• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.11 no.4
• /
• pp.1419-1429
• /
• 2010

Industry structure and environment of the domestic bank have been changed by an influx of large foreign-banks and advanced financial products when the currency crisis erupted in Korea. In a competitive environment, accurate forecasts of changes and tendencies are essential for the survival and development. Forecast of whether to approve loan applications for customer or not is an important matter because that is related to profit generation and risk management on the bank. Therefore, this paper proposes the method to improve forecast accuracy of loan underwriting. Processes in experiments are as follows. First, we select the predictor variables which affect significantly to the result of loan underwriting by correlation analysis and feature selection technique, and then cluster the customers by the 2-Step clustering technique based on selected variables. Second, we find the most accurate forecasting model for each clustering by applying LR, NN and SVM. Finally, we compare the forecasting accuracy of the proposed method with the forecasting accuracy of existing application way.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.8
• /
• pp.958-964
• /
• 2016

In this paper, we propose an i-vector based language recognition system to identify the spoken language of the speaker, which uses multiple discriminative deep neural network (DNN) models analogous to the multi-class support vector machine (SVM) classification system. The proposed model was trained and tested using the i-vectors included in the NIST 2015 i-vector Machine Learning Challenge database, and shown to outperform the conventional language recognition methods such as cosine distance, SVM and softmax NN classifier in open-set experiments.

• Journal of Computing Science and Engineering
• /
• v.5 no.3
• /
• pp.210-222
• /
• 2011

In this work, we study the problem of privacy-preservation data publishing in moving objects databases. In particular, the trajectory of a mobile user in a plane is no longer a polyline in a two-dimensional space, instead it is a two-dimensional surface of fixed width $2A_{min}$, where $A_{min}$ defines the semi-diameter of the minimum spatial circular extent that must replace the real location of the mobile user on the XY-plane, in the anonymized (kNN) request. The desired anonymity is not achieved and the entire system becomes vulnerable to attackers, since a malicious attacker can observe that during the time, many of the neighbors' ids change, except for a small number of users. Thus, we reinforce the privacy model by clustering the mobile users according to their motion patterns in (u, ${\theta}$) plane, where u and ${\theta}$ define the velocity measure and the motion direction (angle) respectively. In this case, the anonymized (kNN) request looks up neighbors, who belong to the same cluster with the mobile requester in (u, ${\theta}$) space: Thus, we know that the trajectory of the k-anonymous mobile user is within this surface, but we do not know exactly where. We transform the surface's boundary poly-lines to dual points and we focus on the information distortion introduced by this space translation. We develop a set of efficient spatiotemporal access methods and we experimentally measure the impact of information distortion by comparing the performance results of the same spatiotemporal range queries executed on the original database and on the anonymized one.

• Steel and Composite Structures
• /
• v.17 no.3
• /
• pp.215-236
• /
• 2014

The flexural behaviour of steel beams significantly affects the structural performance of the steel frame structures. In particular, the flexural overstrength (namely the ratio between the maximum bending moment and the plastic bending strength) that steel beams may experience is the key parameter affecting the seismic design of non-dissipative members in moment resisting frames. The aim of this study is to present a new formulation of flexural overstrength factor for steel beams by means of artificial neural network (NN). To achieve this purpose, a total of 141 experimental data samples from available literature have been collected in order to cover different cross-sectional typologies, namely I-H sections, rectangular and square hollow sections (RHS-SHS). Thus, two different data sets for I-H and RHS-SHS steel beams were formed. Nine critical prediction parameters were selected for the former while eight parameters were considered for the latter. These input variables used for the development of the prediction models are representative of the geometric properties of the sections, the mechanical properties of the material and the shear length of the steel beams. The prediction performance of the proposed NN model was also compared with the results obtained using an existing formulation derived from the gene expression modeling. The analysis of the results indicated that the proposed formulation provided a more reliable and accurate prediction capability of beam overstrength.

• Journal of the Korea Society of Computer and Information
• /
• v.19 no.1
• /
• pp.149-156
• /
• 2014

NGE(Nested Generalized Exemplar Method) can increase the performance of the noisy data at the same time, can reduce the size of the model. It is the optimal distance-based classification method using a matching rule. NGE cross or overlap hyperrectangles generated in the learning has been noted to inhibit the factors. In this paper, We propose the DHGen(Dominant Hyperrectangle Generation) algorithm which avoids the overlapping and the crossing between hyperrectangles, uses interval weights for mixed hyperrectangles to be splited based on the mutual information. The DHGen improves the classification performance and reduces the number of hyperrectangles by processing the training set in an incremental manner. The proposed DHGen has been successfully shown to exhibit comparable classification performance to k-NN and better result than EACH system which implements the NGE theory using benchmark data sets from UCI Machine Learning Repository.

• Journal of Digital Contents Society
• /
• v.2 no.1
• /
• pp.51-61
• /
• 2001

The FLC(Fuzzy Logic Controller) is stronger to the disturbance and has the excellent characteristic to the overshoot of the initialized value than the classical controller, and also can carry out the proper control being out of all relation to the mathematical model and parameter value of the system. But it has the restriction which can't adopt the environment changes of the control system because of generating the fuzzy control rule through an expert's experience and the fixed value of the once determined control rule, and also can't converge correctly to the desired value because of haying the minute error of the controller output value. Now there are many suggested methods to eliminate the minute error, we also suggest the GA-FNNIC(Genetic Algorithm Fuzzy Neural Network Intelligence Controller) combined FLC with NN(Neural Network) and GA(Genetic Algorithm). In this paper, we compare the suggested GA-FNNIC with FLC and analyze the output characteristics, convergence speed, overshoot and rising time. Finally we show that the GA-FNNIC converge correctly to the desirable value without any error.

• Korean Journal of Agricultural Science
• /
• v.47 no.3
• /
• pp.645-655
• /
• 2020

Bacteria are a very common cause of food poisoning. Moreover, bacteria form biofilms to protect themselves from harsh environments. Conventional detection methods for foodborne bacterial pathogens including the plate count method, enzyme-linked immunosorbent assays (ELISA), and polymerase chain reaction (PCR) assays require a lot of time and effort. Hyperspectral imaging has been used for food safety because of its non-destructive and real-time detection capability. This study assessed the feasibility of using hyperspectral imaging and machine learning techniques to detect biofilms formed by Escherichia coli. E. coli was cultured on a high-density polyethylene (HDPE) coupon, which is a main material of food processing facilities. Hyperspectral fluorescence images were acquired from 420 to 730 nm and analyzed by a single wavelength method and machine learning techniques to determine whether an E. coli culture was present. The prediction accuracy of a biofilm by the single wavelength method was 84.69%. The prediction accuracy by the machine learning techniques were 87.49, 91.16, 86.61, and 86.80% for decision tree (DT), k-nearest neighbor (k-NN), linear discriminant analysis (LDA), and partial least squares-discriminant analysis (PLS-DA), respectively. This result shows the possibility of using machine learning techniques, especially the k-NN model, to effectively detect bacterial pathogens and confirm food poisoning through hyperspectral images.

• The KIPS Transactions:PartB
• /
• v.16B no.1
• /
• pp.63-70
• /
• 2009

BCI technology is to control communication systems or machines by brain signal among biological signals followed by signal processing. For the implementation of BCI systems, it is required that the characteristics of brain signal are learned and analyzed in real-time and the learned characteristics are applied. In this paper, we detect feature vector of EEG signal on left and right hand movements based on incremental approach and dimension reduction using the detected feature vector. In addition, we show that the reduced dimension can improve the classification performance by removing unnecessary features. The processed data including sufficient features of input data can reduce the time of processing and boost performance of classification by removing unwanted features. Our experiments using K-NN classifier show the proposed approach 5% outperforms the PCA based dimension reduction.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.19 no.3
• /
• pp.165-171
• /
• 2019

The study of the computer architecture has shown that major improvements in price-to-energy performance stems from domain-specific hardware development. This paper analyzes the tensor processing unit (TPU) ASIC which can accelerate the reasoning of the artificial neural network (NN). The core device of the TPU is a MAC matrix multiplier capable of high-speed operation and software-managed on-chip memory. The execution model of the TPU can meet the reaction time requirements of the artificial neural network better than the existing CPU and the GPU execution models, with the small area and the low power consumption even though it has many MAC and large memory. Utilizing the TPU for the tensor flow benchmark framework, it can achieve higher performance and better power efficiency than the CPU or CPU. In this paper, we analyze TPU, simulate the Python modeled OpenTPU, and synthesize the matrix multiplication unit, which is the key hardware.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.10
• /
• pp.1435-1440
• /
• 2021

With the advent of the low interest rate era, many investors are flocking to the stock market. In the past stock market, people invested in stocks labor-intensively through company analysis and their own investment techniques. However, in recent years, stock investment using artificial intelligence and data has been widely used. The success rate of stock prediction through artificial intelligence is currently not high, so various artificial intelligence models are trying to increase the stock prediction rate. In this study, we will look at various artificial intelligence models and examine the pros and cons and prediction rates between each model. This study investigated as stock prediction programs using artificial intelligence artificial neural network (ANN), deep learning or hierarchical learning (DNN), k-nearest neighbor algorithm(k-NN), convolutional neural network (CNN), recurrent neural network (RNN), and LSTMs.