What are Prediction Markets?

Market forecasting is an analysis and inference process based on various information and materials obtained from market surveys, through analysis and research, and using scientific forecasting techniques and methods for the future commodity supply and demand trends, influencing factors and changes in the market. . ^[1]

market prediction

Forecasting itself must rely on methods such as mathematics and statistics, as well as advanced methods. Let s not talk about technology and methods first. For the managers of an enterprise, what may be the first concern is how to form an effective way of thinking? The following principles may be instructive:

Market forecasts can be classified according to different criteria.

By predicted time span

According to the forecast time span, market forecast can be divided into short, medium, long and short term forecasts.

The short-term forecast is based on the actual situation of demand changes in the market.

Introduction

Venus China Venus China as a case, using operations research and computer-aided management principles to conduct in-depth research and analysis on the marketing history and current status of its products, large-screen color monitors (referred to as color displays), to establish mathematical models And use computers to make scientific predictions and formulate business strategies for the future. This paper uses mathematical models and self-developed software packages to establish an integrated marketing management information system. The system automatically collects, processes, and analyzes useful, timely, and accurate information from marketing transactions and corporate environments. At the same time, it can transfer classified and reassembled information to the company's management and departments in real time.

Product sales profile

Venus' sales situation in the world is optimistic. As display manufacturers in various countries have set up factories in China or released a large number of goods to China, competition in the industry is becoming increasingly fierce, but the company's Chinese company's sales have not increased significantly. In addition to factors, another important factor is that the company's internal potential is not fully tapped, especially the lack of scientific and strategic market observations, and the lack of a set of effective operating management information systems, resulting in the company's sales situation in a "market-driven "Situation. Therefore, when the company is facing an unfavorable macroeconomic environment, it cannot make a sensitive response and formulate effective countermeasures to obtain marketing initiative.

General framework of product market analysis and marketing planning system

In the world, Venus has certain advantages, but the sales situation in the Chinese market shows that the company's products have a potential crisis in the Chinese market. Therefore, Venus China proposed to develop a "marketing management information decision system" Its main function is to provide reliable and timely market information for the company's managers.

In order to achieve the target function, the system includes four functional modules:

(1) Market forecast and analysis

(2) Planning and market research

(3) Ordering and customer service

(4) Transportation and distribution

This article focuses on the research and discussion of marketing forecast analysis and planning modules. Because predictive analysis and planning research are the primary links of market operation management, it is the premise and basis for enterprises to make correct business decisions.

Data flow of marketing management information system

There are two main sources of marketing management information system: the first source is market researchers, who collect data on market conditions for market forecasting and research analysis; the second source is users, which means all The unit and individual who purchased the product, it put forward order requirements to the enterprise, as well as requirements on product quality and performance. After these raw data are input to the system, after proper processing, various market information is generated, some are stored in the corresponding database, and some are output to the relevant departments or other subsystems.

Market forecasting model

Venus China For an enterprise to make correct business decisions, forecasting and analysis play an important role. Through forecasting and analysis, the unknown state in the market is transformed into a scientifically predicted state of expectation, which enables enterprises to avoid market risks to a certain extent. On the basis of conscientious summary of past experience, we must not only strengthen the leading role of qualitative prediction and analysis, but also pay more attention to the research work of quantitative prediction and analysis, especially give full play to the role of computers, so that qualitative prediction analysis and quantitative prediction analysis are closely combined. Together, create a brand-new scientific forecasting and analysis method that is more in line with the actual product market and the company. On the one hand, with the development of China's macro economy, the development of market demand for large-screen displays has a certain continuity. On the other hand, monitors are universal products, and various brands are highly competitive. There are relatively few fixed supporting users for the display, so the development and sales of screen displays also have some uncertain factors, that is, it is difficult to consider the causality of its development. In addition, the market demand for monitors is supported by compatible PC sales, and there are certain seasonal fluctuations. For example, January and February saw sluggish sales, the situation in March turned bright, and then it was maintained in a slow decline. Sales suddenly rose in August due to the climax of buying compatible PCs during the summer. Based on this situation, I think that the forecasting method should adopt two methods: the exponential smoothing method and the seasonal variation method in the time series analysis method. The former mainly predicts short-term sales trends, while the latter focuses on forecasting seasonal changes and long-term sales changes, making up for the short-term forecasting shortcomings. A good prediction result can be obtained by combining the two prediction methods. 1. Forecast short-term sales trends with improved exponential smoothing.

The exponential smoothing method can be used to predict short-term sales trends. The basic principle of this method is to emphasize the effect of data on the predicted value, and the weight of the data can be arbitrarily selected, but the role of long-term data is not completely ignored. The mathematical model of exponential smoothing is as follows:

F [, t + 1] = F [, t] + (V [, t] -F [, t]) (3-1)

Can be written as:

F [, t + 1] = V [, t] + (1-) F [, t] (3-2)

smoothing coefficient, whose value is between 0 and 1 (0 < <1);

V [, t] the actual value of the t-th period (year or month);

F [, t] the forecast value of the t-th period (year or month);

F [, t] in equation (3-1) can be written as:

F [, t] = V [, t-1] + (1-) F [, t-1]

And F [, t-1] = V [, t-2] + (1-) F [, t-2]

...... Successive calculations in this way, and then the predicted values of different periods are substituted into formula (3-2). After expansion, we get:

F [, t + 1] = V [, t] + (1-) V [, t-1] + (1-) [2] V [, t-2] + (4-3 )

The value of in the formula should be selected according to the actual situation. If you want to strengthen the role of the data, the value of can be made larger. Assuming that = 0.9 is substituted into the above formula, we get:

F [, t + 1] = 0.9V [, t] + 0.09V [, t-1] + 0.009V [, t-2] +

It can be seen that the data plays a major role in the above formula, and the role of the remaining historical data decreases rapidly according to the weight of the proportional series (common ratio is 1-). Therefore, this method is an improved version of the weighted moving average method. It can change the weight to adjust the role of the data through the choice of value, and also consider the role of the long-term data. In practice, the choice of value can be determined based on experience. If the data fluctuations are not large and the graph is relatively stable, the value should be smaller. If the data fluctuations are large, the value should be larger. = 0.7 0.8. This enables the predicted value to react quickly to changes in the actual value, thereby reducing the deviation between the predicted value and the actual value. Taking the historical sales data of the monitor as an example, the exponential smoothing method is used to calculate the forecast values of 1990-1996 for each year according to = 0.1 and 0.9, as shown in Table 3-1.

market prediction

Exponentially smoothed predictions:

Actual forecast

Cycle (year) (million yuan) a = 0.1 a = 0.9

1987 1494.0 1494.0 1494.0

1988 1476.6 1494.0 1494.0

1989 1673.0 1492.0 1478.3

1990 1777.8 1506.7 1621.1

1991 1738.6 1533.8 1762.1

1992 2028.5 1554.3 1741.0

1993 2071.9 1601.7 1999.7

1994 2252.0 1648.8 2064.7

1995 2825.0 1709.1 2233.3

1996 2439.0 1820.7 2765.8

Figure 3-1 shows two prediction graphs with different values of in the exponential smoothing method. It can be seen that because the actual data is unstable and fluctuates greatly, in this case, when = 0.9, the predicted value graph is closer to the actual value; when = 0.1, the predicted value graph only reflects the data change The trend deviates greatly from the actual value. The exponential smoothing method is to strengthen the role of data in different periods through artificial adjustment of the alpha value, and can adapt to more complex changes. Less historical data is required. Exponential smoothing is a time series analysis method. A time series is a series that changes under the influence of random factors. Therefore, its prediction cannot be without bias. Therefore, the accuracy of the prediction needs to be explained so that there is a comparison standard when selecting the prediction method. How to determine the accuracy of the prediction? The accuracy of a certain prediction cannot be used as the criterion for evaluating the prediction method, but the average value should be used to judge from a statistical point of view. Now we will use two methods to measure prediction accuracy: mean absolute deviation and mean square error.

The mathematical expressions of the two methods are as follows:

Mean Absolute Deviation (MAD):

1 n

MAD = ( V [, t] -F [, t] ) (i = 1,2,3,, n) (3-4)

n i = 1

Mean square error (MSE):

1 n

MSE = [ (V [, t] -F [, t]) [2] (i = 1,2,3,, n) (3-5)

ni = 1

Based on these two standards, two types of values ( = 0.1 and = 0.9) are used for the error analysis of the same data in the exponential smoothing method in Table 3-1.

market prediction Analysis and comparison. As shown in Table 3-2. From the results calculated in Table 3-2 using two standards, in the case of this set of actual data, the prediction result of = 0.9 is more accurate than the prediction result of = 0.1. Comparison of error analysis:

Exponential smoothing

Period (year) (million yuan) a = 0.1 mean absolute deviation mean square error a = 0.9 mean absolute deviation

1987 1494.0 1494.0 0.0 0.0 1494.0 0.0

1988 1476.6 1494.0 17.4 302.8 1494.0 17.4

1989 1637.0 1492.3 144,7 20938.1 1478.3 158.7

1990 1777.8 1506.7 271.0 73441.0 1621.1 156.7

1991 1738.6 1533.8 204.7 41902.1 1762.1 23.5

1992 2028.5 1554.3 474.1 224770.8 1741.0 287.5

1993 2071.9 1601.7 470.1 220994.0 1999.7 72.1

1994 2252.0 1648.8 603.2 363850.2 2064.7 187.3

1995 2825.0 1709.11115.9 1245232.8 2233.3 591.7

1996 2439.0 1820.7 618.3 382294.9 2765.8 326.8

Total 3919.4 2573726.7 1821.7

Total average absolute difference 391.9 182.2

Mean square error 1 257372.7

Exponential smoothing:

Period (year) Mean square error

1987 0.0

1988 302.8

1989 25185.7

1990 24554.9

1991 552.3

1992 82656.3

1993 5198.4

1994 35081.3

1995 350108.9

1996 106790.2

Total 630430.8

Total mean absolute difference

Mean square error 1 63043.1

2. Forecast seasonal changes in seasonal demand

Although the index smoothing method can better reflect the short-term sales trend, it is not suitable for long-term forecasting. As a supplement to the short-term forecasting method, we use the seasonal change method to forecast seasonal demand changes and long-term sales changes of large-screen displays. Large-screen displays are easily affected by the sales of compatible PCs and other factors, and their market demand changes seasonally or cyclically. In order to do a good job of balanced production and timely supply, it is necessary to grasp the laws of change. Seasonal changes in the demand for large-screen displays are sometimes more complicated. It includes both trendy changes (such as year-on-year demand growth), seasonal changes, and other occasional changes (such as national political and economic conditions Sudden change). Therefore, the analysis and prediction of this changing state need to apply a variety of feasible methods for comprehensive analysis. Based on the monthly sales of Venus in 1995 and 1996, as shown in Table 3-3, the sales in the next two years are forecasted.

Prediction steps:

(1) Mark the distribution map of data points, and determine the form of the change as shown in Figure 3-2. This set of data shows two types of changes. One is a strong seasonal change. The demand in the two seasons was small. First, the trend changed, and the demand for products showed an increasing trend.

market prediction

(2) Determine long-term trend changes

There are two ways to determine the change in growth trend

(I) Draw a straight line using the fixed point of the average monthly growth rate

Drawing {}

According to the data in Table 3-3, the average monthly sales in 1995 and 1996 were obtained:

1688

Monthly average sales in 1995 = = 140.7 million yuan

2370

Monthly average sales volume in 1996 = = 197.5 million yuan

197.5-140.7

Average monthly increase = = 4.73 million yuan / month

This 4.73 million yuan / month is a long-term trend change. If the average monthly sales volume is calculated as the sales volume in the middle of the year (June), two points A and B can be given in Figure 3-3. Among them, point A is June 1995 with coordinate Y value of 140.7; point B is June 1996 with coordinate Y value of 197.5. Connecting the AB line is a long-term trend change.

(Ii) Apply the least squares method to list the linear regression equation:

Assume the equation of the line is:

Y = a + bx

market prediction In the formula:

Regression coefficients nX · Y-X · Y

b =

nX [2]-(X) [2]

YbX

a =

Substituting the data in Table 3-3 into the above two formulas:

24 × 55200-300 × 4058

b = = 3.89

24 × 4900-300 [2]

4058-3.89 × 300

a = = 120.46

twenty four

The trend mathematical model is:

Y = 120.46 + 3.89x (3-6)

(3) Calculate the monthly trend value of the trend line

Substitute the values of each month into the trend model formula (3-6) to get the trend values of each month. All calculated values are listed in item (3) of Table 3-3. The trend value for each month is used to calculate the seasonal coefficient.

(4) Determine the seasonal coefficient

The seasonal coefficient is the quotient obtained by dividing (2) and (3) in Table 3-3. The algorithm for listing the seasonal coefficient for January is:

30 ÷ 124.4 = 0.24

The rest by analogy. There are 24 months seasonal coefficients in the table, which are two complete cycles. Therefore, the monthly seasonal coefficients corresponding to each year should be averaged, and the average value should be taken as the seasonal coefficient value of each month, as shown in Table 3-4 As shown.

market prediction

Table 3-4 Seasonal coefficients

Seasonal coefficient

Month 1995 1996 Average

1 0.24 0.56 0.40

2 0.39 0.93 0.66

3 1.44 1.11 1.28

4 1.22 1.48 1.35

5 1.27 1.19 1.23

6 0.99 1.31 1.15

7 1.88 0.96 1.42

8 0.98 1.10 1.04

9 1.23 1.52 1.38

10 0.81 1.27 1.04

11 0.64 0.50 0.57

12 0.48 0.43 0.45

(5) Establish a prediction model for prediction

Assuming S [, t] is the seasonal coefficient of the t-th month, the predicted value of the t-th month

Y [, t] = (a + bX [, t]) S [, t] (3-7)

If you want the forecast for July 1997, you have:

X [, t] = 24 + 7 = 31

S [, t] = 1.42

So: Y [, t] = (120.46 + 3.89 × 31) × 1.42 = 342.29 million yuan

Also, if the demand forecast for January 1998 is obtained, then:

X [, t] = 24 + 12 + 1 = 37

S [, t] = 0.4

Y [, t] = (120.46 + 3.89 × 37) × 0.4 = 105.76 million yuan

The above discussions are the mathematical models of the two prediction methods of exponential smoothing and seasonal variation and their application examples. It should be pointed out that the use of computers to make predictions mainly lies in the use of mathematical models and improving the accuracy of predictions. The advantage of using a computer for prediction is that it can accurately process a large amount of data, and can often modify the model in time according to changing conditions. At the same time, it can also be linked to other systems to strengthen information communication. Demand data should be collected when using computers to forecast market demand. Generally speaking, the more statistical data, the better. In less important cases, you can find seven points. In important cases, you must find at least twelve points. Observing seasonal demand patterns requires at least two years of data. The time span of the data has an impact on the forecast. The span is too long, and seasonal fluctuations are masked.

For exponential smoothing, time series data is entered into the computer. The output is the predicted value of the next cycle after being calculated by the exponential smoothing method. The computer program should provide a prediction table (ATABLEOFFORECASTS). The smoothing coefficient ranges from 0.1 to 0.9; on the other hand, the program can use the least square method to select a better smoothing coefficient. At the same time, it can also calculate the weighted average according to the number of cycles specified by the user, which will benefit the sensitivity. Analysis performed. For the seasonal change prediction method, the input calculation is also time series data, and the output is the seasonal change trend in the future period. When market demand peaks and valleys, seasonal demand must be considered. Generally speaking, seasonal demand behavior requires that peaks occur at the same time in each cycle, and the peak demand must exceed the average demand MAD / 2 (average absolute deviation). The seasonal demand estimate is expressed in the computer as a trend line and seasonal coefficient.

Market research and marketing plan

The purpose of market research and marketing plan is to conduct sufficient market research and formulate a reasonable sales plan so as to minimize the risks assumed by the enterprise. The market research and marketing planning module has to complete the following three tasks:

(1) Analysis of market survey data, generally based on the competition of large-screen displays and the use of statistical analysis methods to study market issues;

(2) Use the results of the sales forecast to formulate a sales plan.

(3) Advertising analysis to facilitate the formulation of advertising strategies.

What are Prediction Markets?

market prediction

IN OTHER LANGUAGES

RELATED ARTICLES

How can we help?