There are four kinds of changing factors in the time series: ① Long-term trend (T), in which things show a general trend of increasing or decreasing gradually throughout the forecast period; (2) Periodic change (c), a periodic change with a certain time interval, such as the alternation of crisis and recovery; (3) Seasonal changes, that is, periodic changes with a one-year cycle, such as seasonal fluctuations in sales in the clothing industry; (4) Accidental changes (1), irregular changes except the above three situations, also known as random changes. The comprehensive modes of these four factors are addition mode, multiplication mode and mixed mode. If the time series is expressed as (= 1, 2, 3, ..., representing sampling time), the time series in addition mode is the sum of the above four variable factors =(T)+(C)+(S)+(I), while the time series in multiplication mode is the product of the above four variable factors = (t) × (c). Time series method can be divided into two categories: ① Using time series data directly to establish mathematical model and make prediction without subdividing four variables. (2) Pre-processing the four factors with emphasis, thus deducing specific forecasting methods such as eliminating seasonal variation, moving average, exponential smoothing, autoregressive, time function fitting, etc.