Notes on time-series analysis

Time series analysis

1. A very personal conclusion
1) Use time-series analysis not for prediction but to analyze relationships with other factors.

2) Treat time-series analysis not as prediction, but as a function with parameters and a return value (conditional computation).
-> Use the domain expert's know-how as parameters, and treat the computation as empirical evidence (the function's return value).

3) A reasonable scenario
   X - It will turn out like this
  O - If certain conditions (a person's experiential know-how, correlation analysis based on historical data) are met, ...it can turn out like this.

2. Background
1) Time-series analysis is called ARIMA. It's ARMA with 'I (Integrated, cumulative)' added.
2) Definitions ARIMA (Autoregressive Integrated Moving Average)
(1) Autoregressive refers to the autoregressive model, and Moving Average refers to the moving-average model.
(2) AR - Autoregression. A model where the error terms of earlier observations affect later observations.
I - Integrated. "Cumulative", meaning this is a label attached to time-series models that use differencing.
MA - Moving Average. A model where an observation is influenced by the series of previous error terms.
3) Origin?
...(continued)