scholarly journals Novel robust time series analysis for long-term and short-term prediction

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Hiroshi Okamura ◽  
Yutaka Osada ◽  
Shota Nishijima ◽  
Shinto Eguchi
Keyword(s):  
Robust Regression ◽  
Simulated Data ◽  
New Method ◽  
Short Term ◽  
Least Absolute Deviations ◽  
Term Prediction ◽  
Regression Approach ◽  
Short Term Prediction ◽  
Long Term Prediction

AbstractNonlinear phenomena are universal in ecology. However, their inference and prediction are generally difficult because of autocorrelation and outliers. A traditional least squares method for parameter estimation is capable of improving short-term prediction by estimating autocorrelation, whereas it has weakness to outliers and consequently worse long-term prediction. In contrast, a traditional robust regression approach, such as the least absolute deviations method, alleviates the influence of outliers and has potentially better long-term prediction, whereas it makes accurately estimating autocorrelation difficult and possibly leads to worse short-term prediction. We propose a new robust regression approach that estimates autocorrelation accurately and reduces the influence of outliers. We then compare the new method with the conventional least squares and least absolute deviations methods by using simulated data and real ecological data. Simulations and analysis of real data demonstrate that the new method generally has better long-term and short-term prediction ability for nonlinear estimation problems using spawner–recruitment data. The new method provides nearly unbiased autocorrelation even for highly contaminated simulated data with extreme outliers, whereas other methods fail to estimate autocorrelation accurately.

scholarly journals Novel Robust Time Series Analysis for Long-term and Short-term Prediction

2021 ◽  
Author(s):  
Hiroshi Okamura ◽  
Yutaka Osada ◽  
Shota Nishijima ◽  
Shinto Eguchi
Keyword(s):  
Robust Regression ◽  
Simulated Data ◽  
New Method ◽  
Short Term ◽  
Least Absolute Deviations ◽  
Term Prediction ◽  
Regression Approach ◽  
Short Term Prediction ◽  
Long Term Prediction

Abstract Nonlinear phenomena are universal in ecology. However, their inference and prediction are generally difficult because of autocorrelation and outliers. A traditional least squares method for parameter estimation is capable of improving short-term prediction by estimating autocorrelation, whereas it has weakness to outliers and consequently worse long-term prediction. In contrast, a traditional robust regression approach, such as the least absolute deviations method, alleviates the influence of outliers and has potentially better long-term prediction, whereas it makes accurately estimating autocorrelation difficult and possibly leads to worse short-term prediction. We propose a new robust regression approach that estimates autocorrelation accurately and reduces the influence of outliers. We then compare the new method with traditional least squares and least absolute deviations methods using simulated data and real ecological data. Simulations and analysis of real data demonstrate that the new method generally has better long-term and short-term prediction ability for nonlinear estimation problems using spawner--recruitment data. The new method provides nearly unbiased autocorrelation even for highly contaminated simulated data with extreme outliers, whereas the other methods fail to estimate autocorrelation accurately.

scholarly journals On Retrospecting Human Dynamics with Attention

2019 ◽  
Author(s):  
Minjing Dong ◽  
Chang Xu
Keyword(s):  
Human Motion ◽  
Motion Prediction ◽  
Sufficient Information ◽  
Short Term ◽  
Human Dynamics ◽  
Time Horizons ◽  
Term Prediction ◽  
Error Accumulation ◽  
Short Term Prediction

Deep recurrent neural networks have achieved impressive success in forecasting human motion with a sequence to sequence architecture. However, forecasting in longer time horizons often leads to implausible human poses or converges to mean poses, because of error accumulation and difficulties in keeping track of longerterm information. To address these challenges, we propose to retrospect human dynamics with attention. A retrospection module is designed upon RNN to regularly retrospect past frames and correct mistakes in time. This significantly improves the memory of RNN and provides sufficient information for the decoder networks to generate longer term prediction. Moreover, we present a spatial attention module to explore and exploit cooperation among joints in performing a particular motion. Residual connections are also included to guarantee the performance of short term prediction. We evaluate the proposed algorithm on the largest and most challenging Human 3.6M dataset in the field. Experimental results demonstrate the necessity of investigating motion prediction in a self audit manner and the effectiveness of the proposed algorithm in both short term and long term predictions.

scholarly journals Multistep-Ahead Air Passengers Traffic Prediction with Hybrid ARIMA-SVMs Models

2014 ◽  
Vol 2014 ◽  
pp. 1-14 ◽  
Author(s):  
Wei Ming ◽  
Yukun Bao ◽  
Zhongyi Hu ◽  
Tao Xiong
Keyword(s):  
Prediction Models ◽  
Nonlinear Modeling ◽  
Real Data ◽  
Data Preprocessing ◽  
Air Transportation ◽  
Data Sets ◽  
Term Prediction ◽  
Short Term Prediction ◽  
Long Term Prediction

The hybrid ARIMA-SVMs prediction models have been established recently, which take advantage of the unique strength of ARIMA and SVMs models in linear and nonlinear modeling, respectively. Built upon this hybrid ARIMA-SVMs models alike, this study goes further to extend them into the case of multistep-ahead prediction for air passengers traffic with the two most commonly used multistep-ahead prediction strategies, that is, iterated strategy and direct strategy. Additionally, the effectiveness of data preprocessing approaches, such as deseasonalization and detrending, is investigated and proofed along with the two strategies. Real data sets including four selected airlines’ monthly series were collected to justify the effectiveness of the proposed approach. Empirical results demonstrate that the direct strategy performs better than iterative one in long term prediction case while iterative one performs better in the case of short term prediction. Furthermore, both deseasonalization and detrending can significantly improve the prediction accuracy for both strategies, indicating the necessity of data preprocessing. As such, this study contributes as a full reference to the planners from air transportation industries on how to tackle multistep-ahead prediction tasks in the implementation of either prediction strategy.

scholarly journals Empirical regional models for the short-term forecast of <i>M3000F2</i> during not quiet geomagnetic conditions over Europe

2013 ◽  
Vol 31 (10) ◽  
pp. 1653-1671 ◽  
Author(s):  
M. Pietrella
Keyword(s):  
Correction Factor ◽  
Regional Model ◽  
Local Models ◽  
Short Term ◽  
European Area ◽  
Term Prediction ◽  
Starting Point ◽  
Short Term Forecasting ◽  
Long Term Prediction

Abstract. Twelve empirical local models have been developed for the long-term prediction of the ionospheric characteristic M3000F2, and then used as starting point for the development of a short-term forecasting empirical regional model of M3000F2 under not quiet geomagnetic conditions. Under the assumption that the monthly median measurements of M3000F2 are linearly correlated to the solar activity, a set of regression coefficients were calculated over 12 months and 24 h for each of 12 ionospheric observatories located in the European area, and then used for the long-term prediction of M3000F2 at each station under consideration. Based on the 12 long-term prediction empirical local models of M3000F2, an empirical regional model for the prediction of the monthly median field of M3000F2 over Europe (indicated as RM_M3000F2) was developed. Thanks to the IFELM_foF2 models, which are able to provide short-term forecasts of the critical frequency of the F2 layer (foF2STF) up to three hours in advance, it was possible to considerer the Brudley–Dudeney algorithm as a function of foF2STF to correct RM_M3000F2 and thus obtain an empirical regional model for the short-term forecasting of M3000F2 (indicated as RM_M3000F2_BD) up to three hours in advance under not quiet geomagnetic conditions. From the long-term predictions of M3000F2 provided by the IRI model, an empirical regional model for the forecast of the monthly median field of M3000F2 over Europe (indicated as IRI_RM_M3000F2) was derived. IRI_RM_M3000F2 predictions were modified with the Bradley–Dudeney correction factor, and another empirical regional model for the short-term forecasting of M3000F2 (indicated as IRI_RM_M3000F2_BD) up to three hours ahead under not quiet geomagnetic conditions was obtained. The main results achieved comparing the performance of RM_M3000F2, RM_M3000F2_BD, IRI_RM_M3000F2, and IRI_RM_M3000F2_BD are (1) in the case of moderate geomagnetic activity, the Bradley–Dudeney correction factor does not improve significantly the predictions; (2) under disturbed geomagnetic conditions, the Bradley–Dudeney formula improves the predictions of RM_M3000F2 in the entire European area; (3) in the case of very disturbed geomagnetic conditions, the Bradley–Dudeney algorithm is very effective in improving the performance of IRI_RM_M3000F2; (4) under moderate geomagnetic conditions, the long-term prediction maps of M3000F2 generated by RM_M3000F2 can be considered as short-term forecasting maps providing very satisfactory results because quiet geomagnetic conditions are not so diverse from moderate geomagnetic conditions; (5) the forecasting maps originated by RM_M3000F2, RM_M3000F2_BD, and IRI_RM_M3000F2_BD show some regions where the forecasts are not satisfactory, but also wide sectors where the M3000F2 forecasts quite faithfully match the M3000F2 observations, and therefore RM_M3000F2, RM_M3000F2_BD, and IRI_RM_M3000F2_BD could be exploited to produce short-term forecasting maps of M3000F2 over Europe up to 3 h in advance.

Sign in / Sign up

Export Citation Format

Share Document