Asymptotic equivalence between frequentist and Bayesian prediction limits for the Poisson distribution

Bejleri Valbona,Sartore Luca,Nandram Balgobin 한국통계학회 2022 Journal of the Korean Statistical Society Vol.51 No.3

Bayesian prediction limits are constructed based on some maximum allowed probability of wrong prediction. However, the frequency of wrong prediction in a long run often exceeds this probability. The literature on frequentist and Bayesian prediction limits, and their interpretation is sparse; more attention is given to prediction intervals obtained based on parameter estimates or empirical studies. Under the Poisson distribution, we investigate frequentist properties of Bayesian prediction limits derived from conjugate priors. The frequency of wrong prediction is used as a criterion for their comparison. Bayesian prediction based on the uniform and Jeffreys’ non-informative priors yield one sided prediction limits that can be interpreted in a frequentist context. It is shown here, by proving a theorem, that Bayesian lower prediction limit derived from Jeffreys’ noninformative prior is the only optimal (largest) Bayesian lower prediction limit that possesses frequentist properties. In addition, it is concluded as corollary that there is no prior distribution such that Bayesian upper and lower prediction limits obtained from it will both coincide with their respective frequentist prediction limits. Our results are based on asymptotic considerations. An example with real data is included, and the sensitivity of the Bayesian prediction limits with respect to conjugate priors is numerically explored through simulations.

Performance Prediction Model of University Students Based on the Grey BP Neural Network

Liao Yu,Liu Zongxin 보안공학연구지원센터 2016 International Journal of u- and e- Service, Scienc Vol.9 No.10

This article counted the best performance of students entrepreneurship courses from 2005 to 2014, and took the best performance prediction of 2014 entrepreneurship course as the research object. According to the best annual performance of entrepreneurship courses from 2005 to 2014, this article established the grade prediction model of series combination of GM (1, 1) grey prediction model and BP neural network prediction model, and the established model was used to predict the best annual performance of students entrepreneurship course. Through comparing the actual value of the best annual performance of 2014 entrepreneurship course and the predicted value c by the model, this article analyzed the application of grey BP neural network prediction model in the students entrepreneurship performance prediction. The research results showed that for entrepreneurship performance prediction problem, the grey BP neural network prediction model had high prediction precision , simple application, and it can be widely used, and had more advantages than single GM (1, 1) grey prediction model and BP neural network model.

A Climate Prediction Method Based on EMD and Ensemble Prediction Technique

Shuoben Bi,Shengjie Bi,Xuan Chen,Han Ji,Ying Lu 한국기상학회 2018 Asia-Pacific Journal of Atmospheric Sciences Vol.54 No.4

Observed climate data are processed under the assumption that their time series are stationary, as in multi-step temperature and precipitation prediction, which usually leads to low prediction accuracy. If a climate system model is based on a single prediction model, the prediction results contain significant uncertainty. In order to overcome this drawback, this study uses a method that integrates ensemble prediction and a stepwise regression model based on a mean-valued generation function. In addition, it utilizes empirical mode decomposition (EMD), which is a new method of handling time series. First, a non-stationary time series is decomposed into a series of intrinsic mode functions (IMFs), which are stationary and multi-scale. Then, a different prediction model is constructed for each component of the IMF using numerical ensemble prediction combined with stepwise regression analysis. Finally, the results are fit to a linear regression model, and a short-term climate prediction system is established using the Visual Studio development platform. The model is validated using temperature data from February 1957 to 2005 from 88 weather stations in Guangxi, China. The results show that compared to single-model prediction methods, the EMD and ensemble prediction model is more effective for forecasting climate change and abrupt climate shifts when using historical data for multi-step prediction.

Simulation Study on Optimizing Neural Network in Short-Term Electric Load Prediction

Tan Zhongfu,Xin He,Ju Liwei 보안공학연구지원센터 2016 International Journal of Multimedia and Ubiquitous Vol.11 No.3

It researches the short-term electric load prediction and short-term electric load has the characteristics of time-varying, uncertainty, nonlinearity, etc., so the traditional linear prediction method cannot correctly describe the changing rule of the short-term electric load prediction, and neural network has the deficiencies including local minimum value of neural network, over-fitting and weak generalization ability, and the prediction accuracy is lower. In order to improve the accuracy of the short-term electric load prediction, this paper proposes a short-term electric load prediction model (IQPSO-BPNN) based on optimizing BP neural network. Firstly, it improves Quantum Particle Swarm Optimization to optimize the BP neural network parameters, and then adopts the optimized BP neural network to conduct modeling for the nonlinear change law of the short-term electric load prediction. Finally, it takes simulation test for the model performance. The simulation result shows that IPQPSO solves the problems of the BP neural network, and improve the prediction accuracy of the short-term electric load and reduce the prediction error.

복잡지형에 대한 WAsP의 풍속 예측성 평가

윤광용(Yoon Kwang Yong),유능수(Yoo Neung Soo),백인수(Paek In Su) 강원대학교 산업기술연구소 2008 産業技術硏究 Vol.28 No.1

A linear wind prediction program, WAsP, was employed to predict wind speed at two different sites located in complex terrain in South Korea. The reference data obtained at locations more than 7 kilometers away from the prediction sites were used for prediction. The predictions from the linear model were compared with the measured data at the two prediction sites. Two compensation methods such as a self-prediction error method and a delta ruggedness index (RIX) method were used to improve the wind speed prediction from WAsP and showed a good possibility. The wind speed prediction errors reached within 3.5% with the self prediction error method, and within 10% with the delta RIX method. The self prediction error method can be used as a compensation method to reduce the wind speed prediction error in WAsP.

Research on Prediction of Reverse Returned Logistics Based on Grey-Markov Model

Yuming Luo 보안공학연구지원센터 2016 International Journal of u- and e- Service, Scienc Vol.9 No.8

In order to improve the prediction accuracy of reverse returned logistics, considering it has the characteristics of high volatility and uncertainty, the paper used the theory of Markov Chain to modify the result of Grey prediction. And a Grey-Markov prediction model was established. Several parallel region has been divided used the prediction curve of Grey prediction model as symmetric center. And each region was a state interval. A practical example show that the average relative error rate and the variance ratio of Grey-Markov prediction model was smaller, and the prediction accuracy is higher comparing with the Grey prediction model. The model is effective and feasible.

주가지수 옵션 횡단면 정보의 미래 기대변동성 예측력에 관한 연구

손경우 ( Kyoung Woo Sohn ),김상수 ( Sang Su Kim ) 한국금융학회 2016 금융연구 Vol.30 No.1

본 연구에서는 옵션들이 사후적으로 고평가가 심화될 시점에 시장의 방향성이 전환될 가능성이 높다는 점에 착안하여, 코스피200 옵션들의 횡단면 정보에 의한 미래 기대변동성의 변곡점 예측력을 검정하였다. 옵션시장의 횡단면 정보로는 크게 내재 위험중립확률 분포의 고차적률(왜도 및 첨도)과 내재변동성 곡선(기울기 및 곡률)을 사용하였으며, 미래 기대변동성의 변곡점을 예측하기 위해 프로빗 모형을 이용하였다. 실증분석결과, 일정한 첨도(혹은 곡률)의 조건하에서, 왜도(혹은 기울기)가 내포하고 있는 미래 기대변동성의 방향이 전환될 것으로 예측되었다. 첨도가 높고 왜도가 높을 때 변동성의 상승전환이 예측되었고, 첨도가 낮고 왜도가 낮을 때 변동성의 하락전환이 예측되었다. 미래 기대변동성의 변곡점을 예측하기 위해 첨도의 역할이 중요함을 알 수 있다. 특히, 심외가격 옵션들로부터 얻은 변동성 곡선의 정보(기울기 및 곡률)가 내재 고차적률 보다 더 높은 예측력을 보여, 심외가격 옵션이 미래 기대변동성의 변곡점 예측에 있어서 상대적으로 더 효율적 도구임을 알 수 있다. 이러한 결과들을 이용하여 VKOSPI에 대한 가상의 투자전략을 수행한 결과, 표본 외 예측으로도 금융위기에 기대변동성의 변곡점을 잘 포착하였다. 본 연구의 예측 모형이 투자전략 및 위험관리 측면에서 유용할 수 있음을 시사한다. It is often said that investors’ expectation about market volatility and risk premium is reflected in stock index option price (implied volatility), precisely speaking, cross-sectional information of stock index option market. For example, it is well-known that “Volatility smile” or “volatility skew” is one of the phenomena which can represent investors’ expectation or risk premium related to market uncertainties. However, abnormalities may exist in the market as well. Some authors have been argued about “overvaluation puzzle of put option price.” Broadie, Chernov, and Johannes (2009), Bondarenko (2014) have claimed that it is hard to explain every dimension of overvaluation phenomenon of put option price under the existing theoretical models, because these cross-sectional unprecedented phenomena are reflecting demands of abnormal premium for market volatility and uncertainty, and thus the excessive volatility skew or unusual volatility shape sporadically occurs. This study puts a concern about information implied in cross-sectional abnormalities rather than it investigates the abnormalities themselves. Many articles have demonstrated that when investors demand abnormal premium or have extremely biased expectation over economic uncertainty, prices of related options tend to be extremely overvalued or undervalued. Based on this empirical tendency, we can infer that a situation where investors’ overvaluation (undervaluation) of put options deepens might lead to decrease (increase) in market volatility for the future with a high possibilities. Accordingly, using this inference, we can predict volatility (price of options) if we can know when we observe overvaluation or undervaluation. Therefore, our goal is to examine the prediction power of the inflection point of the future expected market volatility using cross-sectional information of the KOSPI 200 options. To this end, we first use probit model to test for the prediction power of the cross-sectional information of the options. The implied volatility of at-the-money option is used as a proxy variable for the expected market volatility in the model. We use two independent variables sets which can be obtained from cross-sectional stock index options. The one of the set is kurtosis and skewness which are higher moments of the implied risk neutral probability density, and the other set is slope and curvature of volatility curve which are main concerns of this study. Second, we present virtual investment strategies using the VKOSPI, the daily volatility index provided by KRX, to examine the effectiveness of the investment strategy of expected volatility based on our prediction model. Our prediction tests are two-folded in terms of dependent variables. The first one is min/max method measured by peaks and troughs of expected volatility and another one is up/down method measured by expected volatility’s % change of volatility at each time interval. And then, each method is categorized into 4 sub-tests by differing independent variables, such as implied higher moments and pairs of information of volatility curves. Thus, we test the 8 sub-models of prediction in total and compare their prediction power of the cross-sectional information of the options. The results show us that first, the expected volatility seems to be near the trough (peak) when skewness or slope is getting higher (lower) and kurtosis or curvature is getting higher. Second, the prediction power of our model is strong enough to capture the expected volatility located in the near-troughs and the near-peak area. Third, prediction power is improved when the implied kurtosis or curvature is added to the model, rather than only using skewness or slope. Fourth, prediction power is relatively high when information of volatility curve, rather than implied risk neutral probability density, is used. Since slope and curvature of volatility curve are more related to the information of deep-out-of-the-money option, it can be interpreted that they are useful to predict the inflection point of expected volatility with higher prediction power. Moreover, Our empirical findings mentioned in the above works well in out-of-sample prediction as well as in-sample prediction. Using the probability obtained from the prediction model with slope and curvature of volatility curve, we found that the strategy that virtually invest in VKOSPI index outperforms others. Surprisingly, the 2008 financial crisis at which the volatility is soared up, and subsequent period at which volatility trend shows decreasing are also well captured from out-of-sample prediction. Therefore, we can conclude that our prediction model is a useful strategy of defensive investment in the practical portfolio point of view.

온라인 언급이 기업 성과에 미치는 영향 분석

정지선(Ji Seon Jeong),김동성(Dong Sung Kim),김종우(Jong Woo Kim) 한국지능정보시스템학회 2015 지능정보연구 Vol.21 No.4

Due to the development of internet technology and the rapid increase of internet data, various studies are actively conducted on how to use and analyze internet data for various purposes. In particular, in recent years, a number of studies have been performed on the applications of text mining techniques in order to overcome the limitations of the current application of structured data. Especially, there are various studies on sentimental analysis to score opinions based on the distribution of polarity such as positivity or negativity of vocabularies or sentences of the texts in documents. As a part of such studies, this study tries to predict ups and downs of stock prices of companies by performing sentimental analysis on news contexts of the particular companies in the Internet. A variety of news on companies is produced online by different economic agents, and it is diffused quickly and accessed easily in the Internet. So, based on inefficient market hypothesis, we can expect that news information of an individual company can be used to predict the fluctuations of stock prices of the company if we apply proper data analysis techniques. However, as the areas of corporate management activity are different, an analysis considering characteristics of each company is required in the analysis of text data based on machine-learning. In addition, since the news including positive or negative information on certain companies have various impacts on other companies or industry fields, an analysis for the prediction of the stock price of each company is necessary. Therefore, this study attempted to predict changes in the stock prices of the individual companies that applied a sentimental analysis of the online news data. Accordingly, this study chose top company in KOSPI 200 as the subjects of the analysis, and collected and analyzed online news data by each company produced for two years on a representative domestic search portal service, Naver. In addition, considering the differences in the meanings of vocabularies for each of the certain economic subjects, it aims to improve performance by building up a lexicon for each individual company and applying that to an analysis. As a result of the analysis, the accuracy of the prediction by each company are different, and the prediction accurate rate turned out to be 56% on average. Comparing the accuracy of the prediction of stock prices on industry sectors, ‘energy/chemical’, ‘consumer goods for living’ and ‘consumer discretionary’ showed a relatively higher accuracy of the prediction of stock prices than other industries, while it was found that the sectors such as ‘information technology’ and ‘shipbuilding/transportation’ industry had lower accuracy of prediction. The number of the representative companies in each industry collected was five each, so it is somewhat difficult to generalize, but it could be confirmed that there was a difference in the accuracy of the prediction of stock prices depending on industry sectors. In addition, at the individual company level, the companies such as ‘Kangwon Land’, ‘KT & G’ and ‘SK Innovation’ showed a relatively higher prediction accuracy as compared to other companies, while it showed that the companies such as ‘Young Poong’, ‘LG’, ‘Samsung Life Insurance’, and ‘Doosan’ had a low prediction accuracy of less than 50%. In this paper, we performed an analysis of the share price performance relative to the prediction of individual companies through the vocabulary of pre-built company to take advantage of the online news information. In this paper, we aim to improve performance of the stock prices prediction, applying online news information, through the stock price prediction of individual companies. Based on this, in the future, it will be possible to find ways to increase the stock price prediction accuracy by complementing the problem of unnecessary words that are added to the sentiment dictionary.

재무변수 및 주식가격 변수를 이용한 회사채 신용등급 예측모형의 개발

김진선,최영문 韓國公認會計士會 2006 회계·세무와 감사 연구 Vol.43 No.-

선행연구에 의해 개발된 회사채 신용등급 예측모형에서는 신용평가 기관이 특정기업의 신용평가시 사용할 것으로 상정된 여러 가지 재무변수들이 예측모형에 독립변수로 포함된 바, 이 독립변수들은 모두 객관적으로 관찰이 가능한 과거 및 현재의 재무적 특성들에 관한 것으로 한정되었다. 그러나 실제로 신용평가기관이 특정기업의 회사채 신용등급을 결정할 때는 해당 기업의 과거 및 현재의 재무적 특성뿐만 아니라 당해 기업의 미래 경영성과(전망)를 고려하리라는 것은 상식에 부합한다. 따라서 본 연구에서는 기업의 미래의 수익성과 관련성을 갖는 주식가격 변수를 선택하고 이들 주식가격 변수가 회사채 신용등급 예측모형에서 유의한 설명력을 가지며 모형의 등급예측력(예측적중률)을 높이는데 기여하는지의 여부를 검토한다. 또한 일부 선행연구에서 사용된 바 있는 회귀모형(OLS)의 강건성을 확인할 목적으로 N-probit 모형을 적용한 예측모형도 함께 검토한다. 본 연구에서는 2000년부터 2004년까지의 5년의 기간에 걸쳐 회사채등급이 존재하는 510개 거래소 상장 및 KOSDAQ 등록기업을 표본으로 하여 이중 340개 기업-년 관찰치를 예측표본으로 하고 나머지 170개 기업-년 관찰치를 예측모형의 정확도(예측적중률)를 검토하기 위한 검증표본으로 하였다. 중요한 검증결과는 다음과 같다. 첫째, 15등급 체계의 경우(+, 0, -를 별개의 등급으로 정의할 때) 최종 예측모형의 설명력은 75.7%에 달하며, 한 등급 차이 내로 예측할 수 있는 정확도가 74.1%에 이른다. 회사채 등급의 +, 0, -를 구분하지 않은 5등급 체계의 경우의 모형의 전체적인 설명력은 70.3%이며, 예측된 등급과 실제 등급이 일치하는 적중률은 70.0%로서 신용등급을 6등급으로 분류하여 예측한 Kaplan and Urwitz(1979)의 예측적중률 69%에 비견할 만하다. 둘째, 기업의 미래 수익성 전망을 대표(proxy)하는 주식가격 변수로서 주가순자산 비율과 누적시장조정수익률의 상대적 기여도를 분석한 결과 예측모형에서 이들 주식가격 변수의 계수추정치가 통계적 유의성을 보이며 이 두 변수가 예측모형의 설명력과 예측정확도에 미치는 기여도가 적지 않은 것으로 나타나고 있다. 이러한 검증결과는 기업신용평가기관이 회사채 신용등급을 결정함에 있어 기업의 과거 및 현재의 재무적 특성뿐만 아니라 미래의 수익성 전망을 일정수준 반영하는 증거로 해석된다. 셋째, 위의 검증결과들을 N-probit 모형을 사용하여 비교․검토한 결과 OLS 회귀모형을 적용한 경우와 거의 동일한 결과가 도출되어 OLS 회귀모형의 강건성을 확인할 수 있다. 15등급 체계 및 5등급 체계의 경우 모두 OLS 회귀모형에 기초하여 도출된 예측모형에 포함된 독립변수들이 N-probit 모형을 사용하여 추정된 예측모형에서도 예외 없이 통계적 유의성을 보이며, 예측모형의 예측정확도는 약간 높거나(15등급 체계의 경우) 유사하다(5등급 체계의 경우). 또한, 예측모형에서 주식가격 변수의 계수추정치가 통계적 유의성을 보이며 예측모형의 예측정확도에 미치는 공헌(기여)도 재차 확인된다. 이러한 연구결과는 향후의 연구가 주식가격 변수를 이용할 수 없는 비상장기업에 적용가능한 회사채 신용등급 예측모형의 개발에 집중될 필요가 있음을 간접적으로 시사하고 있다. Bond rating prediction models developed in the existing studies include observable past and present financial characteristics only as independent variables. However, it may be a reality that bond rating agencies -- when assigning bond ratings -- would take into account future earnings prospects of rated companies as well as those firms' past and present financial characteristics. Hence this study selects two stock price-based variables, which may be potentially correlated with future corporate earnings performance, to investigate if these stock price-based variables add statistically significant explanatory power to the bond rating prediction model and contribute to enhancing the prediction accuracy of the bond rating prediction model. Selected stock price-based variables are Price-to-Book ratios and cumulative market-adjusted returns. The bond rating prediction model developed in this study adopts both OLS regression method and N-probit model, using 510 firm-year observations of a sample of December year-end non-financial firms listed on the Korea Stock Exchange and KOSDAQ over the 2000-2004 period. Among 510 firm-year observations, two thirds are used as an estimation sample, while the remaining constitute a validation sample to measure the prediction accuracy of the model. First of all, the overall explanatory power of the model reached 75.7%, and coefficients of most independent variables showed signs as predicted. The final model predicted correctly 74.1% of the validation sample within a single category of margin across 15-rating hierarchy. In case of 5-rating hierarchy, the final model predicted correctly 70.0% of the validation sample. This hit rate figure is comparable to that of Kaplan and Urwitz(1979). Secondly, two stock price-based variables, which may proxy future profit performance of sample firms, contribute significantly to the prediction accuracy and the overall explanatory power of the model. This result, coupled with significant coefficients of those two variables in the prediction models, indicates that the rating agencies may take into account future earnings prospect as well as past and present financial characteristics of rated firms. Finally, when the prediction model is derived based on an N-probit model, empirical findings are very similar to those of OLS-based prediction models, confirming the robustness of OLS regression models. Both in cases of 15-rating and 5-rating hierarchy, independent variables selected for the OLS regression models show statistical significance in corresponding N-probit models without exception. The prediction accuracy of N-probit models is slightly higher than that of OLS models in case of 15-rating hierarchy, while the results are mixed in case of 5-rating hierarchy. Two stock price-based variables, which proxy future profit performance of sample firms, also contribute significantly to the prediction accuracy of the model, which confirms the finding from OLS-based models. These results indirectly imply that future research should be focused on the development of a bond rating prediction model that can be applicable to non-listed firms for which stock price-based variables are unavailable.

특정인에 대한 범죄예측 시스템의 문제점과 개선방안

김병수(Kim, Byung-Soo) 한국형사법학회 2021 刑事法硏究 Vol.33 No.3

머신러닝 및 딥러닝 기술에 근거한 인공지능이 빅데이터라는 방대한 자료 속에서 숨겨진 정보들 간의 상관관계를 찾아내고 이를 바탕으로 미래에 어떤 일이 발생할 것인지를 통계적으로 예측하고 있다. 이를 활용하여 사회의 많은 분야에서 상당히 높은 수준의 예측을 하고 있다. 이러한 높은 예측력은 범죄예방분야에서도 활용되고 있는데, 이미 선진국에서는 범죄발생률이 높은 대도시를 중심으로 빅 데이터와 인공지능을 이용한 범죄예측 시스템을 도입함으로써 범죄율이 감소하는 등 가시적인 효과를 보고 있다. 이러한 범죄예측 시스템의 방법은 범죄 예측, 범죄 피해자 예측, 범죄자 예측 등 크게 3가지로 구분할 수 있다. 특정한 범죄자와 피해자와 같은 특정인의 범죄를 전망하는 범죄자 예측과 피해자 예측은 개인의 Privacy가 침해된다는 점과 민감한 개인정보의 수집 범위가 제한적이라는 점에서 연구 및 활용이 제한되었다. 이 때문에 범죄예측 시스템은 장소별, 지역별 범죄 위험도를 기반으로 한 범죄발생의 시간과 장소를 전망하는 시공간 범죄예측이 주를 이루고 있다. 이러한 상황에서 본 연구는 특정인에 대한 범죄예측 시스템에서 발생할 수 있는 문제점을 제기하고 이에 대한 신속한 대처방안을 모색하는데 목표를 두고자 한다. 특정인에 대한 범죄예측 시스템은 특정인의 개인정보를 다루기 때문에 필연적으로 헌법상 보장된 사생활의 비밀과 자유 및 개인정보자기결정권 등의 기본권을 침해할 수밖에 없다. 특정인을 대상자로 할 경우 일반인과의 형평성의 문제가 제기될 수 있다. 또한 대상범죄를 어느 범위까지 허용할 것인지 명확한 한계를 설정하기 어렵다. 범죄예측 시스템에 사용되는 인공지능 알고리즘은 불명투명성, 편향성, 부정확성의 문제가 있다. 인공지능 알고리즘의 불투명성은 부정확한 범죄예측으로 인한 피해나 그 책임의 소재를 찾는데 어려움을 줄 수 있다. 또한 인공지능 알고리즘이 지닌 잠재적 편향성과 부정확성 때문에 범죄예측이 공정하지 못하다는 비판을 받고 있다. 그리고 범죄예측 시스템은 시민들에게 새로운 통제수단이 되고 있다. 이러한 문제점을 해결하기 위해서는 먼저, 개인정보보호법 제15조 1항 1호에 의해 정보주체의 사전 동의를 반드시 받도록 하여야 할 것이다. 범죄예측을 위한 특정인에 대한 개인정보수집은 합목적적 범위 내에서 필요최소한에 그쳐야 할 것이고, 잘못된 결과가 발생할 경우를 회피 내지 개선할 수 있는 적당한 조치를 사전에 보장하여야 한다. 모든 개인정보수집은 감독기관의 적절한 감독을 받을 것을 보장하여야 한다. 알고리즘의 불투명성을 해결하는 방법으로는 정보 주체나 일반인을 대신하여 알고리즘의 적절성을 심사하고 감시하는 조직이나 기관을 설치하는 방법을 제시할 수 있다. 편향성 및 부정확성 해결방안으로 범죄예측 시스템을 통한 범죄예방의 효과성이 입증되어 공개되어야 하며 범죄예측 시스템의 알고리즘의 정확성이 담보되어야 할 것이다. 범죄예측 시스템의 적용대상범죄와 대상자의 선정을 위해서는 재범의 중대성과 위험성이 중요한 판단 기준이 될 것이다. 범죄예측 시스템을 사용한 범죄예측은 살인, 강도, 강간, 방화와 같은 중대범죄여야 할 것이다. 범죄예측 시스템이 시민에 대한 통제장치로서 악용될 우려를 불식시키기 위해서는 범죄예측 시스템도 보안처분과 같이 비례성의 원칙에 의하여 대상자의 권리침해를 최소화하여야 한다. 범죄예측 시스템은 형벌이나 다른 보안처분에 의하여 목적을 달성할 수 없는 최후수단으로서 사용되어야 할 것이다. 대상자의 재범위험성을 근거로 개별적으로 사용하여야 한다. 범죄예측시스템은 행정처분에 의하여 행해져서는 안되고 법원의 사법심사에 의하여 부과되어야 한다. Artificial intelligence based on machine learning and deep learning technology finds correlations between hidden information in the vast data of big data and statistically predicts what will happen in the future. Using AI, there are high-level predictions in many fields of society. This high predictive power of AI is also being used in the field of crime prevention. In developed countries, crime prediction systems using big data and artificial intelligence have been introduced intensively in large cities with high crime rates. And they are producing tangible effects such as a decrease in the crime rate. The methods of such a crime prediction system can be broadly divided into 3 categories: ① crime prediction, ② criminal prediction, ③ crime victim prediction. Because of the limited scope of collection of sensitive personal information and protection of personal privacy, crime prediction, which predicts the time and place of a crime based on the crime risk by region and location, is mainly used rather than predicting criminals and victims. In this situation, the purpose of this study is to raise the problems that may appear in the crime prediction system for a specific person and to find a quick response method. Therefore, in this study, the problem of infringement of individual constitutional and legal rights that may occur when the crime prediction system is used for a specific person, the problem of opacity, inaccuracy and bias of the algorithm of the crime prediction system, and the problem of the crime prediction system as a new control means. We will review the potential problems and seek solutions to them. Since the crime prediction system for a specific person deals with the personal information of a specific person, it inevitably violates the basic rights such as privacy and freedom and the right to self-determination of personal information guaranteed by the Constitution. When targeting a specific person, the issue of equity with the general public may be raised. In addition, it is difficult to set clear limits on the extent to which target crimes are permitted. Artificial intelligence algorithms used in crime prediction systems have problems of opacity, bias, and inaccuracy. The opacity of artificial intelligence algorithms can make it difficult to find the cause of damage or responsibility due to inaccurate crime prediction. In addition, due to the potential bias and inaccuracy of artificial intelligence algorithms, crime prediction is being criticized for being unfair. And the crime prediction system is becoming a new means of control for citizens. In order to solve this problem, it is necessary to first obtain the prior consent of the information subject according to Article 15 (1) 1 of the Personal Information Protection Act. The collection of personal information for a specific person for crime prediction should be limited to the minimum necessary within the scope of the purpose of the crime, and appropriate measures to avoid or improve the occurrence of erroneous results should be ensured in advance. It should be ensured that all personal information collection is subject to appropriate supervision by the supervisory authority. As a method of solving the opacity of the algorithm, it can be suggested to establish an organization or institution that examines and monitors the adequacy of the algorithm on behalf of the data subject or the general public. As a solution to bias and inaccuracy, the effectiveness of crime prevention through the crime prediction system should be proven and disclosed, and the accuracy of the crime prediction system algorithm should be guaranteed. The seriousness and risk of recidivism will be an important criterion for the selection of target crimes and targets for the crime prediction system. Crime prediction using the crime prediction system should be serious crimes such as murder, robbery, rape,