What is demand forecasting (Forecasting: Principles and practice Chapter 1)

1. getting started
   1.1 what can be forecast?
   The predictability of an event or quantity depends on
2. How well do we understand the factors that cause this situation
3. How much data is available
4. Predict whether it will affect what we want to predict

1.2
Forecasting , planning and goals Forecasting: All available information, including historical data and any understanding of future events that may affect forecasting, are designed to accurately predict future
goals: what is to happen, goals should be related to forecasting and planning , But not always. When setting goals, there are often no plans to achieve them, nor do we predict whether they will be achieved.
Planning: Response to forecasts and goals; planning involves determining the appropriate actions needed to bring forecasts to goals

Short-term forecast: Arrange personnel, production and transportation required events
Medium-term forecast: need to determine future resource needs
Long-term forecast: used for strategic planning, need to consider market opportunities, environmental factors and internal resources.

1.3
Determining what to forecast Early in the forecast , the forecast content needs to be determined:

1. Predict whether each product line or product group
2. Predict whether it is for each sales outlet, or for regional sales outlets, or for total sales
3. Whether the forecast is for weekly data, monthly data or annual data

1.4 forecasting data and methods
Appropriate forecasting methods largely depend on the available data

If no data is available, or if the available data is not relevant to the forecast, then use qualitative forecast

When two conditions are met, quantitative predictions can be used:

1. Provide digital information about the past
2. It is reasonable to assume that certain aspects of past patterns will continue to arrive

Time series forecasting: The simplest time forecasting method uses only relevant variable information to make predictions without attempting to discover factors that affect its behavior; therefore, they will infer trends and seasonal patterns, but ignore all other information.

The models used for prediction include: decomposition model, exponential smoothing model and ARIMA model.

Forecast Variables and Time Series Forecasting
Forecast variables are often useful in time series forecasting; if you build a model for electricity forecasting

There are always changes in electricity demand in the forecast, and the forecast variables cannot be explained; the 'error' on the right allows random changes and the influence of related variables not included in the model.
Since power demand data forms a time series, we can use time series models to make predictions. At this time, the appropriate time series prediction equation is:

Here, future predictions are based on the past values ​​of variables, not on external variables that may affect the system.
There is a third model that combines the functions of the above two models:

these types of hybrid models can be called dynamic regression models, panel data models, longitudinal models, transfer function models, and linear models.

1.6 The basic steps in a forecasting task
forecast usually involves 5 basic steps:
1)
Definition of the problem To define the problem carefully, you need to understand the way of using forecasting, the person who needs to forecast and how the forecasting function is suitable for the organization that needs to forecast.
2) Collecting information
At least two kinds of information are always required
a) Statistical data
b) Accumulation of expertise for those who collect data and use predictions
3) Preliminary exploratory analysis
Starting with drawing data charts, is there a consistent pattern of exploration? Is there a clear trend? Does seasonality matter?
4) Choosing and fitting the model
The best model to use depends on the availability of historical data, the strength of the relationship between the predictor variable and any explanatory variables, and how the forecast is used.
5) Use and evaluate the prediction model After
selecting the model and estimating its parameters, use the model to make predictions

1.7 the statistical forecasting perspective
In most forecasting situations, as events approach, changes related to our predictions will shrink; in other words, the farther we predict, the more uncertain we will be.

Prediction interval: A series of values ​​that a random variable can obtain with a relatively high probability; for example, a 95% prediction interval contains a series of values, which should include the actual future value with a probability of 95%.