Current location - Education and Training Encyclopedia - Graduation thesis - What time series model should small data use?
What time series model should small data use?
Common time series models include ARIMA model, exponential smoothing model and trend model. For small sample size (sample number

Simple moving average model:

Formula:

Y _ t = (1/k) * (y _ {t-1}+y _ {t-2}+...+y _ {t-k}) where y _ t represents the observed value at t time point and k represents the size of the moving average window. Example:

Suppose you want to predict the daily temperature change in the next month, and you have collected the temperature data of the past 10 days, as shown below:

28, 29, 30, 3 1, 32, 34, 35, 33, 32, 3 1

If the moving average window size k = 3 is selected, the predicted temperature value of 1 1 day can be calculated:

y _ 1 1 =( 1/3)*(35+33+32)= 100/3≈33.33

Exponential smoothing model:

Formula:

Y _ t = α * y _ t-1+(1-α) * yhat _ t-1Here, y _ t represents the observed value at time t, and α is the smoothing constant (0

Example:

Suppose you want to predict the daily sales change of next month based on the sales data of the past 10 days, as follows:

100, 1 10, 105, 120, 1 15, 125, 130, 135, 140, 150

If the smoothing constant α = 0.2 is selected, the sales forecast value of 1 1 day can be calculated:

y _ 1 1 = 0.2 * 150+0.8 * 140 = 30+ 1 12 = 142