Medium and long term predication method for necessities of life
Medium and long term predication method for necessities of life
 CN 103,617,468 A
 Filed: 12/13/2013
 Published: 03/05/2014
 Est. Priority Date: 12/13/2013
 Status: Active Application
First Claim
1. one way of life necessity mediumand longterm forecasting method, is characterized in that, key step is:
 (1), by commodity historical data input database storage, storage data analysis is arranged;
(2) build commodity historical data tendency trend map, judge whether to use described Forecasting Methodology to carry out future trend prediction;
Judge whether to exist exceptional value, if existed, after rejecting, again enter and judge whether to use described Forecasting Methodology to carry out future trend prediction;
(3) according to commodity historical data, build time series { Y _{t};
(4) determine the smoothing factor α
of each index;
(5) determine initial value with (6) each parameter of Comprehensive Model, builds linear trend predictive equation, and formula is as follows;
Chinese PRB Reexamination
Abstract
The invention discloses a medium and long term predication method for necessities of life. The medium and long term predication method comprises the following steps: firstly, collecting historical data of a commodity needing to be predicated to form a database; secondly, determining influence factors in an algorithm model by applying experiencebased judgment, trial and differenceratiomean value computing methods which are scientific and practical; eliminating random fluctuation in a historical statistical series to improve the accuracy of a predication result; then inputting data and outputting a result and forming a trending graph aiming at an obtained new number sequence to obtain a good fitting curve line and a predication result. The medium and long term predication method is simple to operate and low in cost; a formula is simple and easy to understand and can be applicable to nearly all detections taking a time sequence as a basis in all the fields; medium and long term market predication of the necessities of life, which has the development with an obvious tendency characteristic and less fluctuation, is processed well.

6 Citations
For the method and apparatus that prediction object is predicted  
Patent #
CN 106,156,875 A
Filed 04/10/2015

Current Assignee

A kind of power equipment warning information trend forecasting method based on exponential smoothing  
Patent #
CN 107,122,880 A
Filed 03/06/2017

Current Assignee

Based on the clock deviation forecasting procedure for improving exponential smoothing  
Patent #
CN 109,116,716 A
Filed 08/27/2018

Current Assignee

A kind of method for analyzing storage battery production situation  
Patent #
CN 107,886,212 A
Filed 09/25/2017

Current Assignee

A kind of metering table demand computational methods based on data analysis  
Patent #
CN 107,862,476 A
Filed 12/05/2017

Current Assignee

Method for acquiring user natural gas use behavior characteristics and device thereof  
Patent #
CN 104,951,991 A
Filed 03/28/2014

Current Assignee

No References
3 Claims

1. one way of life necessity mediumand longterm forecasting method, is characterized in that, key step is:

(1), by commodity historical data input database storage, storage data analysis is arranged; (2) build commodity historical data tendency trend map, judge whether to use described Forecasting Methodology to carry out future trend prediction;
Judge whether to exist exceptional value, if existed, after rejecting, again enter and judge whether to use described Forecasting Methodology to carry out future trend prediction;(3) according to commodity historical data, build time series { Y _{t};
(4) determine the smoothing factor α
of each index;(5) determine initial value with (6) each parameter of Comprehensive Model, builds linear trend predictive equation, and formula is as follows;


2. daily necessities mediumand longterm forecasting method as claimed in claim 1, is characterized in that:

For the value of smoothing factor α
, method is as follows;A) differenceratioaveraging method The method thinks, the size key of α
value depends on the magnitude of size that t issue itself changes, and specifically asks for step;1) according to time series y _{t}value obtain Δ
y _{t}=y _{t}y _{t1};
2) then according to the differentiated new sequence Δ
y of onelevel _{t}, obtain it is arithmetical mean;
3) use respectively again than upper Δ
y _{t}the value of each phase obtains new sequence { Δ
y _{t}'"'"';
4) to Δ
y _{t}'"'"' ask arithmetic mean, i.e. α
value comparatively accurately;
B) experience method This method mainly depends on seasonal effect in time series development trend and judges; 1) if the irregular fluctuation of time series is less, its basic trend is more stable, and these row α
should get smaller, makes forecast model adopt more valid data;
Moving little when time train wave, α
can be less, between (0.10.3), chooses;2) if there is larger variation in seasonal effect in time series basic trend, should get and make greatly model fast reaction speed, catch up with rapidly seasonal effect in time series and change;
When time train wave is moving larger, α
can be larger, and between (0.30.8), chooses;3) if initial value is to get arbitrarily surely, α
should get largerly, makes model through first few week after date, approaches rapidly real process.α
is larger, thinks more important in the recent period;
In reality, can get the value of several α
more and carry out tentative calculation, get α
that variance reckling is corresponding as weight;C) trial and error procedure According to concrete time series situation, with reference to experience method, roughly determine span, then to get several α
values and carry out tentative calculation, the prediction standard error under more different α
values, chooses the α
that predicts standard error minimum.


3. daily necessities mediumand longterm forecasting method as claimed in claim 1, is characterized in that:

with the big or small important of initial value to exponential smoothing value, determines algorithm initial value, and implementation method is as follows;
A) when time series item number is more than or equal to 15, initial value is relatively little on exponential smoothing impact, and at this moment can directly get first phase data is initial value, with little, desirable on predicted value impact ${S}_{0}^{\left(1\right)}={S}_{0}^{\left(2\right)}={y}_{1};$ B), when time series item number is less than 15, initial value is larger on the impact of exponential smoothing value, at this moment must conscientiously study, and conventionally gets the arithmetical mean of initial a few phase actual values as initial value;

Specification(s)