학술논문
데이터마이닝 기법을 활용한 공공 IT사업의 낙찰률 예측모델 개발 / Development of Bidding-Ratio Prediction Model for Public Information Technology Business Projects Using Data Mining Method
Document Type
Dissertation/ Thesis
Author
김철 / Kim, Cheol
Source
Subject
Language
Korean
Abstract
공공IT 입찰에 참가하는 기업의 낙찰률은 수익성, 수행원가 그리고 성과물의 품질에 직접적인 영향을 주는 가장 민감한 요소로 낙찰률에 영향을 미치는 다양한 요인과 변수들이 사전에 고려되어야 하나, 과거의 경험에 의존하여 직관적으로 의사결정하고 있어 보다 체계적이고 과학적인 낙찰률 예측을 위한 의사결정이 반드시 필요하다. 본 연구를 진행하면서 선행연구 조사를 검토한 결과, 부동산 경매 낙찰률 예측모형에 대한 연구나 공공 건설입찰 낙찰률에 대한 연구는 다수 확인할 수 있었으나 공공IT사업 입찰에서 낙찰률 예측과 관련된 연구는 확인할 수 없었다. 본 연구는 IT기업이 공공IT 입찰에서 낙찰률을 예측하고 있는 방법을 개선하여 보다 실용적이고 경쟁력 있는 낙찰률 예측모형을 개발하는 것을 목적으로 공공IT 조달입찰 데이터를 수집하여 분석을 수행하였다. 2014년부터 2016년까지 공공IT 조달입찰의 입․낙찰자료 10,747건을 대상으로 낙찰률 예측모형을 개발하였다. 낙찰률 예측을 위해 사용한 분석모형은 실무에서 가장 많이 사용하고 있는 Linear Regression과 최근 머신러닝 분석 중 학술적으로 많은 주목을 받고 있는 Support Vector Regression, Random Forest 3가지 분석을 실시했다. 본 연구에서는 공공IT 입․낙찰 특성 데이터를 활용해 낙찰률 예측모형을 개발하고 낙찰률에 영향을 미치는 변수들의 중요도를 파악하여 공공 IT입찰 시 과학적인 의사결정을 위한 활용 방안을 모색하였으며, 입․낙찰 특성 데이터의 분석만으로도 현실과 매우 유사한 낙찰률 예측이 가능한 것을 확인하였다. 본 연구는 데이터마이닝 기법을 활용하여 실제와 매우 유사한 낙찰률 예측모형을 제시하였고, 낙찰률에 영향을 미치는 인자들의 중요도와 실무 전문가들이 실무에서 체감하는 인자들의 중요도 간의 차이점을 파악하였다. 현재로서는 관련 연구가 없는 상황에서 중소기업이 주류를 이루고 있는 우리나라 공공IT 시장에서 분석력이 열악한 중소기업이 공공IT 입찰 참여시 해당 낙찰률 예측모형을 활용하여 보다 과학적인 의사결정에 활용되기를 기대한다.
For companies participating in public IT bidding, the winning rate is the most sensitive factor that directly affects profitability, cost of execution and quality. Since various factors and variables affecting the winning bid rate must be considered in advance, it is necessary to make a decision for the systematic and scientific prediction of the winning rate because there is a limit to the intuitive decision making that depends on the past experience. Since the execution of this study, it was conducted a preliminary research as well. However, many studies on the prediction model of real estate auction success rate or the public construction bid auction rate were confirmed, but there was no result about a study related to the prediction of the successful bid in public IT bidding that it could be confirmed at all. The purpose of this study is to develop a more practical and competitive bidding rate forecasting model by improving the way that IT companies are predicting the winning bid in public IT bidding. From 2014 to 2016, it was developed a prediction model of winning rate with 10,747 entries and winning bids for public IT procurement bidding. The analytical model used for prediction of the winning rate was three kinds of analysis: Linear Regression, which is the most used in practice, and Support Vector Regression and Random Forest, which have been receiving much attention in the recent machine learning analysis. In this study, it was developed a prediction model of public competitive advantage by using public IT bidding characteristics data and figured out the importance of variables affecting the bidding rate. There were searched the application method for scientific decision making in public IT bidding just through the data analysis confirmed that it is possible to predict the success rate very similar to reality. At present, it is hoped that small and medium enterprises with inferior analytical ability in the Korean public IT market, where smaller enterprises are the mainstream, will utilize the successful bid rate forecasting model to make more for sure.
For companies participating in public IT bidding, the winning rate is the most sensitive factor that directly affects profitability, cost of execution and quality. Since various factors and variables affecting the winning bid rate must be considered in advance, it is necessary to make a decision for the systematic and scientific prediction of the winning rate because there is a limit to the intuitive decision making that depends on the past experience. Since the execution of this study, it was conducted a preliminary research as well. However, many studies on the prediction model of real estate auction success rate or the public construction bid auction rate were confirmed, but there was no result about a study related to the prediction of the successful bid in public IT bidding that it could be confirmed at all. The purpose of this study is to develop a more practical and competitive bidding rate forecasting model by improving the way that IT companies are predicting the winning bid in public IT bidding. From 2014 to 2016, it was developed a prediction model of winning rate with 10,747 entries and winning bids for public IT procurement bidding. The analytical model used for prediction of the winning rate was three kinds of analysis: Linear Regression, which is the most used in practice, and Support Vector Regression and Random Forest, which have been receiving much attention in the recent machine learning analysis. In this study, it was developed a prediction model of public competitive advantage by using public IT bidding characteristics data and figured out the importance of variables affecting the bidding rate. There were searched the application method for scientific decision making in public IT bidding just through the data analysis confirmed that it is possible to predict the success rate very similar to reality. At present, it is hoped that small and medium enterprises with inferior analytical ability in the Korean public IT market, where smaller enterprises are the mainstream, will utilize the successful bid rate forecasting model to make more for sure.