A survey of time series forecasting from stochastic method to soft computing

Abstract

Forecasting is a part of statistical modelling that is widely used in various fields because of its benefits in decision-making. The purpose of forecasting is to predict the future values of certain variables that vary with time using its previous values. Forecasting is related to the formation of models and methods that can be used to produce a good forecast. This research is a survey paper research that used a systematic mapping study and systematic literature review. Generally, time series research uses linear time series models, specifically the autoregressive integrated moving average model that has long been used because it has good forecasting accuracy. The successfulness of the Box–Jenkins methodology is based on the reality that various models can imitate the behaviour of various types of series, usually without requiring many parameters to be estimated in the final choice of the model. However, the assumption of stationarity that must be met makes this method inflexible to use. With the advances in computers, forecasting methods from stochastic models to soft computing continue to develop and extend. Soft computing for forecasting can provide more accurate results than traditional methods. Moreover, soft computing has many advantages in terms of the amount of data that can be analysed and the time- and cost-effectiveness of the process.

Short-term forecasting of confirmed daily COVID-19 cases in the Southern African Development Community region

“…, where represent the number of variables. The generalized additive model is then written as: (6) where is a constant parameter, are smooth functions and are independent and identically distributed ( ) error terms. Equation ( 1) is estimated using penalized cubic splines 22,23 given as: (7) The penalty parameter controls the degree of smoothness which is optimized using the generalized cross-validation criterion (GCV) 23 .…”

Traffic Flow Prediction for UTM Application: A Deep Learning Approach

“…A number of reviews in relation to time series algorithms deployed in the assessment of the utilities of varying domains have been published [10], [11], [12]. In the current work, linear stochastic paradigms for prediction have been widely used, e.g., autoregressive, moving average, autoregressive integrated moving average (ARIMA), seasonal ARIMA, autoregressive fractionally integrated moving average, autoregressive conditional heteroscedasticity (ARCH), and generalized ARCH.…”

A Systematic Review of Packages for Time Series Analysis

“…Time series analysis is a broad research field covering many application domains. The literature contains many reviews, either focusing on analysis tasks and methods (see, for instance, these reviews on forecasting [1][2][3][4], clustering and classification [5][6][7][8][9], anomaly detection [10][11][12], changepoint analysis [13][14][15], pattern recognition [16,17], or dimensionality reduction [18]) or focusing on a specific application domain (see, for instance, these surveys on finance [19], IoT and Industry 4.0 [20][21][22], or health [23]). Over time, several formal definitions and reviews of time series analysis tasks have been published; see, for example [24][25][26].…”