sequentially
1. Definitions
Time series (or dynamic series) is the number of columns the number of the same time their statistical indicators of the occurrence of arrayed order. The main purpose of time series analysis is to predict the future based on existing historical data. Most economic data are given in the form of time series. Depending on the time of observation, time-series may be a year, quarter, month or any other time form. For example: Beijing monthly CPI year on year data. http://data.eastmoney.com/cjsj/cpi.html
Mother went to a specific definition of Austria.
2. components
Not so much the constituent elements, but rather that a time series can be decomposed into which
components: long-term trend, seasonal variation, circulation changes, irregular changes.
1) long-term trend (T) formed by the phenomenon of some fundamental factors of the overall change in trend in the longer term.
2) seasonal variation (S) phenomenon within a year of change with the seasons and cyclical changes that occur regularly.
3) cyclic variation (C) phenomena in years as the undulating shape of the cycle exhibited by a regular change.
4) fluctuate irregularly (I) is an erratic changes, including strict random and irregular sudden changes affect major changes in the two types.
Can be found through the components, time series can be used to analyze the development trend of things, according to predict trends, to provide a reference for the development of relevant policies related parts.
3. Combination Model
A combination of time series model There are two main forms of
the additive model: Y = T + S + C + I ( that is, adding several trends above)
multiplicative model: Y = T · S · C · I ( above several trends multiplied by the
specific use to which it is set according to the specific issues it
4. stationarity
Smooth definition: stability refers to all the statistical properties of the time series will not change over time and occurs
mainly in
Strictly stationary
Wide smooth
unstable
Wherein a stationary time series are mainly
(1) the mean and variance not change with time;
(2) the autocorrelation coefficient is only related to the time interval, which is independent of time.
Correlation coefficients are used to quantify the degree of correlation between variables. Autocorrelation coefficient study of the correlation coefficient is a sequence of different periods, which is a series of calculated time series correlation coefficient of the current period and different lag period.
The current mainstream of time series forecasting methods are carried out for stationary time series analysis, but in fact, most of the time sequence we encounter are not stable, so in the analysis, you need to first identify stationary sequence, and not the smooth sequence into a stationary series. A time series smoothing only be treated in order to be controlled and predicted.
The time series smoothing treatment
There are a lot of the time series smoothing the way, the method is based on differential and logarithmic law, because this approach helps us to interpret the time series model. Difference, refers to the difference between before and after the sequence of the adjacent two data. Logarithmic law is the logarithmic value of the time series. (This process basically two nonstationary time series), and there are some that wavelet analysis (here a link ow http://blog.sina.com.cn/s/blog_136949aa50102x750.html wavelet analysis) the weather changes with time series of more. There is a Kalman filter Han https://blog.csdn.net/yzxnuaa/article/details/79450182
in R language, the main function of the differential used in the diff ()
This time series is emmmm acquaintance, after detailed analysis of the specific model