Forecasting big time series

Abstract

Time series forecasting is a key ingredient in the automation and optimization of business processes: in retail, deciding which products to order and where to store them depends on the forecasts of future demand in different regions; in cloud computing, the estimated future usage of services and infrastructure components guides capacity planning; and workforce scheduling in warehouses, call centers, factories requires forecasts of the future workload. Recent years have witnessed a paradigm shift in forecasting techniques and applications, from computer-assisted model-and assumptionbased to data-driven and fully-automated. This shift can be attributed to the availability of large, rich, and diverse time series data sources, posing unprecedented challenges to traditional time series forecasting methods. As such, how can we build statistical models to efficiently and effectively learn to forecast from large and diverse data sources? How can we leverage the statistical power of "similar" time series to improve forecasts in the case of limited observations? What are the implications for building forecasting systems that can handle large data volumes? The objective of this tutorial is to provide a concise and intuitive overview of the most important methods and tools available for solving large-scale forecasting problems. We review the state of the art in three related fields: (1) classical modeling of time series, (2) scalable tensor methods, and (3) deep learning for forecasting. Further, we share lessons learned from building scalable forecasting systems. While our focus is on providing an intuitive overview of the methods and practical issues, we also present technical details underlying these powerful tools.

A Comparison of Feature Selection and Forecasting Machine Learning Algorithms for Predicting Glycaemia in Type 1 Diabetes Mellitus

“…Relationships are indicated by the coefficients and these are generally very simple to calculate, meaning that we can afford to employ a number of features. However, performance with these models may be limited [44]. Good results are achieved if the predictor/response relationship falls on a hyperplane.…”

A Hybridly Optimized LSTM‐Based Data Flow Prediction Model for Dependable Online Ticketing

“…The Light Gradient Boosting Machine (LightGBM), XGBoost, Random forest, support vector regression (SVR), and Seasonal Autoregressive Integrated Moving Average (SARIMA) algorithm are used to build models. These methods have been widely used in recent studies [55][56][57][58][59]. The parameter optimization is conducted for each algorithm model for building the optimal models.…”

A Methodology for Calculating the Contribution of Exogenous Variables to ARIMAX Predictions

“…Time series forecasting is at the core of many business processes to enable data-driven decision-making [1,2]. Among the existing time series forecasting methods, autoregressive integrated moving average with exogenous variables (ARIMAX) has become of the most popular ones in use [3].…”

“…Fig.1depicts the graph of |H(e jω )|. When r 1 and r 2 are real numbers (ϕ2 1 + 4ϕ 2 ≤ 0), its amplitude-frequency response behaves like just the AR(1) process, only displaying peaks at ω = 0 and/or ω = π, as shown by the orange, green and red lines. The blue line shows peak at ω 0 = arccos(ϕ 1 /2 |ϕ 2 |) when the two roots are complex numbers.…”