scholarly journals Causal Inference in Time Series via Supervised Learning

2018 ◽  
Author(s):  
Yoichi Chikahara ◽  
Akinori Fujino
Keyword(s):  
Time Series ◽  
Regression Model ◽  
Causal Inference ◽  
Supervised Learning ◽  
Regression Models ◽  
Multivariate Time Series ◽  
Feature Representation ◽  
Learning Method ◽  
Conditional Distributions ◽  
Learning Framework

Causal inference in time series is an important problem in many fields. Traditional methods use regression models for this problem. The inference accuracies of these methods depend greatly on whether or not the model can be well fitted to the data, and therefore we are required to select an appropriate regression model, which is difficult in practice. This paper proposes a supervised learning framework that utilizes a classifier instead of regression models. We present a feature representation that employs the distance between the conditional distributions given past variable values and show experimentally that the feature representation provides sufficiently different feature vectors for time series with different causal relationships. Furthermore, we extend our framework to multivariate time series and present experimental results where our method outperformed the model-based methods and the supervised learning method for i.i.d. data.

scholarly journals Real-Time Prediction of the COVID-19 Epidemic in Thailand using Simple Model-Free Method and Time Series Regression Model

2021 ◽  
Vol 18 (14) ◽  
Author(s):  
Rati WONGSATHAN
Keyword(s):  
Genetic Algorithm ◽  
Time Series ◽  
Regression Model ◽  
Regression Models ◽  
Gaussian Function ◽  
Logistic Function ◽  
Hyperbolic Tangent ◽  
Time Series Regression ◽  
Model Free ◽  
Tangent Function

The novel coronavirus 2019 (COVID-19) pandemic was declared a global health crisis. The real-time accurate and predictive model of the number of infected cases could help inform the government of providing medical assistance and public health decision-making. This work is to model the ongoing COVID-19 spread in Thailand during the 1st and 2nd phases of the pandemic using the simple but powerful method based on the model-free and time series regression models. By employing the curve fitting, the model-free method using the logistic function, hyperbolic tangent function, and Gaussian function was applied to predict the number of newly infected patients and accumulate the total number of cases, including peak and viral cessation (ending) date. Alternatively, with a significant time-lag of historical data input, the regression model predicts those parameters from 1-day-ahead to 1-month-ahead. To obtain optimal prediction models, the parameters of the model-free method are fine-tuned through the genetic algorithm, whereas the generalized least squares update the parameters of the regression model. Assuming the future trend continues to follow the past pattern, the expected total number of patients is approximately 2,689 - 3,000 cases. The estimated viral cessation dates are May 2, 2020 (using Gaussian function), May 4, 2020 (using a hyperbolic function), and June 5, 2020 (using a logistic function), whereas the peak time occurred on April 5, 2020. Moreover, the model-free method performs well for long-term prediction, whereas the regression model is suitable for short-term prediction. Furthermore, the performances of the regression models yield a highly accurate forecast with lower RMSE and higher R2 up to 1-week-ahead. HIGHLIGHTS COVID-19 model for Thailand during the first and second phases of the epidemic The model-free method using the logistic function, hyperbolic tangent function, and Gaussian function  applied to predict the basic measures of the outbreak Regression model predicts those measures from one-day-ahead to one-month-ahead The parameters of the model-free method are fine-tuned through the genetic algorithm  GRAPHICAL ABSTRACT

scholarly journals Verification of the Returns to Scale of Production Type for the Russian Federation Regions

2019 ◽  
Vol 224 ◽  
pp. 06011
Author(s):  
Igor Kirilyuk ◽  
Oleg Senko
Keyword(s):  
Time Series ◽  
Random Walk ◽  
Regression Model ◽  
Probability Distributions ◽  
Multivariate Time Series ◽  
Returns To Scale ◽  
Production Functions ◽  
Time Series Regression ◽  
Production Type ◽  
Statistical Validity

Monte-Carlo methods to asses a statistical validity of the relationship between coefficients of time series regression model were proposed. In economics such a relationship is present in the case when constant return to scale in production functions is assumed. The techniques being discussed here are virtually free from assumptions about underlying probability distributions and may be used in the case, when target variable or regressors are time series with random walk. This is achieved by comparing the regression model built on truly multivariate time series with those built on simulated time series with random walk. It has been shown that for the production functions of most Russian regions, the returns to scale significantly differs from a constant value at p<0.05.

Sign in / Sign up

Export Citation Format

Share Document