Forecasting Models – an Overview With The Help Of R Software

By Editor IJSMI

Forecasting models involves predicting the future values of a particular series of data which is mainly based on the time domain. Forecasting models are widely used in the fields such as financial markets, demand for a product and disease outbreak.  The objective of the forecasting model is to reduce the error in the forecasting.

Most of the Forecasting models are based on time series, a statistical concept which involves Moving Averages, Auto Regressive Integrated Moving Averages (ARIMA), Exponential smoothing and Generalized Auto Regressive Conditional Heteroscedastic (GARCH) Models. Forecasting models which we deal in this book will be explorative forecasting models which take into account the past data to predict the future values.

Current day forecasting models uses advanced techniques such as Machine Learning and Deep Learning Algorithms which are more robust and can handle high volume of data.

This book starts with the overview of forecasting and time series concepts and moves on to build forecasting models using different time series models. Examples related to forecasting models which are built based on Machine learning also covered. The book uses R statistical software package, an open source statistical package to build the forecasting models. 

• Language: English
• Publisher: IJSMI
• Release date: Jan 21, 2022
• ISBN: 9798201922948

Editor IJSMI

Editor, International Journal of Statistics and Medical Informatics www.ijsmi.com/book.php [e-mail]

Book preview

CHAPTER 1 – FORECASTING INTRODUCTION

Forecasting involves predicting the future value of a series of data wherein the series is mainly depends on the time domain. The time might be represented in terms of number of days, month, and year or even in hour or seconds in some cases. The basic method for forecasting includes the time series method and now currently advanced methods like neural networks; Bayesian and Deep Learning algorithms are used in forecasting due to the availability high end computer systems.  Forecasting is widely used in different fields such as finance, marketing, economics and health care.

Some of the examples which involve forecasting methodology are predicting the future of value of a stock price, demand for a particular product or service, predicting the disease outbreak and Macro Economic forecast such as Gross Domestic Product Forecast. The objective of the forecasting models is used to know the trend in the data set whether it is decreasing or increasing, predict the future value for a specific period of time and minimize the error in the predicted value compared to the actual value.

R an open source statistical software package is used in this book to build the forecasting models along with the Graphic User Interface R studio which is also open source software. R software includes various packages developed by the R community which helps to reduce the time of writing lengthy coding.

Readers are encouraged to refer to the author’s book on Introduction to Statistics for basic statistical concepts as the book assumes that the readers have basic understanding of statistical concepts.

CHAPTER 2 – TIME SERIES

Time series can be defined as a stochastic process which includes the value of a variable spread of over time.

2.1 Decomposition of the time series

Time series can be decomposed into the following components:

Trend component

Seasonal component

Cyclical component

Random or noise component.

Here the seasonal component is usually represented in terms of the four seasons or quarters.

2.2 Auto Correlation

Autocorrelation is a specific terminology widely used in the time series models is defined as the correlation present between the items of time series which are related in time i.e. correlation between current and past value(s) of the time series.

The previous value is termed as lag and if calculate the auto correlation between present value(t) and the immediate previous value (t-1), then the autocorrelation is of order 1 and the previous value is termed as lag 1.

2.3 Types of Time Series

Stationary Time Series

Stationary time series is defined by the characteristic that the mean and variance between of variables are same throughout the time period.

NON STATIONARY TIME Series

A series is defined as non-stationary time series if it includes trend or seasonal component and its mean and variance are not constant over the time period.

Differencing the time series

Differencing the time series helps us to make the non-stationary time series into stationary time series by calculating the difference between successive values which will enable us to remove the trend and seasonality present in the time series. Along with differencing, logarithmic transformation helps us to stabilize the time series variability.

2.4 Time series models

There are different types of Time series models are available which will be discussed and explained with the help of R statistical package in the following chapters of the book.  

Moving Averages

Exponential Smoothing

Auto Regressive Moving Averages (ARMA)

Box Jenkins Auto Regressive Integrated Moving Averages (ARIMA)

Auto Regressive Conditionally Heteroscedasticity (ARCH)

Generalized Auto Regressive Conditionally Heteroscedasticity (GARCH)

Vector Auto Regression (VAR)

Spectral Analysis

Kalman Filtering

Neural Network

Deep Learning Network

Bayesian Time series models

References

Brockwell, P. J., Davis, R. A., & Calder, M. V. (2002). Introduction to time series and forecasting (Vol. 2). New York: springer.

Makridakis, S., Wheelwright, S. C., & Hyndman, R. J. (2008). Forecasting methods and applications. John wiley & sons.

Granger, C. W. J., & Newbold, P. (2014). Forecasting economic time series. Academic Press.

Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: principles and practice. OTexts.

Nowotarski, J., & Weron, R. (2018). Recent advances in electricity price forecasting: A