Historical data is defined as “data that have already been collected and stored, typically in a database, and that are used as the basis for decision making.”1 In other words, historical data are past observations or measurements that can be used to inform present decisions.
Historical data can come from a variety of sources, including surveys, censuses, transactional records, and social media data.2 Text analytics often relies on historical data to train predictive models; for example, a model might be trained on historical customer service transcripts in order to predict future customer needs.3
Outside of the text analytics industry, the term “historical data” may be used in different ways. For example, financial analysts may use historical data to study market trends, while historians may use historical data to reconstruct past events.4 It is important to disambiguate the term “historical data” when using it in dialogue with others, in order to ensure that everyone is on the same page.
Historical Data vs. Other Similar Terms
The terms “historical data” and “past data” are often used interchangeably; however, there is a subtle distinction between the two. Past data are simply observations or measurements that have already occurred, while historical data are past observations or measurements that have been collected and stored for future use.5
The term “time series data” is also similar to “historical data,” but with a focus on the time element. Time series data are a sequence of observations or measurements that are taken over time, often at regular intervals.6 Historical data can be considered a type of time series data, but not all time series data are historical data.
Benefits of Historical Data
There are many benefits to using historical data, especially in the text analytics industry. First, historical data can provide context for present observations. For example, if a text analytics model is trained on historical customer service transcripts, it will be able to better understand the context of new customer service transcripts that it encounters. Second, historical data can be used to identify patterns and trends. For example, analysts may use historical financial data to identify market trends. Finally, historical data can be used to make predictions about future events. For example, a model trained on historical customer service transcripts might be able to predict future customer needs.