Training Sample Set Research Articles

SVM-Based Real-Time Identification Model of Dangerous Traffic Stream State

By comparing and studying the correlation between traffic stream parameters and traffic safety of different highways, the correlations of traffic natural quantity, traffic equivalent, passenger-cargo ratio, car following percentage, congestion degree, and time occupancy rate are obtained. The traffic stream state before the actual accident is used as the criterion to judge the bad traffic stream state. The main parameters are obtained by extracting the parameters from the traffic stream data at the lane level and reducing the dimension of the parameters with the principal component analysis method. Establish a SVM model for RT early warning of traffic stream safety. Compared with other methods, the adaptive parameter selection method can adaptively select parameters according to the training sample set, realize the adaptive ability of the forecast model, and effectively improve the forecast accuracy of traffic stream. This paper studies the risk early warning model of road traffic accidents, which can transform the problem of road traffic safety into active early warning and improve the level of traffic safety. This study provides safety management measures for highway operation departments, which has certain theoretical significance and practical application value.

Genetic algorithm based artificial neural network and partial least squares regression methods to predict of breakdown voltage for transformer oils samples in power industry using ATR-FTIR spectroscopy

The current study proposes a novel analytical method for calculating the breakdown voltage (BV) of transformer oil samples considered as a significant method to assess the safe operation of power industry.Transformer oil samples can be analyzed using the Attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy combined with multivariate calibration methods. The partial least squares regression (PLSR) back propagation–artificial neural network (BP-ANN) methods and a genetic algorithm (GA) for variable selection are used to predict and assess breakdown voltage in transformer oil samples from various Iranian transformer oils.As a result, the root mean square error (RMSE) and correlation coefficient for the training and test sets of oil samples are also calculated. In the GA-PLS-R method, the squared correlation coefficient (R2pred) and root mean square prediction error (RMSEP) are 0.9437 and 2.6835, respectively. GA-BP-ANN, on the other hand, had a lower RMSEP value (0.2874) and a higher R2pred function (0.9891).Considering the complexity of transformer oil samples, the performance of GA-BP-ANN has resulted in an efficient approach for predicting breakdown voltage; consequently, it can be effectively used as a new method for quantitative breakdown voltage analysis of samples to evaluate the health of transformer oil..

Aseismic Optimization of Mega-sub Controlled Structures Based on Gaussian Process Surrogate Model

Due to the complex seismic characteristics of mega-sub controlled structures (MSCS), it is difficult to give full play to their advantages in earthquake resistance by traditional design methods. Meta-heuristic optimization algorithms can be used to improve the seismic performance, but the structural response needs to be calculated repeatedly, which results in high computation cost. To overcome these challenges, an efficient aseismic optimization design procedure for engineering application is developed. In this procedure, the model of optimization problem is established based on time history analysis (THA). Gaussian process regression (GPR) surrogate models are employed to predict the values of the objective and constraint functions. The expected improvement (EI) and constrained expected improvement (CEI) criteria are adopted to update the training sample set and obtain the optimal solution. Then, two examples are presented to validate the effectiveness and efficiency of this method in optimization problems of structures under earthquake loads. Finally, it is applied to optimizations of a MSCS and a mega frame structure (MFS), respectively. The response and cost of the optimized structures are reduced, and the MSCS shows better earthquake resistant capacities.

Research on fault diagnosis of automobile engines based on the deep learning 1D-CNN method

Traditional engine fault diagnosis methods usually need to extract the features manually before classifying them by the pattern recognition method, which makes it difficult to solve the end-to-end fault diagnosis problem. In recent years, deep learning has been applied in different fields, bringing considerable convenience to technological change, and its application in the automotive field also has many applications, such as image recognition, language processing, and assisted driving. In this paper, a one-dimensional convolutional neural network (1D-CNN) in deep learning is used to process vibration signals to achieve fault diagnosis and classification. By collecting the vibration signal data of different engine working conditions, the collected data are organized into several sets of data in a working cycle, which are divided into a training sample set and a test sample set. Then, a one-dimensional convolutional neural network model is built in Python to allow the feature filter (convolution kernel) to learn the data from the training set and these convolution checks process the input data of the test set. Convolution and pooling extract features to output to a new space, which is characterized by learning features directly from the original vibration signals and completing fault diagnosis. The experimental results show that the pattern recognition method based on a one-dimensional convolutional neural network can be effectively applied to engine fault diagnosis and has higher diagnostic accuracy than traditional methods.

Compressive strength prediction of basalt fiber reinforced concrete via random forest algorithm

Basalt fiber is a green non-polluting material with strong tensile mechanical properties. In this paper, the strength prediction model of basalt fiber concrete is constructed by random forest method based on experiments to study the strengthening effect of basalt fiber in concrete composites and explore the impact of different fiber sizes and fiber content on the strength of basalt fiber concrete. First, the compressive strength tests of BFRC under different stress states were performed at fiber volume fractions of 0.2%, 0.4%, 0.6% and fiber lengths of 6 mm, 12 mm and 18 mm to obtain stress–strain curves and peak compressive strength. Then, 70% of the original test data is used to establish a random forest training sample set, and the remaining 30% is used as a test set. According to the error outside the bag, the appropriate number of decision trees and leaf nodes are selected, and the influencing factors are ranked by importance. Then use the Pearson correlation diagram to analyze the correlation of each influencing factor, establish a random forest training model, and output the fitting prediction results of the model training set and the prediction set. Finally, the prediction results of random forest with BP neural network and support vector regression are measured by MSE, RMSE, and R2 evaluation metrics for performance. The results show that the random forest prediction model has a good model fitting ability. Compared with other algorithms, the accuracy of the RF model in the MSE index is increased by 8% and 18.8%, which further verifies the accuracy and reliability of the model.

Designing staggered platelet composite structure with Gaussian process regression based Bayesian optimization

The staggered platelet composite structure, one of the most well-known examples of biomimetics, is inspired by the microstructure of nacre, where stiff mineral platelets are stacked with a small fraction of soft polymer in a brick-and-mortar style. Significant efforts have been made to establish a framework for designing a staggered platelet pattern that achieves an excellent balance of toughness and stiffness. However, because no analytical formula for accurately predicting its toughness is available because of the complexity of the failure mechanism of realistic composites, existing studies have investigated either idealized composites with simplified material properties or realistic composites designed by heuristics. In the present study, we propose a Bayesian optimization framework to design a staggered platelet structure that renders high toughness. Gaussian process regression (GPR) was adopted to model statistically the complex relationship between the shape of the staggered platelet array and the resultant toughness. The Markov chain Monte Carlo algorithm was used to determine the optimal kernel hyperparameter set for the GPR. Starting with 14 initial training data collected with uniaxial tensile tests, a GPR-based Bayesian optimization using the expected improvement (EI) acquisition function was carried out. As a result, it was possible to design a staggered platelet pattern with a toughness 12% higher than that of the best sample in the initial training set, and this improvement was achieved after five iterations of our optimization cycle. As this optimization framework does not require any material theories and models, this process can be easily adapted and applied to various other material optimization problems based on a limited set of experiments or computational simulations.

Multi-objective optimization of fixture locating layout for sheet metal part using Kriging and MOBA

Owing to the multi-objective attribute and excessive computational burden of finite element analysis (FEA) during the optimization of fixture locating layout, a new multi-objective optimization method for the sheet metal fixture design based on the Kriging surrogate model and multi-objective bat algorithm (MOBA) is proposed in this paper. The fixture locating layout is treated as the design variable while the overall deformation and maximum deformation of sheet metal under its dead weight are treated as the objective functions. The training and testing sample sets are generated by Latin hypercube sampling (LHS) and FEA. Kriging surrogate model is built by learning from the training sample set and the multi-objective optimization model for fixture locating layout is constructed based on the established Kriging surrogate model. MOBA is employed to find Pareto optimal solution set for the sheet metal fixture locating layout. Finally, a bi-objective optimization case of fixture locating layout for the sheet metal part based on the “4-2-1” locating principle is conducted to verify the proposed method.

Application of Artificial Neural Network Algorithm in Facial Biological Image Information Scanning and Recognition.

In order to solve the problem that the accuracy of face recognition is not good due to the influence of jitter and environmental factors in the mobile shooting environment, this paper proposes the application of an artificial neural network algorithm in the information scanning and recognition of face biological images. The spatial neighborhood information is integrated into the amplitude detection of multipose face images, and the dynamic corner features of multipose face images are extracted. The structural texture information of multipose face images is compared to a global moving RGB three-dimensional bit plane random field, and the multipose face images are detected and fused. At different scales, appropriate feature registration functions are selected to describe the feature points of multipose face images. The parallax analysis of target pixels and key feature detection of multipose face images are carried out. Image stabilization and automatic recognition are realized by combining artificial neural network learning and feature registration methods. The experimental results show that the experiment is designed with MATLAB, the frame frequency of dynamic face image acquisition is 1200 kHz, the number of collected samples of the multipose face image is 2000, the training sample set is 200, the noise coefficient = 0.24, the number of multipose face image blocks is 120, and the structural similarity is 0.12. It is found that the output signal-to-noise ratio of multipose face image recognition using the method in this paper is high. Conclusion. This method has good performance in feature point registration and high recognition accuracy for multipose face image recognition.

The problem and neural network model of optimization of the farm sown areas structure

Optimization of agriculture is becoming a priority for the modern economy. Today it is important to take measures to form and develop in Ukraine both large agricultural enterprises and small farms. Managers of agricultural enterprises are constantly looking for rational and effective means to reduce costs, which involves the process of optimizing agriculture. However, previously established traditional cost accounting tools are becoming inefficient, time-consuming and inaccessible to business leaders, as qualitatively different means of fixing costs are needed to optimize agriculture. The purpose of this work is to increase the profitability of crop production by developing neural network technology to optimize the structure of sown areas of the farm. The objectives of the work are: – to determine the factors influencing the profitability of crop production and the structure of the training set of crop data for previous years; – to determine the structural elements and build a model to solve the problem of increasing the profitability of crop production, taking into account crop rotations; – experimental verification of the developed technology using a set of training and control data samples. To achieve the goals of the work were used: methods of mathematical modeling, methods of decision theory, evolutionary algorithms, neural networks. To perform these tasks, the concepts of crop rotation were studied and auxiliary data tables were developed, containing information on the level of inexpediency of crops of the same crop, the level of inexpediency of crops of one crop after another. Based on these data, the coefficient of estimated crop yield was calculated. A set of data was formed containing information on 10 fields, each of which was randomly "sown" with agricultural crops, and based on the value of this crop and the coefficient of probable yield of the crop, the probable profit was calculated. A genetic algorithm has been developed in which the value of fitness function was determined using a neural network trained in the generated data set. On the generated data set, the operation of the algorithm was tested, which showed the effectiveness of the proposed technology.

Big Data Application in Forecasting Financial Investment of e-Commerce Industry for Sustainability

With the rapid development of e-commerce, financial investment forecasting in the e-commerce industry has gradually become a concern of relevant personnel. Based on DBN, the study proposes a PVD prediction model. For training and test sample sets, the PLR_VIP algorithm is calculated and min-max normalization is applied to the original financial time series. To determine appropriate network parameters, the DBN network is trained and tested, and then Elliott wave patterns are predicted based on financial time series. The experimental results show that the MSE of the PVD model is 0.4015 and the prediction accuracy is 70.21%, indicating that it can efficiently and accurately identify the Elliott wave pattern of financial time series. Comparing the prediction results of the PVD model with the other five models, the values of the four evaluation indicators of PVD are the lowest among all models, which are 0.6336, 0.4015, 0.9052, and 29.79%, respectively. Compared with the training error changes of other models, it can be seen that the error curve of the DBN network is smoother and the training error is smaller. It shows that it has higher stability, faster convergence speed, higher reliability and accuracy, and shows excellent prediction performance, which is significantly better than other models. Experiments show that under the background of sustainable development, the PVD forecasting model proposed in the study performs well in financial investment forecasting, which provides a reference for the development of financial investment forecasting in the e-commerce industry.

Redox state as assessed using the measurement of human non-mercaptalbumin in embryo culture media is associated with successful embryo development in human in vitro fertilization.

The role of oxidative stress in the pathogenesis of various diseases has been attracting attention. We speculated as to whether the redox state of treatment solutions used for various diseases may play a role in treatment success. In the current study, we focused on the human embryo culture medium used for in vitro fertilization (IVF). A total of 173 oocytes from a total of 91 patients treated with IVF were enrolled. The redox state was assessed by measuring the levels of human non-mercaptalbumin (HNA). We analyzed factors related to blastocyst formation on day 5 or 6 after insemination. We also developed a random forest (RF) model for the prediction of blastocyst formation. The variable importance in the predictive model was assessed using the mean decrease in the Gini impurity. Blastocyst formation was observed in 41.04% (71/173) of the oocytes and was associated with a lower %HNA in the culture medium, a younger patient age, and the fertilization method (standard IVF or intracytoplasmic sperm injection). The RF model developed using these factors and 70% of the samples (training set, n = 121) was validated in the remaining testing set (n = 52) and produced an area under the curve of 0.761, where the %HNA in the culture medium was the most important variable for predicting blastocyst formation. In conclusion, lower levels of oxidative stress in embryo culture media were associated with the success of IVF treatment. The redox state of treatment solutions should be considered to support treatment success.

Training Data Improvement by Automatic Generation of Semantic Networks for Bias Mitigation

The significance of Bias Detection has increased appreciably, due to the increased application of AI. Although syntactic bias is well explored with statistical techniques, there remains semantic bias challenge like for example, Google’s face recognition which excludes colored people. Human expertise is required to detect semantic bias, e.g., for the application of the root-out-bias method. We propose a further automatization to this laborious method, based on the Training Data Improvement for Bias Mitigation (TDIBM). The concept, is to automatically construct a Semantic Network (SN) from the domain description of the training. For the semantic network nouns are extracted. As a second step, synonyms and semantically similar nouns are searched, e.g. in dictionaries, and added to the SNs. As a result, the SN contains nouns that enhances the given domain, with previously unknown knowledge. This SN can be used to check with, e.g., the root-out bias method, whether the training sample is biased, or not. Should the training sample be biased, then the corresponding nouns from the SN can be added to the training sample set to mitigate the bias. The newly developed method, TDIBM is evaluated twofold: Firstly, with the description of the COMPAS system, which is a case management and decision support tool used by U.S. courts to assess the likelihood of a defendant becoming a recidivist. Secondly, an autonomous driving domain is applied, to investigate accidental driving of a Tesla car. Here TDIBM detected among many new features, including one to solve ambiguous scene interpretations for autonomous driving vehicles.

Iterative Spatial-Spectral Training Sample Augmentation for Effective Hyperspectral Image Classification

Factors such as insufficient training samples, high-dimensional data features, and unbalanced data classes can degrade the accuracy of hyperspectral classification. To this end, this paper proposes an iterative spatial-spectral training sample augmentation (ITSA) algorithm and a new classification model incorporating maximum marginal projection (ITSA-MMP). First, ITSA iteratively augments samples by a similar region clustering strategy (SRCS) integrating spatial-spectral metric. Then, box-plot for representative sample selection (BPRSS) is adopted to screen optimal samples for the final augmented sample set (ASS). Next, based on the ASS, MMP projects the hyperspectral image into a low-dimensional subspace to explore the local structure of the data manifold and improve the inter-class separability of the data. Finally, the MMP-reduced data is classified by support vector machine. Experiments on two real hyperspectral datasets validate that ITSA-MMP can effectively increase the training sample set especially for small initial sample set and unbalanced dataset, and obtain a higher classification accuracy.

Meta-Learning Paradigm and CosAttn for Streamer Action Recognition in Live Video

As an emerging field of network content production, live video has been in the vacuum zone of cyberspace governance for a long time. Streamer action recognition is conducive to the supervision of live video content. In view of the diversity and imbalance of streamer actions, it is attractive to introduce few-shot learning to realize streamer action recognition. Therefore, a meta-learning paradigm and CosAttn for streamer action recognition method in live video is proposed, including: (1) the training set samples similar to the streamer action to be recognized are pretrained to improve the backbone network; (2) video-level features are extracted by R(2+1)D-18 backbone and global average pooling in the meta-learning paradigm; (3) the streamer action is recognized by calculating cosine similarity after sending the video-level features to CosAttn to generate a streamer action category prototype. Experimental results on several real-world action recognition datasets demonstrate the effectiveness of our method.

Application of English Vocabulary Presentation Based on Clustering in College English Teaching

As the most basic element in English learning, vocabulary has always been the focus of teaching in college English classes, but the teaching effect is often unsatisfactory. In this paper, the genetic algorithm fitness function design part is integrated with the K-medoids algorithm to form K-GA-medoids, and secondly, it is combined with KNN to form an algorithmic framework for English vocabulary classification. In the classification process, clustering and classification steps are taken to realize the reduction of the training set samples and thus reduce the computational overhead. The experiments show that K-GA-medoids have significantly improved the clustering effect compared with traditional K-medoids, and the combination of K-GA-medoids and KNNs has effectively improved the efficiency of English vocabulary classification compared with the traditional KNN algorithm, while ensuring the classification accuracy. We found that students in college English course consider word memorization as a difficult learning task, and the traditional vocabulary teaching methods are not very effective, and the knowledge of etymology is often little known and rarely covered in classroom lectures. Therefore, the article explores new ideas and strategies for teaching vocabulary in college English from the perspective of etymology.

Unsupervised cross-domain person re-identification by instance and distribution alignment

Most existing person re-identification (re-id) methods assume supervised model training on a separate large set of training samples from the target domain. While performing well in the training domain, such trained models are seldom generalisable to a new independent unsupervised target domain without further labelled training data from the target domain. To solve this scalability limitation, we develop a novel Hierarchical Unsupervised Domain Adaptation (HUDA) method. It can transfer labelled information of an existing dataset (a source domain) to an unlabelled target domain for unsupervised person re-id. Specifically, HUDA is designed to model jointly global distribution alignment and local instance alignment in a two-level hierarchy for discovering transferable source knowledge in unsupervised domain adaptation. Crucially, this approach aims to overcome the under-constrained learning problem of existing unsupervised domain adaptation methods. Extensive evaluations show the superiority of HUDA for unsupervised cross-domain person re-id over a wide variety of state-of-the-art methods on four re-id benchmarks: Market-1501, DukeMTMC, MSMT17 and CUHK03.

Performance of SVM technique for DoA Estimation in 5G mm-Wave band

Applying Machine Learning algorithms in wireless communication has shown increasing interest due to the increase of demand on capacity, the increase of the number of users, and equipment sharing the limited frequency spectrum resources. Also, the need for a reduction in power consumption at base stations and the optimization of radio coverage make ML an attractive and promising technique. In this paper, we investigate the usage of Support Vector Machine (SVM) technique for Direction of Arrival (DoA) estimation in the millimeter-wave band. The objective is to predict the location of a user in a given area by analyzing the received signals at an array of antennas, using an SVM-based model. The first phase of this technique consists of the training phase that aims to identify the characteristics of each class, and that is based on a set of training samples. The second phase consists of testing the trained model using a set of samples/users. We have carried out a set of simulations based on the developed model. The results are promising in terms of the accuracy of determining the DoA, taking into consideration a channel with noise and multipath.

Solution of Mixed-Integer Optimization Problems in Bioinformatics with Differential Evolution Method

The solution of the so-called mixed-integer optimization problem is an important challenge for modern life sciences. A wide range of methods has been developed for its solution, including metaheuristics approaches. Here, a modification is proposed of the differential evolution entirely parallel (DEEP) method introduced recently that was successfully applied to mixed-integer optimization problems. The triangulation recombination rule was implemented and the recombination coefficients were included in the evolution process in order to increase the robustness of the optimization. The deduplication step included in the procedure ensures the uniqueness of individual integer-valued parameters in the solution vectors. The developed algorithms were implemented in the DEEP software package and applied to three bioinformatic problems. The application of the method to the optimization of predictors set in the genomic selection model in wheat resulted in dimensionality reduction such that the phenotype can be predicted with acceptable accuracy using a selected subset of SNP markers. The method was also successfully used to optimize the training set of samples for such a genomic selection model. According to the obtained results, the developed algorithm was capable of constructing a non-linear phenomenological regression model of gene expression in developing a Drosophila eye with almost the same average accuracy but significantly less standard deviation than the linear models obtained earlier.

Classification method of marine target motion pattern based on spatial-temporal trajectories

Classifying the motion pattern of marine targets is of important significance to promote target surveillance and management efficiency of marine area and to guarantee sea route safety. This paper proposes a moving target classification algorithm model based on channel extraction-segmentation-LCSCA-lp norm minimization. The algorithm firstly analyzes the entire distribution of channels in specific region, and defines the categories of potential ship motion patterns; on this basis, through secondary segmentation processing method, it obtains several line segment trajectories as training sample sets, to improve the accuracy of classification algorithm; then, it further uses the Leastsquares Cubic Spline Curves Approximation (LCSCA) technology to represent the training sample sets, and builds a motion pattern classification sample dictionary; finally, it uses lp norm minimized sparse representation classification model to realize the classification of motion patterns. The verification experiment based on real spatial-temporal trajectory dataset indicates that, this method can effectively realize the motion pattern classification of marine targets, and shows better time performance and classification accuracy than other representative classification methods.

Using Vector Agents to Implement an Unsupervised Image Classification Algorithm

Unsupervised image classification methods conventionally use the spatial information of pixels to reduce the effect of speckled noise in the classified map. To extract this spatial information, they employ a predefined geometry, i.e., a fixed-size window or segmentation map. However, this coding of geometry lacks the necessary complexity to accurately reflect the spatial connectivity within objects in a scene. Additionally, there is no unique mathematical formula to determine the shape and scale applied to the geometry, being parameters that are usually estimated by expert users. In this paper, a novel geometry-led approach using Vector Agents (VAs) is proposed to address the above drawbacks in unsupervised classification algorithms. Our proposed method has two primary steps: (1) creating reliable training samples and (2) constructing the VA model. In the first step, the method applies the statistical information of a classified image by k-means to select a set of reliable training samples. Then, in the second step, the VAs are trained and constructed to classify the image. The model is tested for classification on three high spatial resolution images. The results show the enhanced capability of the VA model to reduce noise in images that have complex features, e.g., streets, buildings.