Earticle

One of the latest emerging technologies is artificial intelligence, which makes the machine mimic human behaviour. The most important component used to detect cyber attacks or malicious activities is the intrusion detection system (IDS). Artificial intelligence plays a vital role in detecting intrusions and widely considered as the better way in adapting and building IDS. In modern days, neural network algorithms are emerging as a new artificial intelligence technique that can be applied to real-time problems. The proposed system is to detect a classification of botnet attack which poses a serious threat to financial sectors and banking services. The proposed system is created by applying artificial intelligence on a realistic cyber defence dataset (CSE-CIC-IDS2018), the latest IDS Dataset in 2018 by Canadian Institute for Cybersecurity (CIC) on AWS (Amazon Web Services). The proposed system of Artificial Neural Networks provides an outstanding performance of Accuracy score is 99.97% and an average area under ROC(Receiver Operator Characteristic) curve is 0.999 and an average False Positive rate is a mere value of 0.03. The proposed system of Artificial Intelligence-based Intrusion detection of botnet attack classification is powerful, more accurate and precise. The novel proposed system can be applied to conventional network traffic analysis, cyber-physical system traffic analysis and also can be applied to the real-time network traffic data analysis.

Database systems usually have many parameters that must be configured by database administrators and users. RocksDB achieves fast data writing performance using a log-structured merged tree. This database has many parameters associated with write and space amplifications. Write amplification degrades the database performance, and space amplification leads to an increased storage space owing to the storage of unwanted data. Previously, it was proven that significant performance improvements can be achieved by tuning the database parameters. However, tuning the multiple parameters of a database is a laborious task owing to the large number of potential configuration combinations. To address this problem, we selected the important parameters that affect the performance of RocksDB using random forest. We then analyzed the effects of the selected parameters on write and space amplifications using analysis of variance. We used a genetic algorithm to obtain optimized values of the major parameters. The experimental results indicate an insignificant reduction (-5.64%) in the execution time when using these optimized values; however, write amplification, space amplification, and data processing rates improved considerably by 20.65%, 54.50%, and 89.68%, respectively, as compared to the performance when using the default settings.

Structural failures in steel connections under extreme loading conditions, such as earthquakes or unexpected impact loads, can lead to catastrophic progressive collapse. Among various connection types, the T-stub connection is a fundamental component whose failure mode is highly sensitive to the geometric configuration of its bolts. This study proposes an automated parameter optimization framework for bolt layouts in H-beam T-stub connections to mitigate prying action and enhance structural resilience against disasters. To achieve a high-fidelity analysis, a nonlinear finite element (FE) model was developed using Abaqus. The model incorporates an H-400x200x8x13 section and M20 (F10T) high-strength bolts, accounting for complex surface-to-surface contact, material plasticity, and large deformation effects. To transition from traditional trial-and-error design to a datadriven safety management approach, the FE model was integrated with Isight to automate the optimization of critical design variables: the gauge distance (g) and edge distance (e). The optimization process employed the Optimal Latin Hypercube design of experiments (DOE) to sample the design space efficiently. A surrogate model was subsequently constructed using Response Surface Methodology (RSM) to analyze the sensitivity of the structural response to bolt positioning. The results demonstrate that optimizing the bolt coordinates can significantly reduce the prying force, thereby preventing premature bolt fracture and increasing the energy dissipation capacity of the connection. From a disaster information perspective, this research provides a quantitative foundation for Digital Twin-based structural health monitoring. By establishing a relationship between geometric parameters and failure limits, the proposed methodology enables more accurate safety assessments of existing infrastructure. This study contributes to the development of intelligent disaster prevention systems by integrating advanced numerical simulation with automated optimization techniques to ensure the structural integrity of steel-framed buildings.

Taguchi's robust parameter design is an approach to reduce the performance variation of quality characteristics in products and processes. In robust design, the signal-to-noise ratio (SN ratio) was used to find the optimum condition to minimize the variation of quality characteristics as much as possible and bring the average of quality characteristics closer to the target value. In this paper, we propose a simultaneous optimization method based on a linear model of the SN ratio as a method to find the optimal condition of the control factor in case of multi-characteristics. In addition, the proposed method and the existing method were compared and studied by taking actual cases.

Taguchi has used the signal-to-noise ratio (SN) to achieve the appropriate set of operating conditions where variability around target is low in the Taguchi parameter design. Taguchi has dealt with having constraints on both the mean and variability of a characteristic (the dual response problem) by combining information on both mean and variability into an SN. Many Statisticians criticize the Taguchi techniques of analysis, particularly those based on the SN. In this paper we propose a substantially simpler optimization procedure for parameter design to solve the dual response problems without resorting to SN.

수공구조물을 설계하거나 수자원계획을 수립할 때 제한된 수문자료로 인해 수문모형의 매개변수를 추정하는데 어려움이 따르며 추정된 결과에 신뢰성을 부여하기 위해서 필수적으로 불확실성 분석이 필요하다 하겠다. 이러한 관점에서 본 연구에서는 국내외에서 주로 이용되고 있는 NWS-PC 강우-유출 모형을 대상으로 보다 진보된 매개변수 추정과 불확실성 분석이 가능한 Bayesian Markov Chain Monte Carlo 기법과 결합하여 국내 소양강댐 유역 일유입량 모의에 적용하였다. 실측 일유입량 자료를 대상으로 모형의 검정과정을 수행하였으며 NWS-PC 모형의 총 13개의 매개변수에 대한 사후분포를 추정하여 유출수문곡선의 불확실성 구간을 추정하였다. 검정 및 검증 모두에서 Bayesian Markov Chain Monte Carlo 기법이 모형의 적합성 측면에서 기존 방법론과 비교해보면 다소 우수하거나 비슷한 결과를 나타내었다. 실제로 유역에 발생하는 유출은 다양한 요인에 따라 변화될 수 있으며 이러한 점에서 Bayesian 방법은 강우-유출 관계에서 발생하는 이러한 불확실성을 매개변수의 불확실성으로 인지함으로서 우리가 예상치 못한 유출 사상에 대한 형태를 고려할 수 있는 장점이 있다. 따라서 댐 설계와 같은 대규모 수공 구조물 설계 시에 이러한 불확실성이 접목된 강우-유출 분석이 이루어진다면 보다 합리적인 방법으로 홍수 위험도 분석이 가능하며 더욱이 댐 규모 결정에 있어서 신뢰성 있는 의사 결정 수단을 제공할 수 있을 것으로 사료된다.

It is not always easy to estimate the parameters in hydrologic models due to insufficient hydrologic data when hydraulic structures are designed or water resources plan are established. Therefore, uncertainty analysis are inevitably needed to examine reliability for the estimated results. With regard to this point, this study applies a Bayesian Markov Chain Monte Carlo scheme to the NWS-PC rainfall-runoff model that has been widely used, and a case study is performed in Soyang Dam watershed in Korea. The NWS-PC model is calibrated against observed daily runoff, and thirteen parameters in the model are optimized as well as posterior distributions associated with each parameter are derived. The Bayesian Markov Chain Monte Carlo shows a improved result in terms of statistical performance measures and graphical examination. The patterns of runoff can be influenced by various factors and the Bayesian approaches are capable of translating the uncertainties into parameter uncertainties. One could provide against an unexpected runoff event by utilizing information driven by Bayesian methods. Therefore, the rainfall-runoff analysis coupled with the uncertainty analysis can give us an insight in evaluating flood risk and dam size in a reasonable way.

개별 강우-유출 사상을 대상으로 최적 매개변수를 산정하는 경우 사상별로 매개변수가 서로 다르며 물리적인 범위를 중심으로 변동성이 커 유역의 대표 매개변수를 결정하는데 어려움이 있다. 매개변수 추정 시 변동성 증가는 강우의 시공간적 변동성과 함께, 유역 내 일부 홍수량 산정지점 기준으로 강우-유출 자료만이 이용 가능하여 매개변수의 식별성(identifiability)이 매우 낮다. 추정되는 매개변수의 변동성 확대에 따른 문제점을 개선하기 위하여, 본 연구에서는 다수의 사상들을 동시에 고려한 매개변수 최적화 방법을 제안하였으며, NSE를 목적함수로 하여 매개변수를 최적화하였다. 개별 사상들을 통합적으로 고려하여 최적매개변수를 산정하는 경우 매개변수의 물리적인 특성을 유지함과 동시에 유역의 공동 매개변수 효율적으로 추정이 가능하였다. 개별 매개변수와 공통 매개변수 NSE의 차이가 최대 0.08 정도를 나타내며, 홍수량 재현 측면에서도 개별적으로 최적화를 수행한 경우와 유사하거나 보다 개선된 결과를 확인할 수 있었다.

Estimation of the parameters for individual rainfall-rainfall events can lead to a different set of parameters for each event which result in lack of parameter identifiability. Moreover, it becomes even more ambiguous to determine a representative set of parameters for the watershed due to enhanced variability exceeding the range of model parameters. To reduce the variability of estimated parameters, this study proposed a parameter optimization framework with the simultaneous use of multiple rainfall-runoff events based on NSE as an objective function. It was found that the proposed optimization framework could effectively estimate the representative set of parameters pertained to their physical range over the entire watershed. It is found that the difference in NSE value of optimization when it performed individual and multiple rainfall events, is 0.08. Furthermore, In terms of estimating the observed floods, the representative parameters showed a more improved (or similar) performance compared to the results obtained from the single-event optimization process on an NSE basis.

최근 기후변화, 기후변동으로 인해 저수지 및 소하천의 유량이 감소하는 추세이다. 이와 같은 하천 유량 감소는 건천화 현상을 발생시켜 하천의 생태계 파괴, 자정기능 상실, 친수 기능 상실 등 각종 부작용의 원인이 될 수 있다. 가뭄으로 인해 하천 및 저수지에 저유량이 유지되는 유역은 유출량의 특성 파악이 중요한 요소로 작용한다. 강우-유출 모형을 활용한 유출량 모의를 통해 하천의 유량을 파악할 수 있으며, 그 중 M-RAT (Monthly Runoff Assessment Tool) 모형은 일반적이고 간단한 월 물수지 방정식을 기반으로 유출 분석을 통해 월 유출량을 추정한다. M-RAT 모형을 활용한 유출량 모의 시 하천 유역의 수문 및 기상자료를 사용하여 모형의 매개변수를 보정하는 과정을 통해 모의의 정확도를 높일 수 있다. 본 연구는 머신러닝 기법을 활용하여 13개 유역을 대상으로 M-RAT 모형의 매개변수를 검보정 및 최적화하고, 획득한 매개변수를 활용하여 M-RAT 모형을 통한 월 유출량을 모의하였다. 모형 성능 검증을 위해 미호강, 합천댐, 홍천강 3개 유역에 적용하여 정확성을 검토한 결과, 머신러닝 기반의 매개변수 최적화 기법을 활용했을 때 기존에 수행된 기법에 비해 개선된 결과를 확인할 수 있었다.

Recently, the flow rate has been decreased from the reservoirs and small streams due to climate change. The reduction of stream flow cause the stream desiccation and lead to the various adverse effects such as ecosystem degradation and loss of self-purification capacity. In watersheds where have low flow due to drought, considering the runoff characteristics is important. The runoff of stream can be estimated and predicted by the simulations using rainfall-runoff model such as M-RAT (Monthly Runoff Assessment Tool) model is a widely-used tool. Based on a simple monthly water-budget equation, M-RAT estimates monthly runoff through the analysis. With the M-RAT model, the accuracy of the runoff simulations show the improved results by calibrating the model parameters using hydrological and meteorological data from the watershed. This study proposed the parameter optimization method using the machine learning models to calibrate and optimize the M-RAT model parameters for 13 watersheds, and simulated monthly runoff using the optimized parameters. The proposed methods showed the improved accuracy compared to the traditional parameter optimization methods.

본 논문에서는 4세대 이동통신 기술표준인 LTE/LTE-A 시스템에서 기존 실외 기지국인 eNB 에 집중되는 이동통신 트래픽 부하를 경감시키고 실내 사용자에게 만족스런 통신품질을 제공하여 더 많은 가입자를 수용하기 위해 적용되고 있는 초소형 실내 기지국인 펨토셀(FemtoCell)에서 발생할 수 있는 다양한 유형의 핸드오버 문제의 최적화 방법을 제시하였다. LTE/LTE-A에서는 펨토셀의 최적 설치와 운용을 위해 자가구성네트워크(SON) 기술이 제안되고 있으며, 특히 실내 사용자의 펨토셀간 이동시 발생할 수 있는 링크절체(RLF)에 의한 핸드오버 문제의 최소화를 위해 핸드오버 파라미터를 최적화하는 MRO 기능이 적용된다. 펨토셀간의 핸드오버에서는 너무 이른 핸드오버, 너무 늦은 핸드 오버 그리고 잘못된 셀로의 핸드오버의 세 가지 유형의 핸드오버 문제가 발생한다. 본 논문에서는 이 러한 세 가지 유형의 핸드오버에서 발생하는 RLF에 의한 핸드오버 문제를 해결하기 위한 효과적인 핸드오버 파라미터 최적화 방안을 제시하고 분석하였다. 본 논문에서 제안하는 MRO 기법은 단말기 에서 측정된 펨토셀들의 수신신호 강도를 기반으로 서빙셀과 타겟셀의 CIO 값을 적응적으로 조정할 수 있는 최적화 기법이다. 다섯 개의 펨토셀로 구성된 모의실험환경에 대한 성능분석 연구결과 본 논 문에서 제안한 MRO 기법은 이 알고리즘을 적용하지 않은 환경에 비해 약 80% 정도의 성능개선 효 과를 제공한다. 따라서 본 논문에서 제안한 펨토셀의 핸드오버 문제 해결과 최적화를 위한 MRO 기 법은 실내 사용자를 위한 LTE-A시스템의 펨토셀 환경에서 효과적인 RLF 방지와 개선을 위해 적용 될 수 있을 것으로 기대된다.

The Self-Organizing Network(SON) technology has been developed in 3GPP Release 10 and 11 standard, so called the LTE-Advanced system, for automatic configuration, operation and maintenance of indoor and outdoor base-stations. In particular, the Femtocell, also called a HeNB(Home eNB), is tending to be deployed not only to accommodate indoor mobile communication users, but also to diverge heavy traffic load concentrated to the outdoor base-station(eNB). The Femtocell is connected to the indoor wired communication networks such as Internet. In such LTE/LTE-A Femtocell systems, three types of handover problems such as too late handover, too early handover and to wrong cell handover problems may occur due to inappropriate handover parameters configuration and its resulting radio link failure(RLF). In this paper, a Mobility Robustness Optimization(MRO) algorithm is proposed to tackle the three handover problems in the LTE-A Femtocell systems. Simulation results show that the proposed MRO algorithm is capable to improve such handover problems up to 80% by optimizing handover parameters automatically. The proposed MRO algorithm will play an important role in providing better communication services to the users in the LTE-A indoor environment where many Femtocells are deployed.

비트코인은 정부나 금융기관에 의존되어 있지 않은 전자 거래를 지향하며 만들어진 peer-to-peer 방식의 암호화폐이다. 비트코인은 최초 발행 이후 거대한 블록체인 금융 시장을 생성했고, 이에 따라 기계 학습을 이용한 비트코인 가격 데이터를 예측하는 연구들이 활발해졌다. 그러나 기계 학습 연구의 비효율적인 Hyper-parameter 최적화 과정이 연구 진행에 있어 비용적인 측면을 악화시키고 있다. 본 논문은 LSTM(Long Short-Term Memory) 층을 사용하는 비트코인 가격 예측 모델에서 가장 대표적인 Hyper-parameter 중 Timesteps, LSTM 유닛의 수, 그리고 Dropout 비율의 전체 조합을 구성하고 각각의 조합에 대한 예측 성능을 측정하는 실험을 통해 정확한 비트코인 가격 예측을 위한 Hyper-parameter 최적화의 방향성을 분석하고 제시한다.

Bitcoin is a peer-to-peer cryptocurrency designed for electronic transactions that do not depend on the government or financial institutions. Since Bitcoin was first issued, a huge blockchain financial market has been created, and as a result, research to predict Bitcoin price data using machine learning has been increasing. However, the inefficient Hyper-parameter optimization process of machine learning research is interrupting the progress of the research. In this paper, we analyzes and presents the direction of Hyper-parameter optimization through experiments that compose the entire combination of the Timesteps, the number of LSTM units, and the Dropout ratio among the most representative Hyper-parameter and measure the predictive performance for each combination based on Bitcoin price prediction model using LSTM layer.

본 논문에서는 타겟의 RCS Calibration의 정확도를 향상시키기 위해 Time-Gating 조건에 대한 최적화 기법을 제안하였다. Matlab 기반의 코드를 이용하여 RCS Calibration의 최적화 코드를 개발하였으며, 그 조건을 찾기 위해 S-Parameter 신호 처리 알고리즘을 제시하였다. 안테나와 Far-field 조건을 만족한 거리에 있는 타겟 을 두고 측정실험을 진행하였다. 제안하는 신호 최적화 방법의 핵심은 실험에서 측정된 시간 도메인 S-Parameter 값을 확인하여 폭과 레퍼런스 포인트를 조절하는 것이다. 그 방법의 적용 과정을 본 논문에서 서술하였으며, 시간 축에서의 S21의 신호를 확인하여 타겟의 위치를 기준으로 한 Time-Gating 최적화를 진행하였다. 결과적으로 게 이팅 최적화 코드를 적용하였을 때 최적화가 진행되지 않았을 경우와 비교하여 약 12배만큼 정의된 에러 값이 감 소함을 확인할 수 있었다.

In this paper, an optimization technique for time-gating conditions is proposed to improve the accuracy of target RCS calibration. An RCS calibration optimization code was developed using Matlab-based code, and an S-Parameter signal processing algorithm was presented to find the condition. The measurement experiment was conducted with the antenna and the target at a distance that satisfies the far-field condition. The signal optimization method is to check time domain S-parameter value measured in the experiment and adjust the width and reference point. The process of applying the method was described in this paper, and time-gating optimization was performed based on the position of the target by checking the signal of S21 on the time axis. As a result, when the gating optimization code was applied, it was confirmed that the defined error value was reduced by about 12 times compared to the case where optimization was not performed.

LiDAR는 자율 주행뿐만 아니라 다양한 산업 현장에 적용되어 대상의 크기와 거리를 측정 하는 데 사용되고 있다. 이에 더하여 이 센서는 반사된 빛의 양을 바탕으로 반사 강도 영상 또한 제공한다. 이는 측정 대상의 형상에 대한 정보를 제공하여 센서 데이터 처리에 긍정적인 효과를 일으킨다. LiDAR는 고해상도가 될수록 높은 성능을 보장하지만 이는 센서 비용의 증 가를 야기하는데, 이 점은 반사 강도 영상에도 해당된다. 높은 해상도의 반사 강도 영상을 취득 하기 위해서는 고가의 장비 사용이 필수적이다. 따라서 본 연구에서는 저해상도의 반사 강도 영상을 고해상도의 영상으로 개선하는 인공지능을 개발하였다. 이를 위해서 본 연구에서는 최 적의 초해상화 신경망 모델을 위한 파라미터 분석을 수행하였다. 또한, 초해상화 알고리즘을 2,500여 장의 반사 강도 영상에 적용하여 훈련과 검증을 하였다. 결과적으로 반사 강도 영상의 해상도를 향상시켰다. 바라건대 본 연구의 결과가 향후 자율 주행 분야에 적용되어 주행환경 인식과 장애물 탐지 성능 향상에 기여할 수 있기를 기대하는 바이다.

LiDAR is used in autonomous driving and various industrial fields to measure the size and distance of an object. In addition, the sensor also provides intensity images based on the amount of reflected light. This has a positive effect on sensor data processing by providing information on the shape of the object. LiDAR guarantees higher performance as the resolution increases but at an increased cost. These conditions also apply to LiDAR intensity images. Expensive equipment is essential to acquire high-resolution LiDAR intensity images. This study developed artificial intelligence to improve low-resolution LiDAR intensity images into high-resolution ones. Therefore, this study performed parameter analysis for the optimal super-resolution neural network model. The super-resolution algorithm was trained and verified using 2,500 LiDAR intensity images. As a result, the resolution of the intensity images were improved. These results can be applied to the autonomous driving field and help improve driving environment recognition and obstacle detection performance

In this paper, a study was conducted on the methodology for learning the parameters of a neural network using an evolutionary algorithm such as a particle swarm optimization algorithm. The possibility of using the particle swarm optimization algorithm for deep learning was analyzed, and various methods were considered for practical use.

The SSGAMLP(Small Set Genetic Algorithm Multilayer Perceptron) model helps individual evolution by group evolution. With respect to the MLP, it has better generalization, it can get unknown feature expressions of more possibilities. The model still exist many problems need to be solved. The number of nodes in the hidden layers and the population size of MLP has a great influence on the performance of SSGAMLP. So this paper focuses on the optimization of that two parameters on SSGAMLP. In this paper, the models of several different experiments are designed. By comparing the experimental data, the relationship between the parameter selection and the model performance is obtained.

필드 프로그래머블 포토닉 게이트 어레이(FPPGA)는 광 회로를 사용자의 목적에 맞게 동적으로 재구성할 수 있어, 통신 네트워크, 양자 컴퓨팅, RF 신호 처리 등 다양한 응용 분야에서 활용될 수 있다. 그러나 이러한 다양한 응용 에 적용 가능한 프로그래머블 광 회로를 구현하기 위해서는 입출력 단자와 게이트의 수가 증가하게 되고, 그에 따라 제어 파라미터의 수도 늘어나게 되다. 그리고, 이와 같은 복잡한 광학 시스템에서 원하는 기능을 구현하기 위한 최적의 파라미터를 찾는 것은 어려운 과제이다. 본 연구에서는 MEMS 기반으로 구현 된 FPPGA 를 효과적으로 활용하기 위하 여, 해당 구조를 기반으로 역전파 기법을 적용한 파라미터 최적화 시뮬레이션 도구를 개발하였다. 이 시뮬레이션 도구는 제작된 FPPGA의 게이트 구조와 1:1로 매핑되어, 사용자의 목적에 맞게 광 회로의 파라미터를 효율적으로 설정할 수 있다. 이를 통해 광 신호 라우팅, 주파수 분할 다중화, 신경망 기반 숫자 인식 등 다양한 응용에서 효과적인 파라미터 최적화가 가능함을 입증하였다. 본 연구를 통해 우리는 FPPGA의 실용성을 높이고, 전문 지식이 없는 사용자도 효과적 으로 활용할 수 있는 FPPGA용 인터프리터 제작의 기반을 마련하였다.

Field Programmable Photonic Gate Arrays (FPPGA) can be dynamically reconfigured to meet user objectives, enabling their application in various fields such as communication networks, quantum computing, and RF signal processing. However, implementing programmable optical circuits suitable for such diverse applications requires increasing the number of input/output terminals and gates, which consequently increases the number of control parameters. As a result, finding the optimal parameters to realize desired functionalities in these complex optical systems becomes a challenging task. In this paper, to utilize MEMS-based FPPGA more effectively, we developed a parameter optimization simulation tool that applies backpropagation techniques to their architecture. This simulation tool directly maps onto the gate structure of the fabricated FPPGA, allowing efficient configuration of optical circuit parameters in alignment with user objectives. We demonstrated that effective parameter optimization is achievable in various applications, including optical signal routing, frequency division multiplexing, and neural network-based digit recognition. Our research enhances the practicality of FPPGAs and lays the groundwork for developing an FPPGA interpreter that can be effectively used by users without specialized knowledge.

Accelerator has been widely applied to high energy, low energy physics, Medical environmental and military fields, etc. The magnet power supply is one of the most important parts of the accelerator. While the magnet power supply of accelerator is taken as the control object, and the design of LQR controller of the magnet power supply based on PSO has been proposed. At the beginning, the state-space model of the magnet power supply has been established by modern control theories. Then, the PSO (Particle Swarm Optimization) algorithm has been used for the weight matrices optimization of LQR controller in order to ensure that the magnet power supply can provide a magnetic field quickly which the accelerator required. The experimental results indicate that the method that proposed in the paper can meet the requirements of fast output response of the system and each control index of the LQR controller is obviously superior to that by the traditional method.

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).

The topics on human mobility model have long been researched by various academic and industrial elds. It also has been proven that human mobility has specic patterns and can be predicted up to the probability of 93%, since the mobility of a person cannot be random while peoples have their own frequent visiting places. As a basis of human mobility research, sets of positioning data is used widely. The positioning data of a human can be obtained by GPS or other similar positioning systems, however, it contains inherited environmental errors. It is clear that such position errors harm the correctness of human mobility re- lated results. In this paper we will present ltering method of erroneous positioning data of human mobility with the use of One-Class Support Vector Machine (OCSVM), and we adapted Radial Basis Function (RBF) as kernel function. Experimental values of the criti- cal parameter for RBF have been found for optimal ltering. With this optimal parameter, we ltered raw data set of human mobility trail in order to obtain accurate position data set for further research purpose. By iteratively applying the OCSVM based ltering, like hill climbing approach, we prove that researchers can lter arbitrary rate of raw data for their own purpose. With four sets of positioning data set from various sources, we demonstrate the usefulness of our ltering approach.

The parameters of support vector machine have a great influence on the learning ability and generalization ability, so an improved ant colony optimization algorithm is proposed to optimize the parameters of SVM, then an optimized SVM classifier (IMACO-SVM) is proposed for data classification. In the IMACO-SVM, the adaptive adjustment pheromone strategy is used to make relatively uniform pheromone distribution and the improved pheromone updating method is used to submerge the heuristic factor by the residual pheromone information, in order to effectively solve the contradiction between expanding search and finding optimal solution. The selection of parameters of the SVM is regarded as a combination optimization of parameters in order to establish the objective function of combination optimization. The improved ACO algorithm with good robustness and positive feedback characteristics and parallel searching is used to search for the optimal value of objective function. In order to validate the classification effectiveness of the IMACO-SVM algorithm, some experimental data from the UCI machine learning database are selected in this paper. The classification results show that the proposed IMACO-SVM algorithm has higher classification ability and classification accuracy.