• 제목/요약/키워드: data value prediction

검색결과 1,088건 처리시간 0.024초

슈퍼스칼라 프로세서에서 정적 및 동적 분류를 사용한 혼합형 결과 값 예측기 (A Hybrid Value Predictor using Static and Dynamic Classification in Superscalar Processors)

  • 김주익;박홍준;조영일
    • 한국정보과학회논문지:시스템및이론
    • /
    • 제30권10호
    • /
    • pp.569-578
    • /
    • 2003
  • 데이타 종속성은 명령어 수준 병렬성을 향상시키는데 중요한 장애요소가 되고 있으며, 최근 여러 논문에서 데이타 종속을 제거하기 위하여 결과 값을 예상하는 방법이 연구되고 있다. 혼합형 결과 값 예측기는 여러 예측기의 장점을 이용하여 높은 예상 정확도를 얻을 수 있지만, 동일한 명령어가 여러 개의 예측기 테이블에 중복 엔트리를 갖게되어 높은 하드웨어의 비용을 필요로 한다는 단점이 있다. 본 논문에서는 정적 및 동적 분류 정보를 이용하여 높은 성능을 얻을 수 있는 새로운 혼합형 결과 값 예측기를 제안한다. 제안된 예측기는 반입 단계 동안 정적 분류 정보를 사용하여 적절한 예측기에 할당함으로써 테이블 크기를 효과적으로 감소시켰고 예상정확도를 향상시켰다. 또한 제안된 예측기는 동적 분류를 사용하여“Unknown”유형의 명령어에 가장 적절한 예측방법을 선택하도록 하여 예상 정확도를 더욱 향상시켰다. SimpleScaiar/PISA 툴셋과 SPECint95 벤치마크 프로그램에서 시뮬레이션 한 결과, 정적 분류 정보를 사용하였을 경우 평균 예상 정확도가 85.1%, 정적 및 동적 분류 정보를 모두 사용하였을 경우 87.6%의 평균 예상 정확도를 얻을 수 있었다.

슈퍼스칼라 프로세서에서 동적 분류 능력을 갖는 혼합형 데이타 값 예측기의 설계 (Design of a Hybrid Data Value Predictor with Dynamic Classification Capability in Superscalar Processors)

  • 박희룡;이상정
    • 한국정보과학회논문지:시스템및이론
    • /
    • 제27권8호
    • /
    • pp.741-751
    • /
    • 2000
  • 슈퍼스칼라 프로세서에서 명령어 수준 병렬성(Instruction Level Parallelism)을 적극적으로 활용하기 위해서는 명령들 사이에 존재하는 제어 종속관계 및 데이타 종속관계를 극복하는 것이 필수적이다. 데이타 값 예측은 하나의 명령 결과가 생성되기 전에 미리 결과 값을 예측하고 이 예측된 결과를 사용하여 데이타 종속관계가 있는 명령들을 투기적으로 실행(speculative execution)하는 기법이다. 본 논문에서는 동적 분류 능력을 갖는 혼합형 데이타 값 예측기를 제안한다. 제안된 예측기는 최근 값 예측기, 스트라이드 예측기 및 2 단계 예측기를 결합한 혼합형으로 구성되며, 예측되는 명령은 하드웨어에 의한 동적 분류에 의해 각 예측기로 할당된다. 각 명령들의 특성에 따라 각 예측기로 실행 시에 동적 분류됨으로써 각 예측기는 기존의 혼합형 방식보다도 더욱 효과적으로 활용될 수 있다. 제안된 방식의 타당성 검증을 위해 실행구동방식(execution-driven) 시뮬레이터를 사용하여 SPECint95 벤치마크를 시뮬레이션하여 비교한다. 실험 결과 Instruction Per Cycle 비교실험에서 2 단계 예측기 보다 0.36, 혼합형 예측기 보다 0.0l8의 성능을 보였고, 제안된 방식이 기존의 혼합형 방식보다 예측 정확도가 평균 16%가 향상되었고, 하드웨어 비용을 측정한 결과 45%의 감소효과를 얻었다.

  • PDF

Using Machine Learning Algorithms for Housing Price Prediction: The Case of Islamabad Housing Data

  • Imran, Imran;Zaman, Umar;Waqar, Muhammad;Zaman, Atif
    • Soft Computing and Machine Intelligence
    • /
    • 제1권1호
    • /
    • pp.11-23
    • /
    • 2021
  • House price prediction is a significant financial decision for individuals working in the housing market as well as for potential buyers. From investment to buying a house for residence, a person investing in the housing market is interested in the potential gain. This paper presents machine learning algorithms to develop intelligent regressions models for House price prediction. The proposed research methodology consists of four stages, namely Data Collection, Pre Processing the data collected and transforming it to the best format, developing intelligent models using machine learning algorithms, training, testing, and validating the model on house prices of the housing market in the Capital, Islamabad. The data used for model validation and testing is the asking price from online property stores, which provide a reasonable estimate of the city housing market. The prediction model can significantly assist in the prediction of future housing prices in Pakistan. The regression results are encouraging and give promising directions for future prediction work on the collected dataset.

전자상거래에서 지식탐사기법의 활용에 관한 연구 (An Application of Data Mining Techniques in Electronic Commerce)

  • 성태경;주석진;김중한;홍준석
    • 한국정보시스템학회지:정보시스템연구
    • /
    • 제14권2호
    • /
    • pp.277-292
    • /
    • 2005
  • This paper uses a data mining approach to develop bankruptcy prediction models suitable for traditional (off-line) companies and electronic (on-line) companies. It observes the differences in the composition prediction models between these two types of companies and provides interpretation of bankruptcy classifications. The bankruptcy prediction models revealed the major variables in predicting bankruptcy to be 'cash flow to total assets' and 'gross value-added to net sales' for traditional off-line companies while 'cash flow to liabilities','gross value-added to net sales', and 'current ratio' for electronic companies. The accuracy rates of final prediction models for traditional off-line and electronic companies were found to be $84.7\%\;and\;82.4\%$, respectively. When the model for traditional off-line companies was applied for electronic companies, prediction accuracy dropped significantly in the case of bankruptcy classification (from $70.4\%\;to\;45.2\%$) at the level of a blind guess ($41.30\%$). Therefore, the need for different models for traditional off-line and electronic companies is justified.

  • PDF

LSTM 인공신경망을 이용한 자동차 A/S센터 수리 부품 수요 예측 모델 연구 (A Study on the Demand Prediction Model for Repair Parts of Automotive After-sales Service Center Using LSTM Artificial Neural Network)

  • 정동균;박영식
    • 한국정보시스템학회지:정보시스템연구
    • /
    • 제31권3호
    • /
    • pp.197-220
    • /
    • 2022
  • Purpose The purpose of this study is to identifies the demand pattern categorization of repair parts of Automotive After-sales Service(A/S) and proposes a demand prediction model for Auto repair parts using Long Short-Term Memory (LSTM) of artificial neural networks (ANN). The optimal parts inventory quantity prediction model is implemented by applying daily, weekly, and monthly the parts demand data to the LSTM model for the Lumpy demand which is irregularly in a specific period among repair parts of the Automotive A/S service. Design/methodology/approach This study classified the four demand pattern categorization with 2 years demand time-series data of repair parts according to the Average demand interval(ADI) and coefficient of variation (CV2) of demand size. Of the 16,295 parts in the A/S service shop studied, 96.5% had a Lumpy demand pattern that large quantities occurred at a specific period. lumpy demand pattern's repair parts in the last three years is predicted by applying them to the LSTM for daily, weekly, and monthly time-series data. as the model prediction performance evaluation index, MAPE, RMSE, and RMSLE that can measure the error between the predicted value and the actual value were used. Findings As a result of this study, Daily time-series data were excellently predicted as indicators with the lowest MAPE, RMSE, and RMSLE values, followed by Weekly and Monthly time-series data. This is due to the decrease in training data for Weekly and Monthly. even if the demand period is extended to get the training data, the prediction performance is still low due to the discontinuation of current vehicle models and the use of alternative parts that they are contributed to no more demand. Therefore, sufficient training data is important, but the selection of the prediction demand period is also a critical factor.

CNN-LSTM Coupled Model for Prediction of Waterworks Operation Data

  • Cao, Kerang;Kim, Hangyung;Hwang, Chulhyun;Jung, Hoekyung
    • Journal of Information Processing Systems
    • /
    • 제14권6호
    • /
    • pp.1508-1520
    • /
    • 2018
  • In this paper, we propose an improved model to provide users with a better long-term prediction of waterworks operation data. The existing prediction models have been studied in various types of models such as multiple linear regression model while considering time, days and seasonal characteristics. But the existing model shows the rate of prediction for demand fluctuation and long-term prediction is insufficient. Particularly in the deep running model, the long-short-term memory (LSTM) model has been applied to predict data of water purification plant because its time series prediction is highly reliable. However, it is necessary to reflect the correlation among various related factors, and a supplementary model is needed to improve the long-term predictability. In this paper, convolutional neural network (CNN) model is introduced to select various input variables that have a necessary correlation and to improve long term prediction rate, thus increasing the prediction rate through the LSTM predictive value and the combined structure. In addition, a multiple linear regression model is applied to compile the predicted data of CNN and LSTM, which then confirms the data as the final predicted outcome.

한국형수치예보모델 자료동화에서 위성 복사자료 관측오차 진단 및 영향 평가 (Diagnostics of Observation Error of Satellite Radiance Data in Korean Integrated Model (KIM) Data Assimilation System)

  • 김혜영;강전호;권인혁
    • 대기
    • /
    • 제32권4호
    • /
    • pp.263-276
    • /
    • 2022
  • The observation error of satellite radiation data that assimilated into the Korean Integrated Model (KIM) was diagnosed by applying the Hollingsworth and Lönnberg and Desrozier techniques commonly used. The magnitude and correlation of the observation error, and the degree of contribution for the satellite radiance data were calculated. The observation errors of the similar device, such as Advanced Technology Microwave Sounder (ATMS) and Advanced Microwave Sounding Unit-A shows different characteristics. The model resolution accounts for only 1% of the observation error, and seasonal variation is not significant factor, either. The observation error used in the KIM is amplified by 3-8 times compared to the diagnosed value or standard deviation of first-guess departures. The new inflation value was calculated based on the correlation between channels and the ratio of background error and observation error. As a result of performing the model sensitivity evaluation by applying the newly inflated observation error of ATMS, the error of temperature and water vapor analysis field were decreased. And temperature and water vapor forecast field have been significantly improved, so the accuracy of precipitation prediction has also been increased by 1.7% on average in Asia especially.

머신러닝 기반 가치투자를 통한 주식 종목 선정 연구: 내재가치를 중심으로 (Selecting Stock by Value Investing based on Machine Learning: Focusing on Intrinsic Value)

  • 김윤승;유동희
    • 한국정보시스템학회지:정보시스템연구
    • /
    • 제32권1호
    • /
    • pp.179-199
    • /
    • 2023
  • Purpose This study builds a prediction model to find stocks that can reach intrinsic value among KOSPI and KOSDAQ-listed companies to improve the stability and profitability of the stock investment. And investment simulations are conducted to verify whether stock investment performance is improved by comparing the prediction model, random stock selection, and the market indexes. Design/methodology/approach Value investment theory and machine learning techniques are applied to build the model. Various experiments find conditions such as the algorithm with the best predictive performance, learning period, and intrinsic value-reaching period. This study selects stocks through the prediction model learned with inventive variables, does not limit the holding period after buying to reach the intrinsic value of the stocks, and targets all KOSPI and KOSDAQ companies. The stock and financial data are collected for 21 years (2001-2021). Findings As a result of the experiment, using the random forest technique, the prediction model's performance was the best with one year of learning period and within one year of the intrinsic value reaching period. As a result of the investment simulation, the cumulative return of the prediction model was up to 1.68 times higher than the random stock selection and 17 times higher than the KOSPI index. The usefulness of the prediction model was confirmed in that the number of intrinsic values reaching the predicted stock was up to 70% higher than the random selection.

연간 건물난방 에너지사용량의 예측에 미치는 측정기간의 영향 (Effect of Measuring Period on Predicting the Annual Heating Energy Consumption for Building)

  • 조성환;태춘섭;김진호;방기영
    • 설비공학논문집
    • /
    • 제15권4호
    • /
    • pp.287-293
    • /
    • 2003
  • This study examined the temperature-dependent regression model of energy consumption based on various measuring period. The methodology employed was to construct temperature-dependent linear regression model of daily energy consumption from one day to three months data-sets and to compare the annual heating energy consumption predicted by these models with actual annual heating energy consumption. Heating energy consumption from a building in Daejon was examined experimentally. From the results, predicted value based on one day experimental data can have error over 100%. But predicted value based on one week experimental data showed error over 30%. And predicted value based on over three months experimental data provides accurate prediction within 6% but it will be required very expensive.

역 s-순으로 스캔된 주변 픽셀들에 존재하는 유사성과 에지 특성을 이용한 효율적인 픽셀 값 예측 기법 (An Efficient Pixel Value Prediction Algorithm using the Similarity and Edge Characteristics Existing in Neighboring Pixels Scanned in Inverse s-order)

  • 정수목
    • 한국정보전자통신기술학회논문지
    • /
    • 제11권1호
    • /
    • pp.95-99
    • /
    • 2018
  • 본 논문에서는 영상에서 역 s-순으로 스캔된 주변 픽셀 값들을 이용하여 픽셀 값을 정밀하게 예측할 수 있는 효율적인 픽셀 값 예측 기법을 제안하였다. 영상에는 일반적으로 인접 픽셀 값들 사이에 비슷한 값을 갖는 유사성(similarity)이 존재하고, 방향성이 있는 에지 특성(directional edge characteristics)이 존재할 수 있다. 인접 픽셀간의 유사성과 에지 특성을 이용하여 픽셀 값을 예측하는 GAP(Gradient Adjacent Pixel) 기법을 개선하여 픽셀 값 예측 정확도를 향상시키는 기법을 본 논문에서 제안하였다. 제안된 기법에서는 주변 픽셀들의 위치별 가중치를 사용하여 픽셀 값을 정밀하게 예측하도록 함으로 예측 픽셀 값의 정확도를 증가시켰다. 실제 영상에 대한 실험을 통하여 제안된 기법의 우수성을 확인하였다. 제안된 기법은 가역 데이터 은닉, 가역 워터마킹 및 데이터 압축 등의 응용들에 유용하게 사용될 수 있다.