What do you think?
Missing Values in Time Series Forecasting
A sequential set of data measured over time is referred to as Time series. Time series may be found in various fields, from engineering to economics. The study of time series data is a crucial subject, especially in statistics. Records of data recorded over a long period are likely to have missing observations. It happens when an observation cannot be performed at a certain moment due to malfunctioning equipment, lost records, or an error that cannot be corrected later. When one or more observations are missing, estimating the model and missing value analysis may be essential. With more precise forecasting, a greater understanding of the nature of the data may be acquired by providing approximations of missing variables. Different series may need different methodologies to estimate and generate the greatest possible estimations; it is vital to use these tactics properly. Having incomplete data is an unavoidable process. Data researchers put considerable effort, resources, and time when dealing with missing data in a project or dataset to ensure its accuracy. Despite their efforts, there is always a great possibility of data going missing. Scientists and data analytics consider missing data as one of the greatest obstacles impeding the successful accomplishment of an accurate research project. This thesis will investigate and evaluate the efficacy of several methods for estimating missing values in time series data models.
For univariate data, the method of predicting missing values in time series data includes analysis and modeling. Traditional time series analysis is often used to scalar-valued data and may be represented by autoregressive, moving average, or autoregressive-moving average models developed by Box-Jenkins. In certain circumstances, using an interpolation strategy helps in handling missing data. Time series decomposition, least-squares approximation, and numerical interpolation techniques are other prevalent methodologies for analyzing relational time series data with missing values. Some of these methods are fast to implement, and the model-building processes are simple to follow. These methodologies are often popular among specialists in the time series analysis industry.
Missing data reduces statistical power. Because some missing data is to be anticipated, the desired sample size is adjusted to accommodate it. However, this does not remove the possibility of errors. More attention should be given to missing data in the planning and execution of research and analyzing the resultant data. This study will examine how the existing methods of handling missing values in normal time series forecasting can be used in relational time series forecasting. To further understand this topic, the study will explain and describe several previous studies about handling missing values on time series data. In addition, this study discusses various reasonable strategies for estimating missing data that other researchers in this area might employ.
For univariate data, the method of predicting missing values in time series data includes analysis and modeling. Traditional time series analysis is often used to scalar-valued data and may be represented by autoregressive, moving average, or autoregressive-moving average models developed by Box-Jenkins. In certain circumstances, using an interpolation strategy helps in handling missing data. Time series decomposition, least-squares approximation, and numerical interpolation techniques are other prevalent methodologies for analyzing relational time series data with missing values. Some of these methods are fast to implement, and the model-building processes are simple to follow. These methodologies are often popular among specialists in the time series analysis industry.
Missing data reduces statistical power. Because some missing data is to be anticipated, the desired sample size is adjusted to accommodate it. However, this does not remove the possibility of errors. More attention should be given to missing data in the planning and execution of research and analyzing the resultant data. This study will examine how the existing methods of handling missing values in normal time series forecasting can be used in relational time series forecasting. To further understand this topic, the study will explain and describe several previous studies about handling missing values on time series data. In addition, this study discusses various reasonable strategies for estimating missing data that other researchers in this area might employ.
197 pages, Kindle Edition
Published January 24, 2025
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
No one has reviewed this book yet.