Time series data is defined as data that has been consistently measured and organized by specific time intervals, referred to as the time series frequency. The data is collected into an order sequence that allows researchers to track changes as well as look for trends and patterns. Time series data is unique from other data types because the data points follow a natural order.
Time series data plays a role in several industries for record keeping, analysis, and forecasting. Prevalent examples include:
Weather Forecasting - Data about temperatures, humidity, precipitation, and other factors are measured frequently and stored for analysis. This past data aids meteorologists in interpreting current weather conditions and forecasting future weather patterns.
Inventory Management - Businesses track data about inventory, sales, and materials in order to understand how seasonal trends and other factors affect business. Having an accurate record of sales and inventory helps businesses recognize patterns and prepare for fluctuation in demand.
Machine Learning - Time series data is an important input for many machine learning models designed to recognize patterns or make forecasts. They use time series data to learn to interpret new inputs.
There are two types of time series data. Data points gathered at regular time intervals are called metrics. Daily weather conditions or hourly stock prices are examples of metric data points. The predictability of metrics make them reliable tools for analysis and forecasting. Metric data is required for building predictive models.
Data points collected at irregular intervals are called events. Rather than being measured at regular intervals, event data is collected only when new data is generated. While these events are irregular and sometimes unpredictable, they are still time series data because they are measured and plotted in time order. Bank account transactions and computer log entries are examples of data events. While useful information may be learned from the analysis of events data, the unpredictable intervals mean it is not useful for forecasting or predictive modeling.
Additionally, time series data can be classified as linear or nonlinear. Linear times series data points can be defined as a combination of past and future values. This usually indicates a linear relationship between time and the measured variable. A nonlinear time series has data points that cannot be defined by linear processes and are much less predictable.
Time series data helps researchers learn from the past, recognize patterns, and predict the future. This makes it a valuable asset in understanding our world, improving processes, and developing new technologies. Time series data is also a tool for making informed decisions and predicting outcomes. Whether it’s preparing for the weather, investing money, monitoring health, or making business decisions, time series data plays a role in almost everyone’s life.
When determining data type, the main elements to look at are: how many values are being measured and whether they are being measured over time. Time series data will contain a single value measured repeatedly at regular time intervals. Cross-sectional data is data containing multiple values measured at just one point in time. Unlike in time series data, there is no natural order to the data points in cross-sectional data. If a dataset contains multiple values tracked at intervals across time this is referred to as panel data or pooled data.