A full set of operations research knowledge points
Chapter 2 Forecast
1. Forecast
Prediction is to estimate or judge uncertain events in the future . Forecasting is the basis of decision-making.
2. Prediction methods and classification
- Predicted classification (content):
(1) Economic forecast : divided into macro- economic forecasting and micro- economic forecasting
macroeconomics: refers to the entire national economy forecast to a range of
micro-economic forecast: refers to a single economic entity (company) of The various indicators and the prediction of the domestic and foreign market economic forms involved
(2) Military prediction : study issues related to war and military affairs
(3) Science and technology prediction : divided into scientific prediction and technical prediction
(4) Social prediction : research society Development-related issues, such as population growth forecasts, social buying psychology forecasts, etc.
Social Science Military Economics
- Prediction methods
(1) Qualitative prediction : Using intuitive materials, relying on the subjective judgment and analytical ability of personal experience, to predict the future development, also known as intuitive prediction , mainly includes the expert group method and the Delphi method .
(2) Quantitative prediction : According to historical data and information, apply mathematical statistical methods or use the causal relationship of the development of things to predict the future of things.
The use of historical data to predict is called extrapolation , and time series analysis is commonly used ;
use the causality of the development of things to predict the future of things. Using the causal relationship of the internal factors of things to predict is called the causal method , and the commonly used methods include regression analysis, econometrics, and input-output analysis . - Classification from the forecast time period:
(1) Long-term forecasts
(2) Mid-term forecasts
(3) Short-term forecasts (also called short-term forecasts)
Generally speaking:
economic forecasts are 3-5 years as long-term, 1-3 years as medium-term, and within the year as short-term
Science and technology forecasts 30-50 years as long-term, 10-30 years as medium-term, and 5-10 years as short-term
Short section chief (1, 3, 5 , 10, 30, 50)
Third, the forecasting procedure
- Determine the predicted object or target
- Choose the forecast period.
Long-term forecasts are suitable for product varieties, and the specifications do not change much in a long period of time (such as grain, gasoline). The product life cycle is relatively long or the company
enjoys a longer patent period for the product. - Choose a forecasting method
- Collect relevant information
- Make predictions . Including regular forecasts and quantitative forecasts
Four, qualitative forecasting method
The qualitative forecasting method is also called the judging forecasting method.
Application
(1) The establishment of a model lacks data or information, such as predicting the price of a new product.
(2) The social environment or economic environment has undergone drastic changes, and historical data is no longer representative
Five, the Delphi method
Delphi method : also known as letter inquiry investigation method. Hope to achieve more consensus in the "expert group" method
features:
(1) expert opinions are anonymous, receiving face consultation or inquiry by letter between a back to back in
(2) multiple feedback
(3) Finally, the investigators sorted and summarized the expert opinions, and handed over the more unified and special opinions to the relevant departments for decision-making.
Steps:
(1) Determine the topic
(2) Select experts ( Experts here do not refer to scholars, professors, senior engineers, etc., but also people familiar with forecasting topics)
(3) Design consultation form
(4) Round-by-round consultation and information Feedback
(5) Use statistical analysis methods to quantitatively evaluate and describe the prediction results.
This method requires several rounds of information feedback, and the event is bound to be relatively long, so it is used for long-term or mid-term prediction . In addition, experts should explain the significance of the investigation in advance and pay them so that they can fill in the consultation form carefully. The
Delphi method takes a long time and is suitable for long-term or medium-term forecasts.
Six, expert group method
Definition: Form a group of experts who are consulted to discuss and negotiate face-to-face. In the end, a relatively unanimous opinion was reached on the topics that need to be predicted.
Advantages : Mutual negotiation can be achieved, and mutual supplementation can be achieved.
Disadvantages : When the group meeting is not well organized, it may cause the authority to influence the venue or the opinions of the majority to annihilate the innovative insights of the minority. The
prediction process of this method is relatively tight. So suitable for short-term forecasting
Seven, time series forecasting method
Time series forecasting method: It is to predict the development trend of things based on the historical data of the forecast object and use mathematical statistics to process it.
It is a set of data series arranged in chronological order of historical data and processed by mathematical statistics to predict the development trend of things.
Basic principles:
(1) Recognize the continuity of the development of things . But the accuracy is poor , and it is generally only suitable for short-term forecasts.
(2) Consider the influence and interference of random factors in the development of things.
The composition of time series is very complicated, and can be roughly divided into: long-term trends, seasonal fluctuations, periodic fluctuations and random fluctuations
8. Moving average forecasting method
Moving average forecasting method: divided into simple average forecasting method and weighted average forecasting method
- The simple moving average forecasting method is
actually an arithmetic average forecasting method. Its formula is:
(1) Horizontal comparison method : compare yourself with others at the same time
The month does not appear horizontally from front to back
We can use the average of the peers as our reference price
(2) Longitudinal comparison method : simple moving average method. The example of using the longitudinal comparison method to find the arithmetic mean is one of the simplest time series forecasting methods.
From back to front, the month will appear horizontally, vertically and vertically.
- Weighted average prediction method
According to the different proportions of different values, the corresponding weight can be added to the simple moving average prediction method. The
weighted average calculation formula is:
Nine, exponential smoothing forecasting method
The formula of the exponential smoothing prediction method is:
The value range of a is generally: 0<=a<=1;
when we find that the predicted value of period t has a large error with the actual value, we can increase the value of the smoothing coefficient a,
If the error is not large, a can be smaller.
In special circumstances, that is, when the price of the commodity is bullish or bearish, a can be a number greater than 1.
Doshisha University Doshisha University
10. Regression model prediction method
Regression analysis method is to predict the future development trend of things based on the causality of changes in the internal factors of the development of things. It is a quantitative prediction method that studies the relationship between variables, also known as regression model prediction method or causality method.
- Linear regression equation classification (the independent variable in the variable has a simple relationship with the dependent variable, but according to the number of variables, it can be divided into:)
(1) One-variable linear regression: a regression equation with an indirect relationship between an independent variable and a dependent variable
(2) Multiple Linear regression: the regression equation of the indirect relationship between multiple independent variables and a dependent variable
(3) Non-linear regression: the relationship between variables is not linear but nonlinear - One-variable linear regression model prediction method
Set the regression equation: y=a+bx
The basic idea of using a one-variable linear regression model is to first obtain the values of a and b based on the historical data of x and y , establish a regression model, and then use the model Calculate different values of y relative to different x (fill in the blanks) coefficient of determination: a, b are also called regression model parameters. The principle of coefficient determination applies the least square method. Least Squares Method : A method that requires the trend line to minimize the sum of squares of errors . Using the method of least squares, the calculation formula of the coefficient is obtained:
After finding the regression equation, according to the data of a variable given in the title, the value of another variable can be obtained by bringing it in.
Determine the correlation coefficient R.
Confidence interval: the probability that the actual value lies within this interval should reach 95% or more , If roughly conforms to the normal distribution, the confidence interval is
- Binary linear regression model method. The general formula of the binary linear regression model is as follows:
By using the method of least squares, the values of the regression parameters a, b1, b2 can be obtained
11. Forecast of seasonal changes
When we forecast its sales volume and selling price, we should consider two trends: seasonal changes and general changes. When analyzing and pre-sales of the seasonal changes in product sales and prices, we should focus on The investigation and study of market conditions should focus on the combination of qualitative forecasts. For quantitative forecasting, we use the principle of exponential smoothing.
Summary of Exam Questions
Glossary
fill in the blank
Social Science Military Economics
