Time series forecasting with exogenous variable data
Providing time-series forecasting by receiving target variable data and exogenous variable data, training a plurality of time-series models according to the target variable data and the exogenous variable data, determining a historical error for each of the plurality of time series models, and providing a time-series forecasting model having a lowest historical error.
1 . A computer implemented method comprising:
receiving, by one or more computer processors, target variable data and exogenous variable data;
training, by the one or more computer processors, a differenced multi-variate linear regression (DMLR) model using the target variable data and the exogenous variable data:
training, by the one or more computer processors using differenced target variable and exogenous variable data sets, a base autoregression integrated moving average (ARIMA) model;
determining ARIMA error residuals, by the one or more computer processors using the base ARIMA model;
determining regression residuals, by the one or more computer processors using the ARIMA residuals and ARIMA coefficients;
training, by the one or more computer processors using the DMLR model and the regression residuals, a first portion of a plurality of time-series models according to the target variable data and the exogenous variable data;
determining, by the one or more computer processors, a historical error for each of the plurality of time series models;
providing, by the one or more computer processors, a time-series forecasting model having a lowest historical error; and
using, by the one or more computer processors the time-series forecasting model to forecast target variable time-series data for a user.
2 . The computer implemented method according to claim 1 , further comprising training, by the one or more computer processors using the ARIMA model and the regression residuals, a second portion of the plurality of time-series models according to the target variable data and the exogenous variable data, wherein at least one of the plurality of time-series models comprises a regression with the regression residuals added model.
3 . The computer implemented method according to claim 2 , wherein at least one of the second portion of the plurality of time-series models comprises a regression with the regression residuals subtracted model.
4 . The computer implemented method according to claim 2 , wherein at least one of second portion of the plurality of time-series models comprises a regression with the regression residuals adjusted and subtracted model.
5 . The computer implemented method according to claim 1 , wherein at least one of first portion of the plurality of time-series models comprises a prediction adjusted linear regression model.
6 . The computer implemented method according to claim 1 , wherein at least one of first portion of the plurality of time-series models comprises a training adjusted-linear regression model.
7 . The computer implemented method according to claim 1 , further comprising performing a multi-variate regression analysis of the target and exogenous variable data.
8 . A computer program product comprising one or more computer readable storage devices and collectively stored program instructions on the one or more computer readable storage devices, comprising program instructions which, when executed, cause a computing system to:
receive target variable data and exogenous variable data;
use the target variable data and the exogenous variable data to train a differenced multi-variate linear regression (DMLR) model:
use differenced target variable and exogenous variable data sets to train a base autoregression integrated moving average (ARIMA) model;
use the base ARIMA model to determine ARIMA error residuals;
use the ARIMA residuals and ARIMA coefficients to determine regression residuals;
train a first portion of a plurality of time-series models using the DLMR model and the regression residuals, according to the target variable data and the exogenous variable data;
determine a historical error for each of the plurality of time series models;
provide a time-series forecasting model having a lowest historical error; and
use the time-series forecasting model to forecast target variable time-series data for a user.
9 . The computer program product according to claim 8 , wherein the program instructions further cause the computing system to use the ARIMA model and the regression residuals to train a second portion of the plurality of time-series models according to the target variable data and the exogenous variable data, wherein at least one of second portion of the plurality of time-series models comprises a regression with regression residuals added model.
10 . The computer program product according to claim 9 , wherein at least one of the second portion of the plurality of time-series models comprises a regression with the regression residuals subtracted model.
11 . The computer program product according to claim 9 , wherein at least one of second portion of the plurality of time-series models comprises a regression with the regression residuals adjusted and subtracted model.
12 . The computer program product according to claim 8 , wherein at least one of first portion of the plurality of time-series models comprises a prediction adjusted linear regression model.
13 . The computer program product according to claim 8 , wherein at least one of first portion of the plurality of time-series models comprises a training adjusted-linear regression model.
14 . The computer program product according to claim 8 , wherein the program instructions further cause the computing system to perform a multi-variate regression analysis of the target and exogenous variable data.
15 . A computer system comprising:
one or more computer processors;
one or more computer readable storage devices; and
stored program instructions on the one or more computer readable storage devices for execution by the one or more computer processors, comprising stored program instructions which, when executed cause the computer system to:
receive target variable data and exogenous variable data;
use a differenced multi-variate linear regression to train a (DMLR) model using the target variable data and the exogenous variable data:
use differenced target variable and exogenous variable data sets to train a base autoregression integrated moving average (ARIMA) model;
use the base ARIMA model to determine ARIMA error residuals;
use the ARIMA residuals and ARIMA coefficients to determine regression residuals;
train a first portion of a plurality of time-series models using the DLMR model and the regression residuals, according to the target variable data and the exogenous variable data;
determine a historical error for each of the plurality of time series models;
provide a time-series forecasting model having a lowest historical error; and
use the time-series forecasting model to forecast target variable time-series data for a user.
16 . The computer system according to claim 15 , wherein the program instructions further cause the computing system to use the ARIMA model and the regression residuals to train a second portion of the plurality of time-series models according to the target variable data and the exogenous variable data, wherein at least one of second portion of the plurality of time-series models comprises a regression with regression residuals added model.
17 . The computer system according to claim 16 , wherein at least one of the second portion of the plurality of time-series models comprises a regression with the regression residuals subtracted model.
18 . The computer system according to claim 16 , wherein at least one of second portion of the plurality of time-series models comprises a regression with the regression residuals adjusted and subtracted model.
19 . The computer system according to claim 15 , wherein at least one of first portion of the plurality of time-series models comprises a prediction adjusted linear regression model.
20 . The computer system according to claim 15 , wherein at least one of first portion of the plurality of time-series models comprises a training adjusted-linear regression model.