Shallow Learning Models Research Articles

A Deep Learning Approach for the Estimation of Glomerular Filtration Rate.

An accurate estimation of glomerular filtration rate (GFR) is clinically crucial for kidney disease diagnosis and predicting the prognosis of chronic kidney disease (CKD). Machine learning methodologies such as deep neural networks provide a potential avenue for increasing accuracy in GFR estimation. We developed a novel deep learning architecture, a deep and shallow neural network, to estimate GFR (dlGFR for short) and examined its comparative performance with estimated GFR from Modification of Diet in Renal Disease (MDRD) and Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations. The dlGFR model jointly trains a shallow learning model and a deep neural network to enable both linear transformation from input features to a log GFR target, and non-linear feature embedding for stage of kidney function classification. We validate the proposed methods on the data from multiple studies obtained from the NIDDK Central Database Repository. The deep learning model predicted values of GFR within 30% of measured GFR with 88.3% accuracy, compared to the 87.1% and 84.7% of the accuracy achieved by CKD-EPI and MDRD equations (p = 0.051 and p < 0.001, respectively). Our results suggest that deep learning methods are superior to equations resulting from traditional statistical methods in estimating glomerular filtration rate. Based on these results, an end-to-end predication system has been deployed to facilitate use of the proposed dlGFR algorithm.

Research on adaptive combined wind speed prediction for each season based on improved gray relational analysis.

The stability of the power grid and the operational security of the power system depend on the precise prediction of wind speed. In consideration of the nonlinear and non-stationary characteristics of wind speed in different seasons, this paper employs the weight of wind resource index calculated by triangular fuzzy analytic hierarchy process (TF-AHP), criteria importance through inter-criteria correlation (CRITIC), and entropy weight method (EWM) to improve gray correlation analysis (GRA) and obtain the gray correlation degree of each season. In addition, a wind speed prediction model is provided that includes single-layer and two-layer weighting and is based on both deep and shallow machine learning models. At first, we establish each quarter's wind resource characteristics at typical monthly intervals of 10min, 30min, 60min, and 120min. The GRA's TF-AHP-CRITIC-EWM, enhanced with subjective and objective weights, is used to assess the available wind resources in each season and to compute the forecasted combination of wind speed for each season. As the final prediction results, the prediction values of each layer model are evaluated independently. For the intervals with considerable errors, we apply wavelet denoising and replacement combination. The simulation findings show that the proposed combined model surpasses earlier benchmark models in terms of goodness of fit, prediction accuracy, and generalizability.

Air quality prediction using spatio-temporal deep learning

With the rapid urbanization, many developing countries are suffering from heavy air pollution. Predicting air quality accurately is critical for people’s outside planning and governments’ policy-making in these countries. This paper studies the air quality dataset collected in 22 stations of Beijing city, China, and reveals the spatial correlations, the temporal dependencies and feature correlations hidden in the dataset. Different from existing shallow or deep learning models that do not fully consider these correlations, we propose a spatio-temporal deep learning model called Conv1D-LSTM, which unifies 1-dimensional convolutional neural network (1D CNN) and long short-term memory (LSTM) for spatial and temporal correlation feature extraction, and uses a fully connected network (FCN) to exploit these features for air quality prediction. Moreover, this study investigates the problem of data missing in the air quality dataset and impute the missing values from spatial and temporal views. Extensive experiments are conducted by applying the proposed model to hourly PM2.5 and PM10 with single and multi-step predictions over the study area during 2014–2015. The results show the effectiveness of the proposed model and its superiority over other baseline models.

Gap Filling Cloudy Sentinel-2 NDVI and NDWI Pixels with Multi-Frequency Denoised C-Band and L-Band Synthetic Aperture Radar (SAR), Texture, and Shallow Learning Techniques

Multispectral imagery provides unprecedented information on Earth system processes: however, data gaps due to clouds and shadows are a major limitation. Normalized-Difference Vegetation Index (NDVI) and Normalized-Difference Water Index (NDWI) are two spectral indexes employed for monitoring vegetation phenology, land-cover change and more. Synthetic Aperture Radar (SAR) with its cloud-penetrating abilities can fill data gaps using coincident imagery. In this study, we evaluated C-band Sentinel-1, L-band Uninhabited Aerial Vehicle SAR (UAVSAR) and texture for gap filling using efficient machine learning regression algorithms across three seasons. Multiple models were evaluated including Support Vector Machine, Random Forest, Gradient Boosted Trees and an ensemble of models. The Gap filling ability of SAR was evaluated with Sentinel-2 imagery from the same date, 3 days and 8 days later than both SAR sensors in September. Sentinel-1 and Sentinel-2 imagery from winter and spring seasons were also evaluated. Because SAR imagery contains noise, we compared two robust de-noising methods and evaluated performance against a refined lee speckle filter. Mean Absolute Error (MAE) rates of the cloud gap-filling model were assessed across different dataset combinations and land covers. The results indicated de-noised Sentinel-1 SAR and UAVSAR with GLCM texture provided the highest predictive abilities with random forest R2 = 0.91 (±0.014), MAE = 0.078 (±0.003) (NDWI) and R2 = 0.868 (±0.015), MAE = 0.094 (±0.003) (NDVI) during September. The highest errors were observed across bare ground and forest, while the lowest errors were on herbaceous and woody wetland. Results on January and June imagery without UAVSAR were less strong at R2 = 0.60 (±0.036), MAE = 0.211 (±0.005) (NDVI), R2 = 0.61 (±0.043), MAE = 0.209 (±0.005) (NDWI) for January and R2 = 0.72 (±0.018), MAE = 0.142 (±0.004) (NDVI), R2 = 0.77 (±0.022), MAE = 0.125 (±0.004) (NDWI) for June. Ultimately, the results suggest de-noised C-band SAR with texture metrics can accurately predict NDVI and NDWI for gap-filling clouds during most seasons. These shallow machine learning models are rapidly trained and applied faster than intensive deep learning or time series methods.

Neighborhood graph embedding interpretable fault diagnosis network based on local and non-local information balanced under imbalanced samples

Currently, the monitoring of the health status of mechanical systems is becoming more and more critical, and the actual monitoring data is massive and high-dimensional, and these data are characterized mainly by imbalance. To solve the problem of low fault recognition rate caused by data imbalance, this article proposed a novel local and non-local information balanced neighborhood graph embedding interpretable deep autoencoder (LGBNGEDAE) method for rotating machinery fault diagnosis. Specifically, the local and non-local neighborhood information graph is embedded into the original objective function of the deep autoencoder to smooth the manifold structure of the data in LGBNEDSAE, which endows the deep learning model with better data learning capability and robustness of feature extraction as well as more reasonable network interpretability, making it more suitable for feature learning and classification of unbalanced data. Further, the method’s excellent fault diagnosis capability and generalization performance are verified on two opposed experimental cases of gearboxes. Compared with other deep learning architectures and shallow learning models, the method performs better in the face of unbalanced datasets.

Real-Time Estimation Frameworks for Feeder-Level Load Disaggregation and PEVs’ Charging Behavior Characteristics Extraction

In this article, a model-based real-time approach is proposed to disaggregate a feeder-level load and estimate the total energy of plug-in electric vehicles (PEVs), considering the controlled charging mode and the vehicle-to-grid capability of PEVs. To this end, aggregate demand of load categories participating at the feeder-head as well as the total energy of PEVs are analytically modeled. Then, the state-space representation of the system according to the mentioned models and their relations is proposed. Finally, a Kalman filter-based method is applied to disaggregate the feeder-level load into the aggregate demand of load categories and estimate the total energy of PEVs in real time. The accuracy and complexity of the proposed method are compared with two model-free methods, i.e., a nonlinear autoregressive with exogenous inputs-based shallow learning model and a long short-term memory-based deep learning approach, by using real data. They employ distribution substation measurements along with charging data of a very small subset of PEVs. The comparison results indicate that although the artificial neural network-based methods can effectively represent the nonlinear behavior of the feeder-level load and its components, the Kalman filter-based method significantly improves the PEVs’ total energy estimation by taking into account modeling and measurement uncertainties.

Deep Learning Approaches for Automatic Localization in Medical Images.

Recent revolutionary advances in deep learning (DL) have fueled several breakthrough achievements in various complicated computer vision tasks. The remarkable successes and achievements started in 2012 when deep learning neural networks (DNNs) outperformed the shallow machine learning models on a number of significant benchmarks. Significant advances were made in computer vision by conducting very complex image interpretation tasks with outstanding accuracy. These achievements have shown great promise in a wide variety of fields, especially in medical image analysis by creating opportunities to diagnose and treat diseases earlier. In recent years, the application of the DNN for object localization has gained the attention of researchers due to its success over conventional methods, especially in object localization. As this has become a very broad and rapidly growing field, this study presents a short review of DNN implementation for medical images and validates its efficacy on benchmarks. This study presents the first review that focuses on object localization using the DNN in medical images. The key aim of this study was to summarize the recent studies based on the DNN for medical image localization and to highlight the research gaps that can provide worthwhile ideas to shape future research related to object localization tasks. It starts with an overview on the importance of medical image analysis and existing technology in this space. The discussion then proceeds to the dominant DNN utilized in the current literature. Finally, we conclude by discussing the challenges associated with the application of the DNN for medical image localization which can drive further studies in identifying potential future developments in the relevant field of study.

Adversarial training for deep learning-based cyberattack detection in IoT-based smart city applications

Intrusion Detection Systems (IDS) based on deep learning models can identify and mitigate cyberattacks in IoT applications in a resilient and systematic manner. These models, which support the IDS’s decision, could be vulnerable to a cyberattack known as adversarial attack. In this type of attack, attackers create adversarial samples by introducing small perturbations to attack samples to trick a trained model into misclassifying them as benign applications. These attacks can cause substantial damage to IoT-based smart city models in terms of device malfunction, data leakage, operational outage and financial loss. To our knowledge, the impact of and defence against adversarial attacks on IDS models in relation to smart city applications have not been investigated yet. To address this research gap, in this work, we explore the effect of adversarial attacks on the deep learning and shallow machine learning models by using a recent IoT dataset and propose a method using adversarial retraining that can significantly improve IDS performance when confronting adversarial attacks. Simulation results demonstrate that the presence of adversarial samples deteriorates the detection accuracy significantly by above 70% while our proposed model can deliver detection accuracy above 99% against all types of attacks including adversarial attacks. This makes an IDS robust in protecting IoT-based smart city services.

A systematic review on cough sound analysis for Covid-19 diagnosis and screening: is my cough sound COVID-19?

For COVID-19, the need for robust, inexpensive, and accessible screening becomes critical. Even though symptoms present differently, cough is still taken as one of the primary symptoms in severe and non-severe infections alike. For mass screening in resource-constrained regions, artificial intelligence (AI)-guided tools have progressively contributed to detect/screen COVID-19 infections using cough sounds. Therefore, in this article, we review state-of-the-art works in both years 2020 and 2021 by considering AI-guided tools to analyze cough sound for COVID-19 screening primarily based on machine learning algorithms. In our study, we used PubMed central repository and Web of Science with key words: (Cough OR Cough Sounds OR Speech) AND (Machine learning OR Deep learning OR Artificial intelligence) AND (COVID-19 OR Coronavirus). For better meta-analysis, we screened for appropriate dataset (size and source), algorithmic factors (both shallow learning and deep learning models) and corresponding performance scores. Further, in order not to miss up-to-date experimental research-based articles, we also included articles outside of PubMed and Web of Science, but pre-print articles were strictly avoided as they are not peer-reviewed.

An Empirical Study of the Effectiveness of an Ensemble of Stand-alone Sentiment Detection Tools for Software Engineering Datasets

Sentiment analysis in software engineering (SE) has shown promise to analyze and support diverse development activities. Recently, several tools are proposed to detect sentiments in software artifacts. While the tools improve accuracy over off-the-shelf tools, recent research shows that their performance could still be unsatisfactory. A more accurate sentiment detector for SE can help reduce noise in analysis of software scenarios where sentiment analysis is required. Recently, combinations, i.e., hybrids of stand-alone classifiers are found to offer better performance than the stand-alone classifiers for fault detection. However, we are aware of no such approach for sentiment detection for software artifacts. We report the results of an empirical study that we conducted to determine the feasibility of developing an ensemble engine by combining the polarity labels of stand-alone SE-specific sentiment detectors. Our study has two phases. In the first phase, we pick five SE-specific sentiment detection tools from two recently published papers by Lin et al. [ 29 , 30 ], who first reported negative results with stand alone sentiment detectors and then proposed an improved SE-specific sentiment detector, POME [ 29 ]. We report the study results on 17,581 units (sentences/documents) coming from six currently available sentiment benchmarks for software engineering. We find that the existing tools can be complementary to each other in 85-95% of the cases, i.e., one is wrong but another is right. However, a majority voting-based ensemble of those tools fails to improve the accuracy of sentiment detection. We develop Sentisead, a supervised tool by combining the polarity labels and bag of words as features. Sentisead improves the performance (F1-score) of the individual tools by 4% (over Senti4SD [ 5 ]) – 100% (over POME [ 29 ]). The initial development of Sentisead occurred before we observed the use of deep learning models for SE-specific sentiment detection. In particular, recent papers show the superiority of advanced language-based pre-trained transformer models (PTM) over rule-based and shallow learning models. Consequently, in a second phase, we compare and improve Sentisead infrastructure using the PTMs. We find that a Sentisead infrastructure with RoBERTa as the ensemble of the five stand-alone rule-based and shallow learning SE-specific tools from Lin et al. [ 29 , 30 ] offers the best F1-score of 0.805 across the six datasets, while a stand-alone RoBERTa shows an F1-score of 0.801.

A novel adaptive convolutional neural network for fault diagnosis of hydraulic piston pump with acoustic images

As an essential part of hydraulic transmission systems, hydraulic piston pumps have a significant role in many state-of-the-art industries. Thus, it is important to implement accurate and effective fault diagnosis of hydraulic piston pumps. Owing to the heavy reliance of shallow machine learning models on the expertise and experience of engineers, fault diagnosis based on deep models has attracted significant attention from academia and industry. To construct a deep model with good performance, it is necessary and challenging to tune the hyperparameters (HPs). Since many existing methods focus on manual tuning and use common search algorithms, it is meaningful to explore more intelligent algorithms that can automatically optimize the HPs. In this paper, Bayesian optimization (BO) is employed for adaptive HP learning, and an improved convolutional neural network (CNN) is established for fault feature extraction and classification in a hydraulic piston pump. First, acoustic signals are transformed into time–frequency distributions by a continuous wavelet transform. Second, a preliminary CNN model is built by setting initial HPs. The range of each HP to be optimized is identified. Third, BO is employed to select the optimal combination of HPs. An improved model called CNN-BO is constructed. Finally, the diagnostic efficiency of CNN-BO is analyzed using a confusion matrix and t-distributed stochastic neighbor embedding. The classification performance of different models is compared. It is found that CNN-BO has a higher accuracy and better robustness in fault diagnosis for a hydraulic piston pump. This research will provide a basis for ensuring the reliability and safety of the hydraulic pump.

Deep-Cov19-Hate: A Textual-Based Novel Approach for Automatic Detection of Hate Speech in Online Social Networks throughout COVID-19 with Shallow and Deep Learning Models

Deep-Cov19-Hate: A Textual-Based Novel Approach for Automatic Detection of Hate Speech in Online Social Networks throughout COVID-19 with Shallow and Deep Learning Models

A multimodal approach to chaotic renewable energy prediction using meteorological and historical information

Wind energy, which exhibits non-stationarity, randomness, and intermittency, is inextricably linked to meteorological data. The wind power series can be broken down into several subsequences using data decomposition techniques to make forecasting simpler and more accurate. Because of this, a single prediction model does not perform well in extracting hidden information from each subsequence. To predict different frequency series, this paper employed shallow and deep learning models and proposed an improved hybrid wind power prediction model based on secondary decomposition, extreme learning machines (ELM), convolutional neural networks (CNN), and bidirectional long short-term memory (BiLSTM). To begin, secondary decomposition was employed to break down the wind power series into several components. The ELM was used to forecast the low-frequency components. Following that, CNN was utilized to reintegrate the input characteristics of the high-frequency components, followed by BiLSTM prediction. Finally, the forecasting values for each component were added to generate the final prediction results. For one-, two-, and three-step predictions, the model was applied to the La Haute Borne wind farm. Additionally, four comparative experiments were conducted to validate the model’s usefulness. The suggested model’s mean absolute error (MAE), mean absolute percentage error (MAPE), and R-squared (R2) values for one-step prediction of the March data were 14.87 kW, 22.24 kW, and 0.984, respectively, which indicate the proposed model’s superiority to other prediction models.

Convolutional neural networks for intra-hour solar forecasting based on sky image sequences

Accurate and timely solar forecasts play an increasingly critical role in power systems. Compared to longer forecasting timescales, very short-term solar forecasting has lagged behind in both research and practice. In this paper, we propose deep convolutional neural networks (CNNs) to provide operational intra-hour (10-minute-ahead to 60-minute-ahead) solar forecasts. We develop two CNN structures inspired by a widely-used CNN architecture. The CNNs are tailored to our solar forecasting regression tasks and rely solely on sky image sequences. Case studies based on six years of data (over 150,000 data points) demonstrate that the best CNN model has forecast skill scores of 20%–39% over the naive persistence of cloudiness benchmark, even at these very short timescales. The CNNs also have consistently superior performance when compared to shallow machine learning models with meteorological predictors, where the improvement averages around 7%. The sensitivity analyses show that the sky image length, resolution, and weather conditions have impacts on the deep learning model accuracy. In our intra-hour problem with specific setups, two sky images with a 10-minute 128 × 128 resolution yield the most accurate forecasts. Current limitations, future work, and deployment challenges and solutions are also discussed.

Saturation Line Forecasting via a Channel and Temporal Attention-Based Network

Tailings ponds are places for storing industrial waste. The saturation line is the crucial factor in quantifying the safety of tailings ponds. Existing saturation line time-series prediction methods are mainly based on statistical models or shallow machine learning models. Although these models aim to capture the time dependence of the sequence data, the channel and temporal are even unavailable in principle. To mitigate this problem, in this paper, we present a two-stage forecasting method, which embeds the channel and temporal attention into a hybrid CNN-LSTM model to predict the saturation line. The channel and temporal attention are utilized to capture subtle high-dimensional time-series dependence. In the first stage, the discrete wavelet transform (DWT) is applied to capture the refined sequence information. In the second stage, the CNN-LSTM model is utilized to learn the basic spatial and temporal features in the time series. Furthermore, the channel and temporal attention model are embedded into the CNN-LSTM model to enhance the feature-extracting ability in the channel and temporal dimensions. Consequently, our proposed model is shown to outperform classic models on multiple real-world datasets in terms of RMSE, MAPE, R <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and MAE, respectively.

DRLSTM: A dual-stage deep learning approach driven by raw monitoring data for dam displacement prediction

Dam displacement is an important indicator of the overall dam health status. Numerical prediction of such displacement based on real-world monitoring data is a common practice for dam safety assessment. However, the existing methods are mainly based on statistical models or shallow machine learning models. Although they can capture the timing of the dam displacement sequence, it is difficult to characterize the complex coupling relationship between displacement and multiple influencing factors (e.g., water level, temperature, and time). In addition, input factors of most dam displacement prediction models are artificially constructed based on modelers’ personal experience, which lead to a loss of valuable information, thus prediction power, provided by the full set of raw monitoring data. To address these problems, this paper proposes a novel dual-stage deep learning approach based on one-Dimensional Residual network and Long Short-Term Memory (LSTM) unit, referred to herein as the DRLSTM model. In the first stage, the raw monitoring sequence is processed and spliced with convolution to form a combined sequence. After the timing information is extracted, the convolution direction is switched to learn the complex relationship between displacement and its influencing factors. LSTM is used to extract this relationship to obtain Stage I prediction. The second stage takes the difference between the actual measurement and the Stage I prediction as inputs, and LSTM extracts the stochastic features of the monitoring system to obtain Stage II prediction. The sum of two stage predictions forms the final prediction. The DRLSTM model only requires raw monitoring data of water level and temperature to accurately predict displacement. Through a real-world comparative study against four commonly used shallow learning models and three deep learning models, the root mean square error and mean absolute error of our proposed method are the smallest, being 0.198 mm and 0.149 mm respectively, while the correlation coefficient is the largest at 0.962. It is concluded that the DRLSTM model performance well for evaluating dam health status.

Comparing multi-step ahead building cooling load prediction using shallow machine learning and deep learning models

Accurate building cooling load prediction is beneficial in managing optimal operation to conserve energy user and operational cost. Several physics-based and data-driven models proposed to forecast building cooling load focus on one-step-ahead prediction. Deep learning-based Long-term memory (LSTM) models are shown to perform well for one-step short-term (1Hour Ahead) building cooling load prediction. However, no prior studies have examined the prediction performance of shallow machine learning methods over deep learning models to forecast building cooling demand over multi-steps across diverse real-world datasets. A multi-step model learns a single parametric function from input time series and forecasts an array of building cooling load values (multi-step) simultaneously. A comprehensive study has been carried out to evaluate the performance of six data-driven models (2 shallow learning, 3 deep sequential learning, and 1 heuristic method) to predict multi-step long-term (1Day Ahead) building cooling load. Our results demonstrate variant of the LSTM model, the Recurrent Neural Network Multi-Input Multi-Output (RNN-MIMO) network architecture, performs consistently well compared to its deep learning counterparts and shallow machine learning techniques, both tree boosting and support vector regression. Notable conclusions from results obtained are twofold: Firstly, Long short-term memory (LSTM) based RNN-MIMO architecture performs well in both short-term (1Hour Ahead) and long-term (1Day Ahead) multi-step forecast horizon. RNN-MIMO is up to 33% more accurate, on average, in terms of mean absolute error over existing, state-of-the-art shallow machine learning models both Support Vector Regression (SVR) and tree boosting techniques (XGBoost). Our findings have significant implications for practice. Notably, machine learning models trained on one-step-ahead predictions cannot be deployed readily to predict multiple time steps into the future since longer prediction horizons impose additional training and fine-tuning efforts over each of the multiple steps. RNN-MIMO model’s ability to predict multiple time steps simultaneously eliminates the need for manual fine-tuning of individual models for each required forecast horizon.

A buffered online transfer learning algorithm with multi-layer network

Online transfer learning (OTL) has attracted much attention in recent years. It is designed to handle the transfer learning tasks, where the data of the target domain isn’t available in advance but may arrive in an online manner, which may be a more realistic scenario in practice. However, there typically are two limitations of existing OTL algorithms. 1) Existing OTL algorithms are based on shallow online learning models (SOLMs), e.g., linear or kernel models. Due to this limitation of SOLMs they cannot effectively learn complex nonlinear functions in complicated application and the OTL algorithms based on SOLMs cannot either. 2) Existing algorithms only utilize the latest arrived instance to adjust the model. In this way, the previously arrived instances are not utilized. It may be better to utilize the previously arrived instances as well. In this paper, to overcome the abovementioned two limitations, a buffered online transfer learning (BOTL) algorithm is proposed. In the proposed BOTL algorithm, the learner is designed as a deep learning model, referred to as Online Hedge Neural Network (OHNN). In order to enable the OHNN to be effectively learned in an online manner, we propose a buffered online learning framework that utilizes several previously arrived instances to assist learning. Further, to enhance the performance of the OHNN, a model learned in the source domain is transferred to the target domain. The regret bound of the proposed BOTL algorithm is analyzed theoretically. Experimental results on realistic datasets illustrate that the proposed BOTL algorithm can achieve lower mistake rate than the algorithms compared.

Evaluation of a new dataset for visual detection of cervical precancerous lesions

Automated visual evaluation (AVE) is an emerging method to detect and diagnose cervical precancerous lesions by imaging and analysis via a deep learning classifier. Challenges in AVE development come from not only the limited data available, but also a proper design of the learning protocol. The most analyzed dataset (PEG) is traced to a clinical trial at a single site, where all the images were captured in a well controlled environment. Recently, cervical images have been captured by a light-weight mobile solution where the screening images were collected from a wider user pool at many sites. This paper introduces a new data resource (EVA dataset), collected by providers using a mobile colposcope during their routine practice. Compared to PEG, EVA images contain higher levels of data variations and exhibits a different distribution over multiple image attributes including image sharpness, brightness and colorfulness. In order to evaluate the practical value of EVA dataset for cervical high-grade squamous intraepithelial lesions (SIL) diagnosis, we further present an analysis of how a deep learning based framework can be used with both datasets by evaluating three key technical components: (1) region-of-interest (ROI) detection, (2) data augmentation and (3) pre-trained deep learning model selection. Our results indicate that ROI detection and shallow deep learning models usually help the detection on both datasets. While most data augmentations are effective on the EVA dataset, the improvement is less pronounced on PEG. Overall, using a deep-learning based framework looks promising for high-grade SIL diagnosis but there is still large room for improvements, especially on the EVA dataset. The noted differences indicate that the EVA dataset presents more practical challenges for high-grade SIL diagnosis and AVE classifier development.

Accurate Prediction of Earthquake-Induced Landslides Based on Deep Learning Considering Landslide Source Area

An earthquake-induced landslide (EQIL) is a rapidly changing process occurring at the Earth’s surface that is strongly controlled by the earthquake in question and predisposing conditions. Predicting locations prone to EQILs on a large scale is significant for managing rescue operations and disaster mitigation. We propose a deep learning framework while considering the source area feature of EQIL to model the complex relationship and enhance spatial prediction accuracy. Initially, we used high-resolution remote sensing images and a digital elevation model (DEM) to extract the source area of an EQIL. Then, 14 controlling factors were input to a stacked autoencoder (SAE) to search for robust features by sparse optimization, and the classifier took advantage of high-level abstract features to identify the EQIL spatially. Finally, the EQIL inventory collected from the Wenchuan earthquake was used to validate the proposed model. The results show that the proposed method significantly outperformed conventional methods, achieving an overall accuracy (OA) of 91.88%, while logistic regression (LR), support vector machine (SVM), and random forest (RF) achieved 80.75%, 82.22%, and 84.16%, respectively. Meanwhile, this study reveals that shallow machine learning models only take advantage of significant factors for EQIL prediction, but deep learning models can extract more effective information related to EQIL distribution from low-value density data, which is why its prediction accuracy is growing with increasing input factors. There is hope that new knowledge of EQILs can be represented by high-level abstract features extracted by hidden layers of the deep learning model, which are typically acquired by statistical methods.