Time series is a set of random variables in time, which is usually observed in a given sampling rate within the time period of equal intervals. result. Time series data is essentially reflected in some or some random variables that have changed over time, and the core of the time series prediction method is to excavate this law from the data, and use it to estimate the future data.
constituent elements: long-term trends, seasonal changes, cyclic fluctuations, irregular changes.
1) Long-term trend (T) phenomenon is a total change trend formed by a certain fundamental factor within a longer period of time.
2) Seasonal Change (S) The regular cyclical changes occurred in the year with the season in the year.
3) The phenomenon of cyclic variation (C) has a regular variation in wave undulating forms present in a period of several years.
4) Irregular Changes (I) is a variation in a regular basis, including strict variations in random variations and irregular erosive influences.
1. Can reflect the development of social economic phenomena, describe the development status and results of phenomena.
2. It can be studied the development trend and development speed of social economy.
3. You can explore the law of development and change, and predict some social and economic phenomena.
4. Using the time series can be compared between different regions or countries, which is also one of the important methods of statistical analysis.
(1) Absolute Time Sequence
1. Period Sequence: Arranged by the Total Total Indicators sequentially.
The main feature of the sequence of sequences is:
1) The specifier value in the sequence has addability.
2) The size of each indicator value in the sequence is directly related to the length of the period it reflects.
3) The value of each indicator in the sequence is usually achieved by continuous registration.
2. Time sequence: The main characteristics of the time series
of the time point of the time point
1) sequence The numeric values in the indicator are not additive.
2) The size of each indicator value is not directly contacted with the length of the interval time in the sequence.
3) The value of each indicator in the sequence is usually obtained by regular registration.
(b) relative time series
Time series, a series of relative comparison indicators, is called relative time. sequence.
(3) Average Time Series
Average Time Series refers to a time series arranged in sequential order by a series of similar average indicators.
1, the time series analysis is based on past change trends predicts the future development, it is the premise that the past continues to the future.
Time series analysis is based on the continuous regularity of the development of objective things, using past historical data, through statistical analysis, further speculating future development trends. The past will continue until the future, this hypothesis is included in the premise of the two layers; one is not sudden jump change, is a relatively small pace; two is the past and current phenomena may indicate the development of current and future activities. . This determines that in general, the time sequence analysis method is remarkable for short and recent forecasts, but if extended to the farther future, there will be a lot of limitations, resulting in the predicted value from being deviated from the actual and makes decision failures. .
2, the time series data variations There is a regularity and irregularity of the
time sequence, which is a variety of different factors affecting changes at the same time. Comprehensive result of the role. From the time characteristics of the size and direction change of these influencing factors, changes in the time series data caused by these factors are divided into four types.
(1) Trend: A variable progresses with time, and has a slow and long-term continuous rise, decline, and the changing, the change may not equal.
(2) Periodic: A factor is due to the external impact as the alternation of the natural season has the law of peak and low valleys.
(3) Randomity: Individual is a random change, and the overall statistical law.
(4) Synthetic: The actual variation is a superposition or combination of changes. It is expected that the irregular change is made when it is predicted, highlighting the trend and periodic changes.
to ensure the comparability of the value value in the sequence
(1) period is the best uniform
(two The overall range should be consistent
(3) The economic content of the indicator should be unified
(4) calculation method should be unified
(5) calculation price and measure unit Compared to
Non-stationary (Nonstationarity, also translated as uncomfortable, unstable): The immersed sequence variable cannot present a long period of time Trends and eventually tend to tend to constant or a linear function.
Volumex amplitude Change (Time-VaryingVolatility): That is, the variable variable variable variation changes over time makes effective analysis time sequence variables are difficult .
The stable time number (stationaryTimeSeries) means a time-to-list statistical characteristics will not change over time.
(1) Indicator Analysis
Through the analysis indicators of the time series, the level of development and development changes are revealed.
(b) Component Analysis method
By decomposing the constituent factors affecting the time series, discloses the law evolved with time change.
Time series combination model
1 addition model: y = t + s + c + i (y, T metrology unit is the same Metrics (s, c, i)
2 multiplication model: y = t · s · c · i (common model)
2 (frequencies) (Y, T) The total amount of the unit is the same as the total amount of the unit) (S, C, I increase or decrease)
time series prediction is mainly based on continuity principles . The continuity principle refers to the development of objective things. It has a consecutive continuity, and the development of things is carried out in accordance with its own law. Under certain conditions, the basic development trend of things will continue in the future as long as the conditions are regularly changed.
Time series prediction is to use statistical techniques and methods to identify evolution modes from predictive indicators, establish mathematical models, and quantitatively estimate the future development trend of forecast indicators.