• International journal of advanced smart convergence
• /
• 제12권4호
• /
• pp.88-97
• /
• 2023

Traffic flow prediction is of great significance in urban planning and traffic management. As the complexity of urban traffic increases, existing prediction methods still face challenges, especially for the fusion of spatiotemporal information and the capture of long-term dependencies. This study aims to use the fusion model of graph neural network to solve the spatio-temporal information fusion problem in traffic flow prediction. We propose a new deep learning model Spatio-Temporal Information Fusion using Graph Neural Networks (STFGNN). We use GCN module, TCN module and LSTM module alternately to carry out spatiotemporal information fusion. GCN and multi-core TCN capture the temporal and spatial dependencies of traffic flow respectively, and LSTM connects multiple fusion modules to carry out spatiotemporal information fusion. In the experimental evaluation of real traffic flow data, STFGNN showed better performance than other models.

• 지능정보연구
• /
• 제29권3호
• /
• pp.419-437
• /
• 2023

국내 씬파일러(Thin Filer)의 수가 1200만명을 넘어서며, 금융 업계에서 씬파일러의 신용을 정확히 평가하여 우량고객을 선별해 대출을 공급하는 시도가 많아지고 있다. 특히, 차주의 신용정보에 존재하는 비선형성을 반영하여 채무불이행을 예측하기 위해서 다양한 머신러닝 알고리즘을 활용한 연구가 진행되고 있다. 그 중 그래프 신경망 구조(Graph Neural Network)는 일반적인 신용정보 외에 대출자 간의 네트워크 정보를 반영할 수 있다는 점에서 데이터가 부족한 씬파일러의 채무 불이행 예측에서 주목할 만하다. 그러나, 그래프 신경망을 활용한 기존의 연구들은 신용정보에 존재하는 다양한 범주형 변수를 적절히 처리하지 못했다는 한계가 있었다. 이에 본 연구는 범주형 변수의 맥락적 정보를 추출할 수 있는 트랜스포머 메커니즘(Transformer mechanism)과 대출자 간 네트워크 정보를 반영할 수 있는 그래프 합성곱 신경망(Graph Convolutional Network)를 결합하여 효과적으로 씬파일러의 채무 불이행 예측이 가능한 TeGCN (Transformer embedded Graph Convolutional Network)를 제안한다. TeGCN는 일반 대출자 데이터셋과 씬파일러 데이터셋에 대하여 모두 베이스 라인 모델 대비 높은 성능을 보였으며, 특히 씬파일러 채무 불이행 예측에 우수한 성능을 달성했다. 본 연구는 범주형 변수가 많은 신용정보와 데이터가 부족한 씬파일러의 특성에 적합한 모델 구조를 결합하여 높은 채무 불이행 예측 성능을 달성했다는 시사점이 있다. 이는 씬파일러의 금융소외문제를 해결하고 금융업계에서 씬파일러를 대상으로 추가적인 수익을 창출하는데 기여할 수 있을 것이다.

• 한국콘텐츠학회논문지
• /
• 제9권2호
• /
• pp.142-151
• /
• 2009

기존의 문서 군집화 기법 NSTC은 문서 군집화 과정 내에서 TF-IDF를 이용하여 문서간 유사도를 측정한다. 본 논문에서는 TF-IDF가 아닌, 공통 Phrase의 관계 그래프를 이용한 새로운 문서간 유사도 측정을 제안한다. 이 방법은 문서 집합 내의 공통 Phrase들의 관계를 나타낸 관계 그래프를 통해 공통 Phrase의 가중치를 부여하는 방법을 제시한다. 또한 실험을 통해 NSTC와 비교하여 본 논문에서 제안한 문서간 유사도 측정 기법이 문서 군집화에 더욱 효과적임을 보였다.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2001년도 ICCAS
• /
• pp.120.3-120
• /
• 2001

We propose a new method of path planning for cleaning robots. Path planning problem for cleaning robots is different from conventional path planning problems in which finding a collision-free trajectory from a start point to a goal point is focused. In the case of cleaning robots, however, a planned path should cover all area to be cleaned. To resolve this problem in a systematic way, we propose a method based on a graph model as follows: at first, partition a given map into proper regions, then transform a divided region to a vertex and a connectivity between regions to an edge of a graph. Finally, a region is divided into sub-regions so that the graph has a unary tree which is the simplest Hamilton path. The effectiveness of the proposed method is shown by computer simulation results.

• Journal of Communications and Networks
• /
• 제15권3호
• /
• pp.321-328
• /
• 2013

In this paper, we present a technique for obtaining conflict-free schedules for real-time automation protocol for industrial Ethernet (RAPIEnet) switches. Mathematical model of the switch is obtained using graph theory. Initially network traffic entry and exit parts in a single RAPIEnet switch are identified, so that a bipartite conflict graph can be constructed. The obtained conflict graph is transformed to three kinds of matrices to be used as inputs for our simulation model, and selection of any of the matrix forms is application-specific. A greedy edge-coloring algorithm is used to schedule the network traffic and to solve the minimum coloring problem. After scheduling, empty slots are identified for forwarding the non real-time traffic of asynchronous devices. Finally, an algorithm for synchronizing the schedules of adjacent switches is proposed using edge-contraction and minors. All simulations were carried out using Matlab.

• 대한안전경영과학회지
• /
• 제13권2호
• /
• pp.231-241
• /
• 2011

Business processes are often of long duration, and include internal worker's decision making, which makes business processes to be exposed to many exceptional situations. These properties of business processes makes it difficult to guarantee successful termination of business processes at the design phase. The behavioral properties of business processes mainly depends on the data aspects of business processes. To formalize the data aspect of process modeling, this paper proposes a graph-based model, called Data Dependency Graph (DDG), constructed from dependency relationships specified between business data. The paper also defines a mechanism of describing a set of mapping rules that generates a process model semantically equivalent to a DDG, which is accomplished by allocating data dependencies to component activities.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2004년도 심포지엄 논문집 정보 및 제어부문
• /
• pp.65-67
• /
• 2004

Change of state of Channel between two wireless terminals which is caused by noise and multiple environmental conditions for happens frequently from the Wireles Network. So, When it is like that planning a wireless network protocol or performance analysis, it follows to change of state of time-varying channel and packet the analysis against a transmission efficiency is necessary. In this paper, analyzes transmission time of a packet and a packet in a time-varying and packet based Wireless Network. To reflecte the feature of the time-varying channel, we use a Signal Flow Graph model. From the model the mean of transmission time and the mean of queue length of the packet are analyzed in terms of the packet distribution function, the packet transmission service time, and the PER of the time-varying channel.

• Journal of Information Science Theory and Practice
• /
• 제6권1호
• /
• pp.18-30
• /
• 2018

Interest in open data is continuing to grow around the world. In particular, open government data are considered an important element in securing government transparency and creating new industrial values. The South Korean government has enacted legislation on opening public data and provided diversified policy and technical support. However, there are also limitations to effectively utilizing open data in various areas. This paper introduces an administrative district knowledge model to improve the sharing and utilization of open government data, where the data are semantically linked to generate a knowledge graph that connects various data based on administrative districts. The administrative district knowledge model semantically models the legal definition of administrative districts in South Korea, and the administrative district knowledge graph is linked to data that can serve as an administrative basis, such as addresses and postal codes, for potential use in hospitals, schools, and traffic control.

• 대한수학회보
• /
• 제56권5호
• /
• pp.1199-1210
• /
• 2019

In this paper, we consider a class of inhomogeneous random intersection graphs by assigning random weight to each vertex and two vertices are adjacent if they choose some common elements. In the inhomogeneous random intersection graph model, vertices with larger weights are more likely to acquire many elements. We show the Poisson convergence of the number of induced copies of a fixed subgraph as the number of vertices n and the number of elements m, scaling as $m={\lfloor}{\beta}n^{\alpha}{\rfloor}$ (${\alpha},{\beta}>0$), tend to infinity.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권12호
• /
• pp.3383-3397
• /
• 2023

Scene graphs serve as semantic abstractions of images and play a crucial role in enhancing visual comprehension and reasoning. However, the performance of Scene Graph Generation is often compromised when working with biased data in real-world situations. While many existing systems focus on a single stage of learning for both feature extraction and classification, some employ Class-Balancing strategies, such as Re-weighting, Data Resampling, and Transfer Learning from head to tail. In this paper, we propose a novel approach that decouples the feature extraction and classification phases of the scene graph generation process. For feature extraction, we leverage a transformer-based architecture and design an adaptive calibration function specifically for predicate classification. This function enables us to dynamically adjust the classification scores for each predicate category. Additionally, we introduce a Distribution Alignment technique that effectively balances the class distribution after the feature extraction phase reaches a stable state, thereby facilitating the retraining of the classification head. Importantly, our Distribution Alignment strategy is model-independent and does not require additional supervision, making it applicable to a wide range of SGG models. Using the scene graph diagnostic toolkit on Visual Genome and several popular models, we achieved significant improvements over the previous state-of-the-art methods with our model. Compared to the TDE model, our model improved mR@100 by 70.5% for PredCls, by 84.0% for SGCls, and by 97.6% for SGDet tasks.