Deformation Source Revealed From Leveling Survey in Jigokudani Valley, Tateyama Volcano, Japan

We modeled vertical deformation detected from leveling survey in Jigokudani valley, Tateyama volcano, central Japan. In Jigokudani valley, uplift of 4 cm/year was previously detected during the period from 2007 to 2010 by Interferometric Synthetic Aperture Radar (InSAR). To confirm whether this inflation has continued to the present, we conducted leveling survey in Jigokudani valley since 2015. Most bench marks showed subsidence up to 5.6 cm during the four-year period from October 2016 to September 2020, while a bench mark locates at the center of the leveling route uniquely showed uplift of 1.6 cm. We applied a dislocation source model to the deformation using a grid search method. A crack with a length of 350 m, a width of 100 m, a strike of N117°E and a dip of 61° is located at a depth of 50 m near the center of Jigokudani valley (Koya jigoku and the new fumarolic area) where highly activating recently. Closing of the crack of 344 cm yields volume decreases of 120,400 m3. Striking direction of the crack is parallel to the line of which are old explosion craters (Mikurigaike and Midorigaike ponds) and corresponds to current maximum compressive stress field in the region of Hida Mountains including Tateyama volcano. The deformation source of the previous period from 2007 to 2010 detected from InSAR was estimated to be at a depth of 50 m and a gas chamber was correspondingly found from the audio-frequency magnetotelluric (AMT) survey. The estimated crack in this study is also located at a similar position of the gas chamber which was also identified from AMT survey. During the period from 2015 to 2016, the crack opened (i.e., inflated) and the inflation stopped during the next one-year period from 2016 to 2017. During the period from 2017 to 2020, the crack turned to closing (i.e., deflation), probably because of the increase in emission of volcanic fluid or gas with a formation of a new crater at the western side of Jigokudani valley (Yahata jigoku) during the period from 2017 to 2018.

Permanent, seasonal, and episodic seismic sources around Vatnajökull, Iceland, from the analysis of correlograms

In this study, we locate and characterise the main seismic noise sources in the region of the Vatnajökull icecap (Iceland). Vatnajökull is the largest Icelandic icecap, covering several active volcanoes. The seismic context is very complex, with glacial and volcanic events occurring simultaneously and the classification between the two can become cumbersome. Using seismic interferometry on continuous seismic data (2011–2019), we calculate the propagation velocities and locate the main seismic sources by using hyperbolic geometry and a grid-search method. We identify and characterise permanent oceanic sources, seasonal glacial-related sources, and episodic volcanic sources. These results give a better understanding of the background seismic noise sources in this region and could allow the identification of seismic sources associated with potentially threatening events in real-time.

Support Vector Regression Modelling of an Aerobic Granular Sludge in Sequential Batch Reactor

Support vector regression (SVR) models have been designed to predict the concentration of chemical oxygen demand in sequential batch reactors under high temperatures. The complex internal interaction between the sludge characteristics and their influent were used to develop the models. The prediction becomes harder when dealing with a limited dataset due to the limitation of the experimental works. A radial basis function algorithm with selected kernel parameters of cost and gamma was used to developed SVR models. The kernel parameters were selected by using a grid search method and were further optimized by using particle swarm optimization and genetic algorithm. The SVR models were then compared with an artificial neural network. The prediction results R2 were within >90% for all predicted concentration of COD. The results showed the potential of SVR for simulating the complex aerobic granulation process and providing an excellent tool to help predict the behaviour in aerobic granular reactors of wastewater treatment.

Optimally-Weighted Image-Pose Approach (OWIPA) for Distracted Driver Detection and Classification

Distracted driving is the prime factor of motor vehicle accidents. Current studies on distraction detection focus on improving distraction detection performance through various techniques, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs). However, the research on detection of distracted drivers through pose estimation is scarce. This work introduces an ensemble of ResNets, which is named Optimally-weighted Image-Pose Approach (OWIPA), to classify the distraction through original and pose estimation images. The pose estimation images are generated from HRNet and ResNet. We use ResNet101 and ResNet50 to classify the original images and the pose estimation images, respectively. An optimum weight is determined through grid search method, and the predictions from both models are weighted through this parameter. The experimental results show that our proposed approach achieves 94.28% accuracy on AUC Distracted Driver Dataset.

Optimized Hyperparameter Tuned Random Forest Regressor Algorithm in Predicting Resale Car Value based on Grid Search Method

The price of a car depreciates right from the time it is bought. The resale value of cars is influenced by many factors and influences both buyers and sellers, making it a prominent problem in the machine learning field. Diverse methodologies in machine learning can help us use all the varied factors and process a large amount of data to predict the cost. For our dataset, the Random Forest Regression algorithm shows a significant increase in the prediction rate. In order to optimise the Random Forest Regressor model, best hyperparameters can be found using hyperparameter tuning strategies. On comparing Grid Search and Randomized Search, a better prediction rate is accounted for using the former. These parameters are then passed to the algorithm as hyperparameter tuning can help collect the best batch of decision trees in the random forest for the most optimised prediction rate.

Klasifikasi Dialek Pengujar Bahasa Inggris Menggunakan Random Forest

Speech recognition is one of the important research fields which is currently widely used for various applications. However, speech recognition performance is affected by the dialect of the speaker. Therefore, dialect recognition is often used as an additional feature in speech recognition. The process of recognizing dialects is not easy. Currently, Machine Learning technology is widely applied in dialect recognition. One of the challenges in the introduction of machine learning-based dialects is the imbalance of classes and overlaps in a wide variety of classification techniques. This study applies Random Forest-based oversampling technology for dialect recognition. For hyper-parameter optimization of the random forest algorithm, we apply the Grid Search method. Experiments on Speech Accent Archive data using the MFCC feature resulted in an accuracy of 0.91 and AUC of 0.95

Optimal Feature Set Size in Random Forest Regression

One of the most important hyper-parameters in the Random Forest (RF) algorithm is the feature set size used to search for the best partitioning rule at each node of trees. Most existing research on feature set size has been done primarily with a focus on classification problems. We studied the effect of feature set size in the context of regression. Through experimental studies using many datasets, we first investigated whether the RF regression predictions are affected by the feature set size. Then, we found a rule associated with the optimal size based on the characteristics of each data. Lastly, we developed a search algorithm for estimating the best feature set size in RF regression. We showed that the proposed search algorithm can provide improvements over other choices, such as using the default size specified in the randomForest R package and using the common grid search method.

Non-destructive detection of highway hidden layer defects using a ground-penetrating radar and adaptive particle swarm support vector machine

In this paper, a method that uses a ground-penetrating radar (GPR) and the adaptive particle swarm support vector machine (SVM) method is proposed for detecting and recognizing hidden layer defects in highways. Three common road features, namely cracks, voids, and subsidence, were collected using ground-penetrating imaging. Image segmentation was performed on acquired images. Original features were extracted from thresholded binary images and were compressed using the kl algorithm. The SVM classification algorithm was used for condition classification. For parameter optimization of the SVM algorithm, the grid search method and particle swarm optimization algorithm were used. The recognition rate using the grid search method was 88.333%; the PSO approach often yielded local maxima, and the recognition rate was 86.667%; the improved adaptive PSO algorithm avoided local maxima and increased the recognition rate to 91.667%.

Research on a Multi-Parameter Fusion Prediction Model of Pressure Relief Gas Concentration Based on RNN

The effective prediction of gas concentration and the reasonable formulation of corresponding safety measures have important significance for improving the level of coal mine safety. To improve the accuracy of gas concentration prediction and enhance the applicability of the models, this paper starts with actual coal mine production monitoring data, improves the accuracy of gas concentration prediction through multi-parameter fusion prediction, and constructs a recurrent neural network (RNN)-based multi-parameter fusion prediction of coal face gas concentration. We determined the performance evaluation index of the model’s prediction method; used the grid search method to optimize the hyperparameters of the batch size; and used the number of neurons, the learning rate, the discard ratio, the network depth, and the early stopping method to prevent overfitting. The gas concentration prediction models—based on RNN and PSO-SVR and PSO-Adam-BP neural networks—were compared and analyzed experimentally with the mean absolute percentage error (MAPE) as the performance evaluation index. The result show that using the grid search method to adjust the batch size, the number of neurons, the learning rate, the discard ratio, and the network depth can effectively find the optimal hyperparameter combination. The training error can be reduced to 0.0195. Therefore, Adam’s optimized RNN gas concentration prediction model had higher accuracy and stability than the BP neural network and SVR. During training, the mean absolute error (MAE) could be reduced to 0.0573, and the root mean squared error (RMSE) could be reduced to 0.0167; however, the MAPE could be reduced to 0.3384% during prediction. The RNN gas concentration prediction model and parameter optimization method based on Adam optimization can effectively predict gas concentration. This method shows high accuracy in the prediction of gas concentration time series and can be used as a reference model for predicting mine gas concentration.