Junyoung Heo

AdaBoost based bankruptcy forecasting of Korean construction companies

AdaBoost (adaptive boosting) is an appropriate model to judge the financial risk of Korean construction companies.We classified construction companies into three groups - large, middle, and small based on the capital of a company.We analyzed the predictive ability of the AdaBoost and other algorithms for each group of companies.The AdaBoost has more predictive power than others, especially for the large group of companies that has the capital more than 50 billion won. A lot of bankruptcy forecasting model has been studied. Most of them uses corporate finance data and is intended for general companies. It may not appropriate for forecasting bankruptcy of construction companies which has big liquidity. It has a different capital structure, and the model to judge the financial risk of general companies can be difficult to apply the construction companies. The existing studies such as traditional Z-score and bankruptcy prediction using machine learning focus on the companies of non-specific industries. The characteristics of companies are not considered at all. In this paper, we showed that AdaBoost (adaptive boosting) is an appropriate model to judge the financial risk of Korean construction companies. We classified construction companies into three groups - large, middle, and small based on the capital of a company. We analyzed the predictive ability of the AdaBoost and other algorithms for each group of companies. The experimental results showed that the AdaBoost has more predictive power than others, especially for the large group of companies that has the capital more than 50 billion won.

Enhancing the Reliability of Head Nodes in Underwater Sensor Networks

Underwater environments are quite different from terrestrial environments in terms of the communication media and operating conditions associated with those environments. In underwater sensor networks, the probability of node failure is high because sensor nodes are deployed in harsher environments than ground-based networks. The sensor nodes are surrounded by salt water and moved around by waves and currents. Many studies have focused on underwater communication environments in an effort to improve the data transmission throughput. In this paper, we present a checkpointing scheme for the head nodes to quickly recover from a head node failure. Experimental results show that the proposed scheme enhances the reliability of the networks and makes them more efficient in terms of energy consumption and the recovery latency compared to the previous scheme without checkpointing.

A Hybrid Approach for Improving the Data Quality of Mobile Phone Sensing

Few studies have researched the temporal and spatial effects of insufficient exposure of sensors in mobile phone sensing. In this paper, the missing data problem in mobile phone sensing is addressed by using a hybrid approach to design an estimation model. This estimation model reflects the effects of participatory and opportunistic nodes based on the success probability model. The proposed model considers the spatial and temporal correlation of sensing data to accurately estimate the missing information. By applying the linear regression and linear interpolation models to sample data from neighboring nodes of the missing data, the spatial and temporal context can be described. The experiment results show that the proposed model can estimate the missing data accurately in terms of simulated and real-world datasets.

Energy efficient program updating for sensor nodes with flash memory

Updating sensor node programs is an essential task for maintaining stability and modifying the characteristics of wireless sensor networks. The updating mechanism must consider energy and memory efficiency, because of resource constraints of sensor nodes. In this paper, we propose a novel program updating mechanism, which considers resource constraints of sensor nodes. The proposed mechanism was designed for sensor nodes with the NOR flash memory. This is generally used to store program image. It was designed to minimize the number of flash write/erase operations, which consume a great deal of energy, and to provide wear-leveling for the NOR flash memory. We set a function as the basic unit of program updating, and partition a function into fixed-sized blocks that can be separately relocated in memory. Experimental results show that the proposed mechanism outperforms other mechanisms in terms of energy, memory and wear-leveling for flash memory.

SWICOM: An SDR-Based Wireless Communication Gateway for Vehicles

A wide range of emerging and promising wireless communication protocols are rapidly being introduced into vehicles. They are commonly used for in-car infotainment, telematics, and safety applications. However, adopting new wireless communications into vehicles requires them to be equipped with the corresponding hardware devices. This hardware dependency incurs extra costs to customers to deploy and maintain wireless services in vehicles. To alleviate this problem, this paper proposes a novel wireless communication gateway for vehicles that is called the software-defined radio (SDR)-based wireless communication gateway (SWICOM). It exploits the SDR technology that uses software running on a generic hardware platform to perform signal processing instead of dedicated hardware. The SWICOM can thus integrate multiple wireless hardware devices into a single generic wireless gateway device, which improves flexibility, adaptability, and connectivity of wireless communications. We built its prototype implementation and performed measurements to quantify its run-time performance. The worst-case execution time (WCET) analysis is also given using synchronous data flow (SDF) graphs. All entire results clearly show the viability of the SWICOM.

Routing Protocol based on Data Aggregation with Energy Efficiency in Underwater Wireless Sensor Networks

In underwater wireless sensor networks, sensing data such temperature and salinity can be merged by averaging them. Data aggregation is a good choice to reduce the amount of packets and save energy in underwater wireless sensor networks. However, data aggregation could bring about packet delay and non-directional transmission to the sink. In this paper, we propose a new path building algorithm based on data aggregation to mitigate these problems. The paper reduces the delay without wireless interferences and maximizes the energy efficiency by removing the non-directional transmission to the sink. Experimental results show that the proposed algorithm outperforms in terms of the energy efficiency and the packet delay.

Sybil Attack Detection with Energy Efficiency in Wireless Sensor Networks

Abstract There are lots of vulnerability and chance to be attacked in wireless sensor networks, which has many applications. Among those attacks, sybil attack is to generate a lot of false node and to inject false information into networks. When a user uses such false information without recognizing the attack, there might be a disaster. Although authentication method can be used to protect such attack, the method is not a good choice in wireless sensor networks, where sensor nodes have a limited battery and low power. In this paper, we propose a novel method to detect sybil attack with a little extra overhead. The proposed method use the characteristics that there is a weak connection between a group of normal nodes and a group of false nodes. In addition, the method uses energy aware routing based on random routing and adds a little information into the routing. Experimental results show that the proposed method detects false node by more than 90% probability with a little energy overhead.

Fault tolerant framework and techniques for component-based autonomous robot systems

Due to the benefits of its reusability and productivity, the component-based approach has become the primary technology in service robot software frameworks, such as MRDS (Microsoft Robotics Developer Studio), RTC (Robot Technology Component), ROS (Robot Operating System) and OPRoS (Open Platform for Robotic Services). However, all the existing frameworks are very limited in fault tolerance support, even though the fault tolerance function is crucial for the commercial success of service robots. In this paper, we present a rule-based fault tolerant framework together with widely-used, representative fault tolerance measures. With our observation that most faults in components and applications in service robot systems have common patterns, we equip the framework with the required fault tolerant functions. The system integrators construct fault tolerance applications from non-fault-aware components by declaring fault handling rules in configuration descriptors or/and adding simple helper components, considering the constraints of the components and the operating environment. Much more consistency in system reliability can be obtained with less effort of system developer. Various fault scenarios with a test robot system on the proposed OPRoS fault tolerant framework demonstrate the benefits and effectiveness of the proposed approach.

Dynamic Offloading Algorithm for Drone Computation

In this paper, we propose a mobility-aware offloading decision method for tracking moving object on drone. It aims to shorten response time because the time constraint is a crucial criteria for offloading decision making for tasks involved in moving object recognition and tracking. Through the dwell time model of target moving object, we investigate the impact of target mobility and derive a dynamic offloading decision method considering the expected dwell time. Our proposed model is also designed to give feedback on the network failure rate so that it can estimate accurately the response time.

Deep-learning based web UI automatic programming

The GUI building is an important part of web application development. Various studies such as WYSWYG web editor have been conducted to make this job convenient, where the job is composed of sketching of GUI and coding of HTML/CSS from the sketch. In this paper, we propose a novel way of web GUI building with computer vision and deep-learning. The proposed method requires only a hand-drawn sketch to build GUI. It recognizes web layout using computer vision algorithm, and web widgets using Faster R-CNN. With the recognized information, it makes HTML code automatically.