models in R software. The evaluation results showed a consistent and promising performance of RF, SVR, and DNN models for all five districts with an overall MAPE and nRMSE value of less than 6% during validation at all three growth stages. These models exhibited outstanding performance during validation for the Faridkot, Ferozpur, and Gurdaspur districts. Based on accuracy parameters MAPE, RMSE, nRMSE, and percentage deviation, the RF model was found better followed by SVR and DNN models and, hence can be used for district-level wheat crop yield estimation at different crop growth stages.

Diagnosis of liver hemangioma using magnetic resonance diffusion-derived vessel density (DDVD) pixelwise map: a preliminary descriptive study.

Liver hemangiomas (HGs) are characterized by cavernous venous spaces delineated by a lining of vascular endothelial cells and interspersed with connective tissue septa. Typically, a liver HG has higher apparent diffusion coefficient (ADC) and T2 values than those of hepatocellular carcinomas (HCCs) and liver metastases, and lower ADC and T2 values than those of liver simple cysts. However, a portion of HGs shows ADC and T2 overlapping with those of HCC, liver metastasis, and simple cyst. When MRI is the first line examination for the liver, contrast enhanced imaging is commonly used to confirm the diagnosis of liver HG. Magnetic resonance diffusion-derived vessel density (DDVD) is a physiological surrogate of the area of microvessels per unit tissue area. DDVD is calculated according to: DDVD(b0b2) = Sb0/ROIarea0 - Sb2/ROIarea2, where Sb0 and Sb2 refer to the tissue signal when b is 0 or 2 (s/mm2). Sb2 and ROIarea2 can also be approximated by other low b-values (such as b=10) diffusion-weighted imaging (DWI). In this study, we conducted a preliminary evaluation of magnetic resonance DDVD pixelwise map (DDVDm) for liver HG diagnosis. Three testing datasets were included. All imaging data were acquired at 3.0T. Dataset-1 consisted of 16 HGs (lesion diameter: 1.5-8.85 cm), 4 focal nodular hyperplasia (FNHs, lesion diameter: 1.72-5.7 cm), and 24 HCCs (lesion diameter: 1.83-12.77 cm), and DDVDm was reconstructed with b=0 and b=2 images. Dataset-2 consisted of 6 HGs (lesion diameter: 1.14-6.2 cm), and DDVDm was reconstructed with b=0 and b=10 images. Dataset-3 consisted of 28 HCCs (lesion diameter: 1.91-3.52 cm), and DDVDm was reconstructed with b=0 and b=2 images. For dataset-1 and dataset-2, a trained reader was required to make a diagnosis for a lesion solely based on DDVDm with 4 choices: (I) HG with confidence; (II) HG without confidence; (III) solid mass-forming lesion (MFL) with confidence; (IV) solid MFL without confidence. Then, three readers attempted to confirm whether DDVDm features summarized from dataset-1 and dataset-2 would be generalizable to dataset-3. For dataset-1 and dataset-2 together, the correct diagnosis was made by the trained reader in 90.9% (20/22) of the HGs (77.7% with confidence) and 96.4% (27/28) of the MFLs (85.7% with confidence). HG generally showed substantially higher DDVD signal relative to background liver parenchyma. Though not necessarily, HG DDVD signals could be similar to those of blood vessels. Some HGs showed DDVD signals higher or similar to that of kidneys which have a higher perfusion than the liver. MFL generally showed DDVD signals only slightly higher, similar to, or even slightly lower, than that of background liver parenchyma. The DDVDm features of dataset-3 were all consistent with MFL. When DDVDm is used to evaluate the liver, HG can be diagnosed with confidence in a substantial portion of patients without the need for a contrast enhanced scan. Our result will be relevant for HG confirmation when MRI is the first line examination for the liver.

Comparative Time Series Analysis of SARIMA, LSTM, and GRU Models for Global SF6 Emission Management System

To maintain the balance of the atmosphere, the amount of change in greenhouse gas emissions must be under control. In order to create a management system and take forward-looking steps in this regard, there is no concrete data other than prediction models today. The success of prediction methods is better understood by comparing multiple methods. This research estimates the changes in the emissions of Sulfur Hexafluoride gas (SF6) in the atmosphere using Seasonal Autoregressive Integrated Moving Average Model (SARIMA), Long-Short Term Memory Neural Network (LSTM) and Gated Recurrent Unit (GRU) forecast models and compares their accuracies. Focusing on monthly SF6 emission values between 1998; 2023, time series analysis was performed to predict future emission figures. The actual values and forecast results were compared and evaluated with performance criteria such as R2, RMSE, NSE, MAE and MAPE%. The findings of this research highlight a continious upward trend in SF6 emissions and project that emission levels could approximately double from current levels by 2050. During the analysis process, all three methods performed well in estimating global SF6 gas emissions. The LSTM model generally outperformed SARIMA and GRU models, having the lowest MAPE (0.003%), MAE (0.0003), RMSE (0.0003), and R2 (1) values. It also exhibited very high predictive success with an NSE value of 0.9991. Therefore, it was determined to be the most suitable estimation method with the least error. The aim of this study is to contribute scientifically to the reduction strategies of SF6 emissions.

Short-term Traffic Flow Prediction：A Method of MEA-LSTM Model Based on Chaotic Characteristics Analysis

To effectively improve the accuracy of short-term traffic flow prediction, an improved Long Short-Term Memory (LSTM) method is proposed using the Mind Evolution Algorithm (MEA). Firstly, to address the issues of abnormal and missing traffic flow data, a Neighborhood Stacked Denoising AutoEncoder (NSDAE) is used for data repair. Then, the maximum Lyapunov exponent is used to determine the chaotic characteristics. Meanwhile, based on Bayesian estimation theory, the features of three-parameter sequences are fused in high-dimensional space using phase space reconstruction technique to obtain reconstructed multi-parameter fused traffic flow data. Finally, by taking advantage of the fact that MEA can divide the data into several subpopulations for optimal search separately, a prediction model based on MEA to improve LSTM is proposed. The results show that compared to the other two traditional data restoration methods, the NSDAE has higher accuracy, with the lowest average values of RMSE, MAE, and MAPE. Through the phase space reconstruction technique, the feature fusion of three parameters of traffic flow is realized in high-dimensional space, which makes up for the insufficiency of a single time-series data that cannot comprehensively levy the characteristics of traffic flow. The MEA-LSTM model outperforms the LSTM model in terms of prediction accuracy, computational efficiency, and generalization ability, and its RMSE, MEA, and MAPE are reduced by 24.3%, 28.9%, and 30.1%, respectively.

EKLT: Kolmogorov-Arnold attention-driven LSTM with Transformer model for river water level prediction

Water level prediction is crucial for water resource management and flood warnings. Hybrid LSTM-based methods are widely used but face the following challenges: (1) The attention mechanism based on the universal approximation theorem (UAT) is the mainstream method for improving performance, but its essence is infinite approximation of functions, and the accuracy is difficult to improve; (2) LSTM struggles with modeling complex, long-term dependencies. To address these problems, we apply empirical mode decomposition (EMD) and propose input-spatial attention based on the Kolmogorov-Arnold theorem (KAT) for precise water level feature representation. Cascaded LSTM and Transformer structures capture long-term dependencies. Finally, temporal attention is embedded to achieve accurate water level prediction. Experiments show the proposed method achieves RMSE, MAE, MAPE and R2 values of 0.1870, 0.1328, 1.1228 and 0.9540 in the Liaohe River of Liaoning Province, China, and 0.3027, 0.1844, 4.6034 and 0.8659 in the Hunhe River, respectively, indicating its strong predictive performance.

Time-efficient, high-resolution 3T whole-brain relaxometry using Cartesian 3D MR Spin TomogrAphy in Time-Domain (MR-STAT) with cerebrospinal fluid suppression.

Current three-dimensional (3D) MR Spin TomogrAphy in Time-Domain (MR-STAT) protocols use transient-state, gradient-spoiled gradient-echo sequences that are prone to cerebrospinal fluid (CSF) pulsation artifacts when applied to the brain. This study aims to develop a 3D MR-STAT protocol for whole-brain relaxometry that overcomes the challenges posed by CSF-induced ghosting artifacts. We optimized the flip-angle train within the Cartesian 3D MR-STAT framework to achieve two objectives: (1) minimization of the noise level in the reconstructed quantitative maps, and (2) reduction of the CSF-to-white-matter signal ratio to suppress CSF-associated pulsation artifacts. The optimized new sequence was tested on a gel/water phantom for accuracy evaluation of the quantitative maps, and on healthy volunteers to explore the effectiveness of the CSF artifact suppression and robustness of the new protocol. An optimized sequence with high parameter-encoding capability and low CSF signal response was proposed and validated in the gel/water phantom experiment. From in vivo experiments with 5 volunteers, the proposed CSF-suppressed sequence produced quantitative maps with no CSF artifacts and showed overall greatly improved image quality compared with the baseline sequence. Statistical analysis indicated low intersubject and interscan variability for quantitative parameters in gray matter and white matter (1.6%-2.4% for T1 and 2.0%-4.6% for T2), demonstrating the robustness of the new sequence. We present a new 3D MR-STAT sequence with CSF suppression that effectively eliminates CSF pulsation artifacts. The new sequence ensures consistently high-quality, 1-mm3 whole-brain relaxometry within a rapid 5.5-min scan time.

Application of state of health estimation and remaining useful life prediction for lithium-ion batteries based on AT-CNN-BiLSTM

Ensuring the long-term safe usage of lithium-ion batteries hinges on accurately estimating the State of Health \\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$(\ extrm{SOH})$$\\end{document} and predicting the Remaining Useful Life (RUL). This study proposes a novel prediction method based on a AT-CNN-BiLSTM architecture. Initially, key parameters such as voltage, current, temperature, and SOH are extracted and averaged for each cycle to ensure the uniformity and reliability of the input data. The CNN is utilized to extract deep features from the data, followed by BiLSTM to analyze the temporal dependencies in the data sequences. Since multidimensional parameter data are used to predict the SOH trend of lithium-ion batteries, an attention mechanism is employed to enhance the weight of highly relevant vectors, improving the model’s analytical capabilities. Experimental results demonstrate that the CNN-BiLSTM-Attention model achieves an absolute error of 0 in RUL prediction, an \\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$R^{2}$$\\end{document} value greater than 0.9910 , and a MAPE value less than 0.9003 . Comparative analysis with hybrid neural network algorithms such as LSTM, BiLSTM, and CNN-LSTM confirms the proposed model’s high accuracy and stability in SOH estimation and RUL prediction.

Short-term wind power prediction based on IBOA-AdaBoost-RVM

This study introduces an innovative model, namely IBOA-AdaBoost-RVM, which leverages the Improved Butterfly Optimization Algorithm (IBOA), Adaptive Boosting (AdaBoost), and Relevance Vector Machine (RVM). This model is used to solve the problem of low precision of wind power prediction. Initially, normalization is applied to reduce the influence of varying data dimensions. Subsequently, input variables are determined through the Pearson correlation method. Lastly, the efficacy of the introduced model is assessed across four distinct seasonal monthly data sets. The observed outcomes indicate that the proposed model outperforms other models in terms of evaluation metrics, with the average R2, RMSE, MAE, and MAPE values across the four datasets being 0.954, 10.403, 7.032, and 0.645, respectively, show that the proposed method has potential in the field of wind power prediction.

Analisis Peramalan Persediaan Sparepart Menggunakan Metode Exponential Smoothing di PT. XYZ

XYZ is a manufacturing company that produces Yamaha brand motorbike parts such as electrical switches, electrical sockets, lever assembly and horns. Many of the products sent were returned by customers because they did not meet quality standards, for example in March 2024 there were 1,420 pcs out of a total of 19,900 pcs that were returned. This means that companies have to increase production time and costs to replace defective products that are returned because they do not have safety stock. Forecasting is needed to control the production system so that it does not experience over stock and safety stock shortages. This research aims to provide recommendations for forecasting methods that companies can use to minimize the occurrence of production excesses and shortages in the company. Forecasting is done based on historical company sales data for 12 months. The method used is the exponential smoothing method which is then tested whether this method can be used in the future. This calculation uses the help of POM-QM (Production Operation Management – Quantitative Method) software. Calculations are carried out by testing the MAD, MSE, and MAPE values to obtain calculation error values. The results obtained by forecasting the main sw srtg lock assembly product using the exponential smoothing method were 15,708 for demand for the next period, MAD 3.38, MSE 22.84 and MAPE value 26%. Based on these results, the exponential smoothing method can be a recommendation for companies to forecast future demand. This is because the value of forecasting accuracy or MAPE is reasonable. The smallest percentage of MAPE values has a fairly minimal possibility of error in forecasting results.

Research on Forecasting Sales of Pure Electric Vehicles in China Based on the Seasonal Autoregressive Integrated Moving Average–Gray Relational Analysis–Support Vector Regression Model

Aiming to address the complexity and challenges of predicting pure electric vehicle (EV) sales, this paper integrates a time series model, support vector machine and combined model to forecast EV sales in China. Firstly, a seasonal autoregressive integrated moving average (SARIMA) model was constructed using historical EV sales data, and the model was trained on sales statistics to obtain forecasting results. Secondly, variables that were highly correlated with sales were analyzed using gray relational analysis (GRA) and utilized as input parameters for the support vector regression (SVR) model, which was constructed to optimize sales predictions for EVs. Finally, a combined model incorporating different algorithms was verified against market sales data to explore the optimal sales prediction approach. The results indicate that the SARIMA-GRA-SVR model with the squared prediction error and inverse method achieved the best predictive performance, with MAPE, MAE and RMSE values of 12%, 1.45 and 2.08, respectively. This empirical study validates the effectiveness and superiority of the SARIMA-GRA-SVR model in forecasting EV sales.

Optimizing deep neural network architectures for renewable energy forecasting

An accurate renewable energy output forecast is essential for energy efficiency and power system stability. Long Short-Term Memory(LSTM), Bidirectional LSTM(BiLSTM), Gated Recurrent Unit(GRU), and Convolutional Neural Network-LSTM(CNN-LSTM) Deep Neural Network (DNN) topologies are tested for solar and wind power production forecasting in this study. ARIMA was compared to the models. This study offers a unique architecture for Deep Neural Networks (DNNs) that are specifically tailored for renewable energy forecasting, optimizing accuracy by advanced hyperparameter tuning and the incorporation of essential meteorological and temporal variables. The optimized LSTM model outperformed others, with MAE (0.08765), MSE (0.00876), RMSE (0.09363), MAPE (3.8765), and R2 (0.99234) values. The GRU, CNN-LSTM, and BiLSTM models predicted well. Meteorological and time-based factors enhanced model accuracy. The addition of sun and wind data improved its prediction. The results show that advanced deep neural network (DNN) models can predict renewable energy, highlighting the importance of carefully selecting characteristics and fine-tuning the model. This work improves renewable energy estimates to promote a more reliable and environmentally sustainable electricity system.

Evaluation of Single-Event Effect Current-Carrier Mapping Based on Experimental Data.

For single-event radiation damage of power MOSFET devices, this paper aims to establish a statistical analysis method based on external observation (gate/drain current characteristics in irradiation environment) to recognize and evaluate the radiation evolution process and damage mechanism of microscopic physical quantities inside the devices, namely current-carrier (CC) mapping. Firstly, a special data fluctuate-collapse transform analysis method is proposed according to the temporal characteristics of the gate/drain current. Secondly, a carrier dynamic balance ratio based on current data is defined to evaluate the radiation damage degree of the device. TCAD is used to deeply study the relationship between the external current characteristics and the evolution process of internal physical quantities and the damage mechanism. The results show that the current data timing analysis based on fluctuate-collapse transformation can better peer into the evolution process of irradiation events inside the device, and the statistical analysis based on the dynamic balance ratio of carriers can evaluate the severity of irradiation damage to a certain extent.

A new quantitative method for evaluating the impact of garden greening on outdoor thermal environment in summer - A case study of Japanese residential gardens

Garden greening can improve the outdoor thermal environment of buildings, but there is a lack of quantitative research on this aspect. First, this research quantitatively analyzes the effect of garden greening on the outdoor thermal environment by conducting field measurements at various distances from measurement points set up in a traditional Japanese residential garden during the summer, and a TDTE model was established. The results show that the outdoor thermal environment evaluation indices at each measurement point reach the maximum value of 13:00 daily. The outdoor thermal environment evaluation indices such as the average air temperature value at the nearest measurement point to the garden greening were 12 % lower than those at the farthest measurement point. Secondly, the TDTE model was established based on this foundation, and the validation of the measured data showed that the TDTE model had high reliability (the average RMSE value was 1.791, and the average MAPE value was 2.978). Finally, the model can provide ideas as well as the theoretical basis for quantitative research related to greening and outdoor thermal environments. This research provides scientific guidance for planning as well as reconstruction of residential garden greening.

Efficient High Heating Value estimation using Latin Hypercube Sampling and Artificial Neural Network-based approach.

To maximize energy recovery in waste-to-energy (WTE) systems, the High Heating Value (HHV) of municipal solid waste (MSW) must be accurately estimated. To forecast the HHV of MSW, this study proposes a unique method that combines an Artificial Neural Network (ANN) model with Latin Hypercube Sampling (LHS), with a focus on Solan City, Himachal Pradesh, India. In the present study, the elemental characteristics of waste have been used to deal with uncertainty and to find the suitable parameters responsible for the HHV of the MSW. Initially, Latin Hypercube Sampling (LHS) has been used to deal with uncertainty in the elemental composition of MSW, which includes carbon (C), hydrogen (H), nitrogen (N), sulfur (S), and oxygen (O) content. This elemental composition has been used as input parameters to the ANN model for predicting the HHV of MSW. The network 5-28-5-1 offered a minimum MAPE value of 2.18%, MSE, RMSE and R2 values are 0.012, 0.107 and 0.767, respectively. Thereafter, a synaptic weight approach was used to find the most significant parameters responsible for HHV in MSW. It was observed that carbon is the most suitable parameter for HHV of MSW. By dealing with the uncertainty in MSW characteristics, the integration of LHS strengthens the robustness of the model. The results offer an accurate and economical approach for HHV estimation, which will be useful for improving the MSW management and WTE conversion procedures.

Transformative Transparent Hybrid Deep Learning Framework for Accurate Cataract Detection

This paper presents a transformative explainable convolutional neural network (CNN) framework for cataract detection, utilizing a hybrid deep learning model combining Siamese networks with VGG16. By leveraging a learning rate scheduler and Grad-CAM (Gradient-weighted Class Activation Mapping) for explainability, the proposed model not only achieves high accuracy in identifying cataract-infected images but also provides interpretable visual explanations of its predictions. Performance evaluation metrics such as accuracy, precision, recall, and F1 score demonstrate the model’s robustness, with a perfect accuracy of 100%. Grad-CAM visualizations highlight the key image regions—primarily around the iris and pupil—that contribute most to the model’s decision-making, making the system more transparent for clinical use. Additionally, novel statistical analysis methods, including saliency map evaluation metrics like AUC (Area Under the Curve) and the Pointing Game, were employed to quantify the quality of the model’s explanations. These metrics enhance the interpretability of the model and support its practical applicability in medical image analysis. This approach advances the integration of deep learning with explainable AI, offering a robust, accurate, and interpretable solution for cataract detection with the potential for broader adoption in ocular disease diagnosis and medical decision support systems.

Landslide Displacement Prediction Stacking Deep Learning Algorithms: A Case Study of Shengjibao Landslide in the Three Gorges Reservoir Area of China

Computational models enable accurate, timely prediction of landslides based on the monitoring data on-site as the development of artificial intelligence technology. The most existing prediction methods focus on finding a single prediction algorithm with excellent performance or an integrated and efficient hyperparameter optimization algorithm with a highly accurate regression prediction algorithm. In order to break through the limitation of generalization of prediction models, this paper proposes an ensemble model that combines deep learning algorithms, with a stacking framework optimized with the sliding window method. Multiple deep learning algorithms are set as the first layer of the stacking framework, which is optimized with the sliding window method to avoid confusion in the time order of datasets based on time series analysis. The Shengjibao landslide in the Three Gorges Reservoir is used as a case study. First, the cumulative displacement is decomposed into a trend and a periodic term using a moving average method. A single-factor and a multi-factor superposition model based on multiple deep learning algorithms are used to predict the trend and periodic term of the displacement, respectively. Finally, the predicted values of the trend and periodic terms are added to obtain the total predicted landslide displacement. For monitoring point ZK2-3, the values of RMSE and MAPE of the total displacement prediction with the stacking model are 15.93 mm and 0.54%, and the values of RMSE and MAPE of the best-performing individual deep learning model are 20.00 mm and 0.64%. The results show that the stacking model outperforms other models by combining the advantages of each individual deep learning algorithm. This study provides a framework for integrating landslide displacement prediction models. It can serve as a reference for the geological disaster prediction and the establishment of an early warning system in the Three Gorges Reservoir Area.