Time Series Processing and Applications
introduce
In general, when doing data analysis, almost all tasks related to time series are encountered. For example: what will happen to a certain indicator of a certain company on the next day or every week, or every month; how many downloads of a certain application per day, etc. In this section of experiments, different approaches will be used to handle these time series related tasks.
In this section, it will also introduce how to use time series processing tools in Python, which methods and models can be used for forecasting, what is double exponential and triple exponential fitting, what to do if you encounter non-stationary data, and how to establish SARIMA model, how to use XGBoost for forecasting, etc. In addition, these algorithms will also be applied to data from actual collection this time.
knowledge points
- sequentially
- SARIMA
- linear regression
- XGBoost
- Prophet
Introduction to Time Series Data
Let's start with a simple *definition* of a time series . In general, a time series can be defined as follows:
Time series refers to a set of numerical sequences formed by arranging the values of the same statistical index according to the time sequence of their occurrence.
Therefore, time series data is usually time-related sequence data, and compared with random sample data, time series may contain some useful historical information.
In this experiment, the Statsmodels library is mainly used to process sequences. This library has many statistical functions, including functions for processing time series. Statsmodels will definitely feel more familiar to anyone coming to Python from R#