다항식 회귀분석을 이용한 전자저울의 비선형 특성 개선 연구

채규수 중소기업융합학회 2019 융합정보논문지 Vol.9 No.6

본 연구에서는 다항식 회귀분석(Polynomial regression analysis) 방법을 이용하여 비선형 특성을 갖는 전자저울 의 질량 추정 모델 개발이 이루어 졌다. 전자저울에 사용되는 로드셀의 출력 단자 전압을 기준 질량 추를 사용하여 직접 측정하였고 이 데이터를 이용하여 MS Office 엑셀의 행렬식 계산과 데이터 추세선 분석 기능을 이용하여 다항식 회귀 모델을 구하였다. 5kg까지 측정 가능한 로드셀 전자저울을 사용하여 100g단위로 질량을 측정하였고 다항식 회귀분석 (Multiple regression analysis) 모델을 구하였으며, 단순(1차), 2차, 3차 다항식 회귀분석에 대한 오차를 구하였다. 각 모델에 대한 회귀 방정식의 적합도 분석을 위해 결정계수(Coefficient of determination)를 제시하여 추정 질량과 측정 데이터와의 상관관계를 나타내었다. 본 연구에서 제안하는 3차 다항식 모델을 이용하여 추정 값의 표준편차가 10g, 결정계 수 1.0으로 상당히 정확한 모델을 얻었다. 본 연구에 사용된 선형 회귀 분석 이론을 바탕으로 최근 인공지능 분야에서 많이 사용되고 있는 로지스틱 회귀 분석(Logistic regression analysis)을 활용하여 기상예측, 신약개발, 경제지표 분석 등의 분야에 대한 다양한 연구를 수행할 수 있을 것으로 생각된다. In this study, the development of a weight estimation model of electronic scale with nonlinear characteristics is presented using polynomial regression analysis. The output voltage of the load cell was measured directly using the reference mass. And a polynomial regression model was obtained using the matrix and curve fitting function of MS Office Excel. The weight was measured in 100g units using a load cell electronic scale measuring up to 5kg and the polynomial regression model was obtained. The error was calculated for simple(1st), 2nd and 3rd order polynomial regression. To analyze the suitability of the regression function for each model, the coefficient of determination was presented to indicate the correlation between the estimated mass and the measured data. Using the third order polynomial model proposed here, a very accurate model was obtained with a standard deviation of 10g and the determinant coefficient of 1.0. Based on the theory of multi regression model presented here, it can be used in various statistical researches such as weather forecast, new drug development and economic indicators analysis using logistic regression analysis, which has been widely used in artificial intelligence fields.

Robust Nonparametric Regression Method using Rank Transformation

Park, Dongryeon The Korean Statistical Society 2000 Communications for statistical applications and me Vol.7 No.2

Consider the problem of estimating regression function from a set of data which is contaminated by a long-tailed error distribution. The linear smoother is a kind of a local weighted average of response, so it is not robust against outliers. The kernel M-smoother and the lowess attain robustness against outliers by down-weighting outliers. However, the kernel M-smoother and the lowess requires the iteration for computing the robustness weights, and as Wang and Scott(1994) pointed out, the requirement of iteration is not a desirable property. In this article, we propose the robust nonparametic regression method which does not require the iteration. Robustness can be achieved not only by down-weighting outliers but also by transforming outliers. The rank transformation is a simple procedure where the data are replaced by their corresponding ranks. Iman and Conover(1979) showed the fact that the rank transformation is a robust and powerful procedure in the linear regression. In this paper, we show that we can also use the rank transformation to nonparametric regression to achieve the robustness.

Application of Model Tree and Evolutionary Polynomial Regression for Evaluation of Sediment Transport in Pipes

Mohammad Najafzadeh,Daniele Biagio Laucelli,Abdolreza Zahiri 대한토목학회 2017 KSCE Journal of Civil Engineering Vol.21 No.5

Prediction of critical velocity for sediment deposition is a significant component in design of sewer pipes. Because of the abrupt changes in velocity and shear stress distributions, traditional equations based on regression analysis can fail in evaluating sediment transport efficiently. Therefore, different artificial intelligence approaches have been applied to investigate sediment transport in sewer pipes. This study proposes two different approaches to predict the critical velocity for sediment deposition in sewer networks: Model Tree (MT) and the Evolutionary Polynomial Regression (EPR), a hybrid data-driven technique that combines genetic algorithms with numerical regression. The hydraulic radius, average size of sediments, volumetric concentration, total friction factor, and non-dimensional sediment size were considered as input parameters to characterize sediment transport in clean sewer pipes. The present study implements data collected from different works in literature. The proposed modeling approaches are compared to some benchmark formulas from literature, and discussed from the accuracy and knowledge discovery points of view, highlighting the advantage of both proposed techniques. Results indicated that both techniques have similar accuracy in predictions, but EPR allows to physical validation of returned formulas, allowing identifying the most influent inputs on the phenomenon at stake.

호텔 산업의 자아이미지 일치성 측정 방법의 타당성에 관한 연구 - 간접 측정과 다항 회귀식을 중심으로 -

백주아,윤설민,서원석 한국관광학회 2010 관광학연구 Vol.34 No.9

Given the theoretical significance of self-image congruence, it is important to develop a valid measure of self-image congruence. Traditionally, indirect measure has been the most widely used method for measuring self-image congruity. However, several researchers have suggested that indirect measure has several problems. Thus, this study was to empirically compare the explanatory power of the indirect measure and the polynomial regression in measuring self-Image congruity. Data were conveniently collected from 278 people, who had brand image of five star hotels in Seoul, between December 2008 and January 2009. Results showed that the significant self-image congruity effect of indirect measure disappeared under the statistical test of polynomial regression. Findings from this study imply that the traditional indirect measure of self-image congruity might not be appropriate to measure the self-image congruity.

다항식희귀분석을 이용한 기능성곡면의 모델링

윤상환,황종대,정윤교 한국공작기계학회 2002 한국공작기계학회 추계학술대회논문집 Vol.2002 No.-

This research presents modeling of a functional surface which is a constructed free-formed surface. The modeling introduced in this paper adopts polynomial regression that is utilizing approximating technique. The measured data are obtained from measuring with Coordinate Measuring Machine. This paper introduces efficient methods of Reverse Engineering using Polynomial Regression.

머신러닝 기술을 사용한 비트코인 특성들의 비트코인 가격에 미친 영향 분석

윤성욱 한국정보과학회 2019 정보과학회 컴퓨팅의 실제 논문지 Vol.25 No.7

Bitcoin (BTC ticker) is the most popular crypto-currency in the world, the first release of which took place in 2009. In this respect, in a release date for this currency the price was equal basically to none and was not considered as popular as other known crypto-currency available at the time. Today bitcoin is in high demand from users around the globe. It can be exchanged through special exchanges for ordinary money, or used directly as a means of payment for anything of choice by the users. Bitcoin is accepted by many of the largest online stores and online services for products, goods and services worldwide. Quotes of crypto currency are not regulated by any legislative or legal authorities, and therefore are considered fluid, whereby the value depends totally on the current natural demand and supply. Recently, bitcoin has achieved a great success in use within open global markets. By being motivated from a review of these factors, we decided to take a deep look into the main factors and features characteristic of bitcoin. In this project, we tried to take a vision regarding the use and impact of bitcoin features from a dataset based on Blockchain Wallet API, which is derived from one of the most popular bitcoin digital wallets – Blockchain Wallet API [1]. As a target, we set different phases for the analysis and gathering of relevant data from API. In these terms, for calculating the impact of feature we have used machine learning techniques; which included a pure linear regression and expended version of a linear regression-polynomial regression. 2009년에 처음 소개된 비트코인은 전 세계적으로 출시되어있는 암호 화폐들 중 가장 대중적이다. 출시 될 당시에 가치는 아주 낮았고, 다른 암호 화폐처럼 대중적이지도 않았다. 비트코인에 오늘날 많은 사람들이 관심을 가지고 있다. 비트코인은 일반 화폐와 교환될 수도 있고 지불 용도로 사용할 수도 있다. 비트코인은 많은 온라인 상점들과 온라인 서비스에서 통용되고 있다. 암호 화폐는 특정 당국에 의해 규정되지 않고 일반적인 수요와 공급에 따라 규정되지도 않는다. 비트코인은 최근에 상당한 성공을 이루었다. 비트코인 가격이 어느 주요한 요소들에 의해 결정되는지에 대한 호기심이 본 연구를 진행하게 되었다. 가장 대중적인 비트코인 디지털 월렛(www.blockchain.com) 으로부터 Blockchain Wallet API에 기반해서 데이터셋으로부터 특징들을 뽑았다. 머신러닝 기법으로 polynomial regression을 이용하여 특징들의 영향도를 계산하였다. 결과적으로 어떤 특징들이 비트코인 가격 변동과 연관이 깊은지 살펴봤다.

Analyzing Impact of Bitcoin Features to Bitcoin Price via Machine Learning Techniques

Seongwook Youn(윤성욱) 한국정보과학회 2019 정보과학회 컴퓨팅의 실제 논문지 Vol.25 No.7

2009년에 처음 소개된 비트코인은 전 세계적으로 출시되어있는 암호 화폐들 중 가장 대중적이다. 출시 될 당시에 가치는 아주 낮았고, 다른 암호 화폐처럼 대중적이지도 않았다. 비트코인에 오늘날 많은 사람들이 관심을 가지고 있다. 비트코인은 일반 화폐와 교환될 수도 있고 지불 용도로 사용할 수도 있다. 비트코인은 많은 온라인 상점들과 온라인 서비스에서 통용되고 있다. 암호 화폐는 특정 당국에 의해 규정되지 않고 일반적인 수요와 공급에 따라 규정되지도 않는다. 비트코인은 최근에 상당한 성공을 이루었다. 비트코인 가격이 어느 주요한 요소들에 의해 결정되는지에 대한 호기심이 본 연구를 진행하게 되었다. 가장 대중적인 비트코인 디지털 월렛(www.blockchain.com) 으로부터 Blockchain Wallet API에 기반해서 데이터셋으로부터 특징들을 뽑았다. 머신러닝 기법으로 polynomial regression을 이용하여 특징들의 영향도를 계산하였다. 결과적으로 어떤 특징들이 비트코인 가격 변동과 연관이 깊은지 살펴봤다. Bitcoin (BTC ticker) is the most popular crypto-currency in the world, the first release of which took place in 2009. In this respect, in a release date for this currency the price was equal basically to none and was not considered as popular as other known crypto-currency available at the time. Today bitcoin is in high demand from users around the globe. It can be exchanged through special exchanges for ordinary money, or used directly as a means of payment for anything of choice by the users. Bitcoin is accepted by many of the largest online stores and online services for products, goods and services worldwide. Quotes of crypto currency are not regulated by any legislative or legal authorities, and therefore are considered fluid, whereby the value depends totally on the current natural demand and supply. Recently, bitcoin has achieved a great success in use within open global markets. By being motivated from a review of these factors, we decided to take a deep look into the main factors and features characteristic of bitcoin. In this project, we tried to take a vision regarding the use and impact of bitcoin features from a dataset based on Blockchain Wallet API, which is derived from one of the most popular bitcoin digital wallets – Blockchain Wallet API [1]. As a target, we set different phases for the analysis and gathering of relevant data from API. In these terms, for calculating the impact of feature we have used machine learning techniques; which included a pure linear regression and expended version of a linear regression-polynomial regression.

Minimum Message Length and Classical Methods for Model Selection in Univariate Polynomial Regression

Murlikrishna Viswanathan,양영규,황보택근 한국전자통신연구원 2005 ETRI Journal Vol.27 No.6

The problem of selection among competing models has been a fundamental issue in statistical data analysis. Good fits to data can be misleading since they can result from properties of the model that have nothing to do with it being a close approximation to the source distribution of interest (for example, overfitting). In this study we focus on the preference among models from a family of polynomial regressors. Three decades of research has spawned a number of plausible techniques for the selection of models, namely, Akaike’s Finite Prediction Error (FPE) and Information Criterion (AIC), Schwartz’s criterion (SCH), Generalized Cross Validation (GCV), Wallace’s Minimum Message Length (MML), Minimum Description Length (MDL), and Vapnik’s Structural Risk Minimization (SRM). The fundamental similarity between all these principles is their attempt to define an appropriate balance between the complexity of models and their ability to explain the data. This paper presents an empirical study of the above principles in the context of model selection, where the models under consideration are univariate polynomials. The paper includes a detailed empirical evaluation of the model selection methods on six target functions, with varying sample sizes and added Gaussian noise. The results from the study appear to provide strong evidence in support of the MML- and SRM- based methods over the other standard approaches (FPE, AIC, SCH and GCV).

다항회귀 기반 기상 요인과 피해 면적의 상관관계 분석

김문성(Moonsung Kim),박건(Geon Park),추현창(Hyeonchang Chu),함현준(Hyeonjun Ham),이승헌(Seungheon Lee),엄희승(Heeseung Eom),유명한(Myeonghan Yu),정경용(Kyungyong Chung) 한국정보기술학회 2022 Proceedings of KIIT Conference Vol.2022 No.6

매년 산불 발화는 꾸준하게 발생하며 산불로 인한 산림의 손실은 산불 발생 건수에 비례하여 발생한다. 이로 인해 국내에서도 산불로 인한 산림의 피해와 인적·물적 피해를 줄이기 위해 국가 산불 예보 시스템을 구축하는 등 많은 노력을 하고 있다. 산불의 원인은 기상, 지형 등과 같은 자연적인 요인보다는 쓰레기 소각, 입산자 방화, 담배꽁초 등과 같은 인위적인 요인과 더 높은 상관관계를 가진다. 이로 인해 발생한 산불이 대형 산불로 확산하는 데에는 기상적인 요인과 높은 상관관계를 가진다. 본 연구는 다항회귀를 기반으로 기상 요인과 피해 면적 사이의 상관관계 분석을 제안한다. 산림청에서 제공하는 통계자료를 활용하여 산불 발생 시점의 풍속, 기온 등과 같은 기상 데이터와 해당 산불의 피해면적 사이의 상관관계를 다항회귀 분석하고 시각화한다. Every year, forest fires occur continuously, and the loss of forests due to forest fires is proportional to the number of forest fires. Accordingly, in Korea, many efforts are being made, such as establishing a national forest fire forecasting system to reduce damage to forests and human and material damage caused by forest fires. The causes of forest fires in Korea have a higher correlation with artificial factors such as waste incineration, arson of tenants, and cigarette butts rather than natural factors such as weather and topography. There is a higher correlation with meteorological factors in the spread of forest fires caused by this to large forest fires. This study proposes a correlation analysis between weather factors and damage areas based on polynomial regression. This uses statistical data provided by the Korea Forest Service to analyze and visualize the correlation between weather data such as wind speed and temperature at the time of forest fire and the damage area of the forest fire.

Kernel methods for estimating derivatives of conditional quantiles

이영경,이은령 한국통계학회 2008 Journal of the Korean Statistical Society Vol.37 No.4

A collection of quantile regression functions gives a picture of the conditional distribution of the response given the covariates. However, it cannot be used directly to make a firm conclusion on the effects of the covariates. The derivatives of conditional quantiles, instead, are of immediate use for this purpose. They measure how rapidly the conditional quantiles change as the covariates vary, not only in the center of the population, but also in its upper and lower tails. In this paper we consider estimation of the derivatives of conditional quantiles. The estimators suggested in this paper are based on the double-kernel approach of [Yu, K., & Jones, M. C. (1998). Local linear quantile regression. Journal of the American Statistical Association, 93, 228–237] and on the local logistic approach of [Lee, Y. K., Lee, E. R., & Park, B. U. (2006). Conditional quantile regression by local logistic regression. Journal of Nonparametric Statistics, 18, 357–373].We derive the asymptotic distributions of the two estimators, and compare their finite sample performance via a simulation study.