전이학습을 이용한 시계열 데이터의 결측치 대체와 예측 성능과의 상관성 분석 = A Transfer Learning for Missing Value Imputation and Its Relationship with Prediction Performance in Time Series Data

다국어 초록 (Multilingual Abstract)

Missing values incur the lack of data availability and/or inaccurate predictions in the problem of time series prediction. We consider a transfer learning method for missing data imputation in time series data and test two research hypothesis; the first hypothesis is that the high similarity between two time series, one containing missing values and the other used for transfer learning, improves the imputation performance. Second, a better imputation performance results in a better prediction accuracy. Empirical analysis reveals that the transfer learning with high similarity in two time series improves the imputation performance. As known in the literature, we found a positive correlation between imputation performance and prediction accuracy. However, the correlation between imputation performance and prediction accuracy becomes insignificant when the time series has low volatility and a short length of consecutive missing data. It means that a simple method for missing data imputation is preferred to an expensive but effective method such as transfer learning if the time series is highly stable and predictable.

참고문헌 (Reference)

1 박재현 ; 이진오 ; 강상조 ; 강민수, "결측치 처리: 어떤 방법이 최선인가?" 한국체육학회 44 (44): 385-398, 2005

2 Fawaz, H. I., "Transfer Learning for Time Series Classification" 1367-1376, 2018

3 Ma, J., "Transfer Learning for Long-interval Consecutive Missing Values Imputation Without External Features in Air Pollution Time Series" 44 : 101092-, 2020

4 Sanneh, J., "Spatiotemporal and Machine Learning-Based Time Series Assessment of Drinking Water Quality Complaints in New York City" 2021

5 Schnaars, S. P., "Situational Factors Affecting Forecast Accuracy" 21 (21): 290-297, 1984

6 Tian, Y., "Similarity-based Chained Transfer Learning for Energy Forecasting with Big Data" 7 : 139895-139908, 2019

7 Ramosaj, B., "On the Relation between Prediction and Imputation Accuracy under Missing Covariates" 24 (24): 386-, 2022

8 Yang, J., "Multistage Large Segment Imputation Framework Based on Deep Learning and Statistic Metrics"

9 Burgette, L. F., "Multiple Imputation for Missing Data Via Sequential Regression Trees" 172 (172): 1070-1076, 2010

10 Lin, W. -C., "Missing Value Imputation: A Review and Analysis of the Literature (2006–2017)" 53 (53): 1487-1509, 2020

11 Pati, S. K., "Missing Value Estimation for Microarray Data Through Cluster Analysis" 52 (52): 709-750, 2017

12 Yu, L., "Missing Data Preprocessing in Credit Classification: One-hot Encoding or Imputation?" 58 (58): 472-482, 2022

13 Zhang, Z., "Missing Data Imputation: Focusing on Single Imputation" 4 (4): 2016

14 Pan, R., "Missing Data Imputation by K Nearest Neighbours Based on Grey Relational Structure and Mutual Information" 43 (43): 614-632, 2015

15 Murphy, K. P., "Machine Learning: A Probabilistic Perspective" MIT press 2012

16 Le Morvan, M., "Linear Predictor on Linearly-generated Data with Missing Values: Non Consistency and Solutions" 3165-3174, 2020

17 Feng, R., "Imputation of Missing Well Log Data by Random Forest and Its Uncertainty Analysis" 152 : 104763-, 2021

18 Ma, J., "Improving Air Quality Prediction Accuracy at Larger Temporal Resolutions Using Deep Learning and Transfer Learning Techniques" 214 : 116885-, 2019

19 Türkmen, A. C., "Forecasting Intermittent and Sparse Time Series: A Unified Probabilistic Framework Via Deep Renewal Processes" 16 (16): e0259764-, 2021

20 Zhu, X., "Efficient Utilization of Missing Data in Cost-sensitive Learning" 33 (33): 2425-2436, 2019

21 Huang, J., "Cross-validation Based K Nearest Neighbor Imputation for Software Quality Datasets: An Empirical Study" 132 : 226-252, 2017

22 Chen, Y., "Cross-position Activity Recognition with Stratified Transfer Learning" 57 : 1-13, 2019

23 Cao, W., "Brits:Bidirectional Recurrent Imputation for Time Series" 31 : 2018

24 Weisberg, S., "Applied Linear Regression" John Wiley & Sons 2005

25 Ho, N., "Amic: An Adaptive Information Theoretic Method to Identify Multi-scale Temporal Correlations in Big Time Series Data" 7 (7): 128-146, 2019

26 Xia, J., "Adjusted Weight Voting Algorithm for Random Forests in Handling Missing Values" 69 : 52-60, 2017

27 Chen, Z., "A Transfer Learning-Based LSTM Strategy for Imputing Large-Scale Consecutive Missing Data and Its Application in a Water Quality Prediction System" 602 : 126573-, 2021

28 Emmanuel, T., "A Survey on Missing Data in Machine Learning" 8 (8): 1-37, 2021

29 Fu, T. -C., "A Review on Time Series Data Mining" 24 (24): 164-181, 2011

30 Rahman, G., "A Decision Tree-based Missing Value Imputation Technique for Data Pre-processing" 121 : 2011

31 Ma, J., "A Bi-directional Missing Data Imputation Scheme Based on LSTM and Transfer Learning for Building Energy Data" 216 : 109941-, 2020