Earticle

플라스틱 사출 제품은 다양한 가전제품과 하이테크 제품에 널리 사용되고 있다. 그러나 현재의 치열한 경쟁적 비즈니스 환경에서 플라스틱 사출 제품 제조업자들은 고객을 만족시키면서 경쟁력을 얻기 위하여 다른 경쟁자들보다 먼저 새로운 제품을 시장에 출시하고 신제품의 개발기간을 줄이기 위한 노력을 할 여유가 부족하다. 따라서 무한 경쟁의 시장에서 살아남기 위해서는 제조업자들은 시장 마켓 점유를 빠르게 올리는 것과 동시에 제품의 가격 경쟁력을 가져야 한다. 특징기반 모델의 구조는 현재 연구에서 3D 제작 도구로서 일반적으로 적용되고 있으며 신제품 개발 엔지니어들이 새로운 제품의 개념을 개발하는 데에도 널리 사용되고 있다. 본 연구에서는 특징기반 플라스틱 사출제품을 위한 유전자 알고리즘과 Support Vector Regression (SVR) 기반의 새로운 하이브리드 비용 평가 모델을 제안한다. 제안하는 하이브리드 모델은 기존의 플라스틱 사출제품의 비용평가절차와 계산을 위해 필요로 하는 변수들을 극적으로 간단하게 하고 줄일 수 있다. 사례연구에서는 제안하는 하이브리드 모델과 기존의 multilayer perceptron networks (MLP) 및 pure SVR과의 비교분석을 통하여 제안모델이 플라스틱 사출 제품의 개발단계에서의 비용평가문제를 해결하는데 효율성과 효과성이 있음을 입증한다.

SSD(Solid-State Drive)는 다수의 HDD(Hard Disk Drive)로 구성된 영상 서버에서 캐시로 활용될 수 있으며, 제한된 캐싱 공간을 효율적으로 사용하기 위해서는 캐싱될 영상의 미래 인기도를 정확하게 예측하는 것이 요구된다. 본 논문에서는 SVR(Support Vector Regression) 기반의 영상 인기도 예측 기법을 통해 영상 서버에서 SSD 캐 시를 할당하는 방법을 제안한다. 먼저 예측 영상 인기도를 도출하기 위해 영상의 제목, 대표 이미지, 선호도 지표, 과거 조회수 기록 등 다양한 특징 벡터를 통해 영상 인기도 예측 모델을 구축하고 특징 벡터 구성에 따라 모델의 예 측 성능과 연산 복잡도가 어떻게 변화하는지 분석하며, 그 결과를 기반으로 영상 인기도를 인자로 사용하는 SSD 캐 싱 할당 모델을 제시한다. 9만개의 실제 유튜브 영상의 데이터를 수집하고, SSD 캐싱 동작을 시뮬레이션하여 1) 영상 인기도 예측 성능과 연산 복잡도간의 관계, 2) 최적해 대비 SSD 캐싱 할당 성능 면에서 제안하는 기법의 효용성을 검증하였다.

Solid-state drives (SSDs) can be effectively used as a cache for video servers that consist of a lot of hard disk drives (HDDs), for which exact prediction of future video popularity is essential to make effective use of limited cache space. We propose a new SSD cache allocation scheme by making use of video popularity prediction based on a machine learning technique called support vector regression (SVR). To derive future popularity, we first construct a video popularity prediction model based on various feature types including title, thumbnail, and a ratio of “likes" and “dislikes" and viewing history to predict the popularity of video clips and analyze how complexity and accuracy of the SVR vary with each feature type. Based on this, we develop an SSD cache allocation model that uses video popularity as a parameter. We simulated a video server that stores 90,000 actual YouTube video clips to evaluate our scheme in terms of 1) popularity prediction accuracy and computational complexity 2) SSD caching set determination performance compared with optimal solution. The results confirm that the proposed scheme can accurately predict future popularity at a modest machine learning cost, which can be effectively used to SSD caching set determination.

This research is about a study on the flow stress of Inconel 601 under hot deformation. For Inconel 601, hot compression tests on gleeble 3500 system under 925℃, 1050℃ and 1150℃ and 0.001/s, and 5/s of strain rates were done. The flow behavior of the Inconel 601 was studied and modeled. In this study, the flow stress was modeled using deep neural network and support vector regression algorithm. The flow stress of Inconel 601 was dependent on strain rate and temperature. It was found that both the deep neural network and support vector regression adequately described the flow stress variation of Inconel 601. However, the model by the support vector regression was found to be superior to the model by the deep neural network. The construction of the model by SVR was more efficient than the construction by DNN. Also the prediction accuracy of the model by SVR was better than the accuracy of the model by DNN. It is found that the MAPE(Mean absolute percentage error) of the DNN based model was 4.89% while the MAPE of the SVR based model was 1.98%.

In recent years, data mining techniques such as neural networks, support vector Regression have been applied extensively to the task of predicting financial variables. As influenced by various factors, the volatility of stock shows a non-linear characteristic, which demonstrates that the forecasting is a non-linear problem. Support vector regression (SVR) is proven to be useful in dealing with non-linear forecasting problems in recent years. The key point in using SVR for forecasting is how to determine the appropriate parameters. An improved Artificial Neural Networks(ANN) algorithm is used to optimize the parameter set of (C, σ), which influences the performance of this model directly. By doing so, this model can deal with the nonlinearity and multi-factors of volatility, and ensure stability and accuracy of support vector machine based regression. Finally, we study a case with the satisfactory result by the SPA test which is showing that this model is more accurate than other models, which guarantees its application.

Cloud manufacturing combined with information manufacturing, the Internet of things(IoT), cloud computing, and semantic web technology. Through extending and changing the service technology and network manufacturing, it makes the manufacturing resources and manufacturing capabilities virtualization and servicization. It can make centralized and unified intelligent management for manufacturing. Therefore, the establishment of appropriate cloud manufacturing QoS(quality of service) model is the basis and prerequisite for the development of related research of cloud manufacturing. The existing QoS model ignored the effect of time on QoS. This directly leads to the lack of accuracy of modeling, and further affects the implementation effect of the follow-up study on the prediction, service selection and so on. With cloud manufacturing evaluation criteria, this paper proposes a five dimensional QoS evaluation criteria and the corresponding calculation formula, it is consistent with the cloud manufacturing background. Secondly, we establish the QoS prediction model based on the support vector machine with arbitrary penalty band. This model can cover the shortage for the previous studies, and effectively solve the problem of the dimension of QoS prediction in cloud manufacturing. Thirdly, this paper predicts 5 important QoS indexes. The experimental data from a cloud manufacturing provider's historical data, and the experimental use a large number of historical QoS value to train the support vector machine with arbitrary penalty band. Network training is terminated when the coefficient is greater than 0.995. Then, the QoS value of the seven nodes in the future is predicted by the QoS value of the sample data set. Finally, the experiment shows that the support vector regression machine with arbitrary penalty has a good prediction effect on QoS of cloud manufacturing. But the artificial neural network and the gray forecast model has poor effect on it.

Many comparative studies on the performance of machine learning (ML) techniques for web cost estimation (WCE) have been reported in the literature. However, not much attention have been given to understanding the conceptual differences and similarities that exist in the application of these ML techniques for WCE, which could provide credible guide for upcoming practitioners and researchers in predicting the cost of new web projects. This paper presents a comparative analysis of three prominent machine learning techniques – Case-Based Reasoning (CBR), Support Vector Regression (SVR) and Artificial Neural Network (ANN) – in terms of performance, applicability, and their conceptual differences and similarities for WCE by using data obtained from a public dataset (www.tukutuku.com). Results from experiments show that SVR and ANN provides more accurate predictions of effort, although SVR require fewer parameters to generate good predictions than ANN. CBR was not as accurate, but its good explanation attribute gives it a higher descriptive value. The study also outlined specific characteristics of the 3 ML techniques that could foster or inhibit their adoption for WCE.

Finance fraud of companies is an international difficult problem with a long history. The finance fraud problem is concerned by lots of people. Some researchers make a lot of qualitative or quantitative researches and get some valuable conclusions. In this article , we mainly applies empirical research method, combined with normative research method. First of all, this paper reviews the relevant literatures of financial fraud detecting of listed companies, expounds existing research results from the aspects of motives, signs and detecting methods. We appraise these results are ordering to national conditions and characteristics, analyze the definition of financial fraud. We established a new method which is partial least squares (PLS) and support vector regression (SVR) to solve the above problem in finance. The PLS are able to reduce dimension effectively, acquire nonlinear factor matrix, and SVR has many advantages, such as high imitation degree, effective classification and strong robustness. The model which combines PLS and SVR has great recognition effect.

The observation and forecasting of vessel traffic flow is the foundmental of design for ships’ routeing system. An integrated Genetic Algorithm (GA) based Support Vector Machine (SVM) model for vessel traffic flow forecasting with input factors selection procession is presented in this paper. GA based SVM forecasting model is established whose parameters were optimized through genetic algorithms. Finally, the prediction model is used for ningbo-zhoushan port and the prediction result shows that the improved model reflects the actual growth of vessel traffic flow trend more reasonable and effectively.

In the interest of deriving regressor that is robust to outliers, we propose a support vector regression (SVR) based on non-convex quadratic insensitive loss function with flexible coefficient and margin. The proposed loss function can be approximated by a difference of convex functions (DC). The resultant optimization is a DC program. We employ Newton’s method to solve it. The proposed model can explicitly enhance the robustness and sparseness of SVR. Numerical experiments on six benchmark data sets show that it yields promising results.

Due to the normal forecasting methods for subgrade settlement using observation data have different applicabilities, and the predicting results has bigger volatility and lower accuracy. In view of the above problems, based on the twin support vector regression tool, the settlement prediction model is established by combining with the measured roadbed settlement data; The related parameters of the prediction model are given and compared with the standard support vector regression machine, the comparison tests show that the twins support vector regression is a new method to predict the settlement of the roadbed, and is superior in forecasting accuracy to the standard support vector regression.

Predicting fire nature is artistry as much as it’s a science. Forecasting the burnt area and range of field plays a vital role in resource abatement and renewal efforts. Literature studies have shown that machine learning techniques achieved better performance in forecasting and trend perusal. The purpose of this paper is to investigate the relevance of the state-of-the-art machine learning techniques epsilon Support Vector Regression and Nu-SVR to predict forest fire occurrence and burned area utilizing the meteorological data. The goals of this research are to (1) Identifying the best parameter settings using a grid-search and pattern search technique; (2) comparing the prediction accuracy among the models using different data sorting methods, random sampling and cross-validation. In conclusion, the experiments show that E-SVR performs better using various fitness-functions and variance analysis. The study is carried out to build predictive models for guesstimating the risk of the outbreaks in Montesinho Natural Park.

Significant color discrepancies between the different camera views can be observed in multi-view video sequences. In this paper, a color adjustment algorithm based on support vector regression is proposed. A mapping function is established by extracted feature points from original-image and target-image. Then the mapping function is applied to the original image to obtain the corrected image. Experimental results show that the proposed method can produce good correction result. It also shows that the color differences between multi-view video can be effectively reduced by SVR.

In this article, aiming at complex prediction problems of the rail transit passenger flow and prediction problems combined with the actual situation of rail transit in Beijing. We will propose a fusion model of passenger flow for predicting. It can improve the prediction accuracy for passenger flow forecasting. We use the partial least squares regression method to solve multicollinearity between the dependent variable. The method of principal component analysis can rescreen the all the factors which are affect the passenger flow. To extract comprehensive variable this has the best ability to explain the passenger flow from all of the information, in order to solve the relevance of statistics, noise and information redundancy. The nonlinear prediction model which is between comprehensive variable and passenger flow will be established. Finally, through the passenger flow forecasting of exchange station of Beijing Metro Line 1 to verify the effectiveness of the method.

It has been found that poor quality images decrease the performance of finger vein recognition system, due to missing, vague or spurious features. Therefore, it is important for a finger vein recognition system to evaluate the quality of finger vein images. In this paper, a new method based on Support Vector Regression (SVR) is proposed for finger vein image quality evaluation. In our method, we first manually annotate quality scores for finger vein images in training set and extract five quality features of these images. Then quality scores and quality features are used to build a SVR model, which will be applied to evaluate quality for testing images. In addition, we explore the use of quality score and ascertain that quality score can be used as ancillary information to enhance recognition accuracy for finger vein. Experimental results show that our proposed method is effective for finger vein image quality evaluation.

In urban management, it is important to precisely forecast the short-term demand for necessary resources, including water, electric power, and gas. Although a variety of prediction models have been proposed in literature, the underlying defects and limitations confine the effectiveness and forecasting precision of these models. In this paper, the short-term prediction problem is modeled as a non-linear multivariate regression problem, which is solved by support vector regression (SVR). The parameters in SVR are optimized by artificial fish-swarm algorithm (AFSA). The proposed prediction model (termed SVR-AFSA) overcomes the defects of existing prediction models, thus promoting forecasting precision. In order to verify the effectiveness and prediction precision of SVR-AFSA, this paper conducts experiments on a real dataset of two-month hourly water consumption. It also compares SVR-AFSA with two commonly adopted models, i.e., traditional BP neural network, and SVR optimized by grid method (SVR-grid). The experiments results show that SVR-AFSA outperforms these two models in prediction precision in terms of mean squared error (MSE) and mean absolute percentage error (MAPE).

Financial time series prediction is regarded as one of the most challengingtasks due to the inherent noise and non-stationality of the data. This paperproposed a two-stage financial time series prediction approach hybridizingsupport vector regression (SVR) with hierarchical clustering (HC). By averaging the variables within the clusters obtained from hierarchical clustering, we define super predictors and use them as the input variables of the SVRforecasting model. Although averaging is a simple technique, it plays animportant role in reducing variance. To evaluate the performance of theproposed approach, the Shanghai-Shenzhen 300 index is used as illustrativeexample. The experimental results show that the proposed approach outperforms both the SVR model with principal component analysis and the SVRmodel with genetic algorithms in average prediction error and predictionaccuracy.

As embedded devices prevail in daily life, high energy consumption caused by embedded software caught academic attentions. Multifarious testing and predicting methods are developed accordingly. This paper proposes a model about energy consumption of embedded device based on analysis of embedded software structure and support vector machine regression. The nonlinear relationship between energy consumption and software structure is revealed. The research finds software structure is determined by features like number of components, complexity of component interface, component coupling, and path length. These features are qualified and modeled by using support vector machine regression and energy consumption is predicted based on this model. The experiments results confirm the proposed model.