Time Series
What is a Time Series?
A time series is a sequence of data points ordered by time. Each data point represents an observation at a specific point or time period. A time series has a constant granularity, such as month, week, or day.
A time series has a start point (the first observation) and an end point (the last observation).
Time series can consist of various types of data, such as sensor measurements or monthly sales of a product. They serve as the basis for analysis in areas where patterns are sought and meaningful information is derived from historical data. Examples include fields like statistics, signal processing, financial mathematics, and weather forecasting.
A time series analysis often aims to make predictions about future developments based on given data. The methods used in such analysis are diverse and can differ significantly in approach. However, most methods share the assumption that data points closer in time are more strongly related to one another than those that are further apart.
Time series can also differ greatly in their structure, and as a result, they are classified into certain types. Not every method should or can be applied to all time series types. To ensure the highest prediction quality, the correct classification of the time series and the appropriate selection of the forecasting method are essential.
Different Types of Time Series
Time series can vary greatly in their qualities, and therefore, different techniques are required to model them. A simple classification is as follows:
Smooth Time Series:
Few to no values equal to 0, small coefficient of variation
Erratic Time Series:
Few to no values equal to 0, higher coefficient of variation
Sporadic/Intermittent Time Series:
Many values equal to 0, no clustering
Lumpy Time Series:
If non-zero values appear, they are clustered