JP2012038041A - Winning and losing line forecasting system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system for forecasting a winning and losing line at an election of the local assembly before voting is started using a statistical method.SOLUTION: The winning and losing line forecasting system is realized by an election forecasting device communicably connected to a user terminal. When the election forecasting device receives numerical data necessary for forecasting of a winning line at the election of the local assembly such as the number of voters, the number of assembly seats, the number of candidates and estimated voting rate from the user terminal, the election forecasting device, first of all, calculates forecast vote ratio to be obtained by winners based on a forecast formula obtained by a statistical method from past actual data, and applies next the forecast vote ratio to be obtained by winners to normal distribution and calculates the forecast winning line. Then the election forecasting device transmits screen data displaying the calculated forecast winning line contrasted with the presumed vote ratio to be obtained by the user himself or herself to the user terminal.

Description

  It is related to a system that predicts the winning line in local council elections before the start of voting using statistical methods.

As a simulation of winning predictions, a system and method for creating a vote prediction graph for each candidate and checking whether or not the winning zone is reached are currently in operation.
This method predicts each candidate's vote based on the result of the early vote of each candidate in the number of votes, based on the pre-election situation survey and taking into account the remaining vote acquisition rate.
In other words, it uses Bayesian statistics that make full use of prior probabilities and posterior probabilities, and is being tried in a wide range of fields, not limited to election prediction. For election predictions, there are also methods to increase the accuracy by adding so-called exit survey results.
The present invention uses a statistical method that is completely different from such existing methods. That is, the prediction system combines regression analysis and a normal distribution approximation method. When we searched for prior patent documents in the patent electronic library, many prediction systems using regression analysis such as the present invention were filed, but all of them were systems related to fields unrelated to the election prediction. .

In the above winning prediction simulation, the prediction can be made only after the vote is closed and the voting rate is known on the election day. In addition, in the case of elections for congressional members competing for multiple seats, conditions such as the implementation of exit surveys become more severe.
The applicant has two characteristics: firstly, it can be applied to local council elections that contend for multiple seats, and secondly, prospective candidates can predict their winning line before the voting date. We have been searching for the realization of the system for many years.
Based on this awareness of issues, we analyzed the municipal assembly elections conducted in Japan under the Public Constitution Election Law under the new constitution. First, the competition rate (number of candidates ÷ constant) and the winners We found that there is a very strong correlation between the percentage of votes cast (winning vote rate). Secondly, it was also found that a candidate's vote distribution can be approximated by a normal distribution with the aid of statistical analysis techniques.
The present invention is based on these two findings and aims to realize a system that predicts a winning line in a municipal assembly election, which was impossible in the past, and enables candidates to simulate their own chances of winning. .

In order to achieve the above object, the invention according to claim 1 is a winning line prediction for realizing a pre-election prediction of a winning line in a congressional election by means of an election prediction device that is communicably connected to a user terminal. A system,
The election prediction device is
Predictive request receiving means for receiving numerical data necessary for prediction of the winning line (for example, the number of voters for the election of members, the number of members, the number of candidates, the estimated vote rate, etc.)
A winner vote rate predicted value calculation means for calculating a predicted winner vote rate based on a prediction formula calculated by a statistical method based on past performance data;
A winning line predicted value calculating means for calculating a predicted winning line by applying the calculated predicted winner percentage to a normal distribution,
And a prediction result output means for transmitting the calculated prediction winning line to the user terminal.
The “winner vote rate” is the ratio of the number of votes of all winners to the number of votes, and is the most important concept for the prediction system of the present invention as will be described in detail later.
In this way, a user such as a candidate for candidate can predict a winning line simply by inputting numerical data such as the number of voters and the number of legislators. Uses a prediction method that uses only statistical data, not qualitative data, so the attributes of the user (gender, age, political party, etc.) and the attributes of the municipality where the candidate is planned (geographical location, urban or local) Universally relevant results, regardless of the other). Because the user can easily obtain results by inputting only a few input items, every time the number of candidate candidates or the expected number of votes fluctuates, this system is used to make predictions corresponding to the latest situation. You can get a winning line.
In addition, the effectiveness of this approach has not yet been verified in the chief or national elections.

In order to achieve the above object, the data received by the prediction request receiving means also includes the expected number of votes, and the prediction result output means includes the inputted expected number of votes and the calculated predicted winning line. It is preferable to create screen display data that visually corresponds to and transmit it to the user terminal.
Thereby, a prediction result can be grasped easily. The result display using only letters and numbers is not what a busy user is waiting for the election. The ease of input and the ease of understanding the output are important along with the high accuracy.

In order to achieve the above object, the winning voter predicted value calculation means may use parameters calculated by performing regression analysis with the competition rate as an independent variable and the winners vote rate as a dependent variable.
Here, the “competition rate” is a value obtained by dividing the number of candidates by a member of the Diet.

Candidates, etc., in the electoral elections in the municipality to which attention is paid, calculate the predicted value of the winning line instantly if the specified items are entered before the election, and where the expected number of votes will be based on the winning line Can be displayed.
In addition, when the system of the present invention was prototyped and verified in the 462 elections recently conducted among the elections of a few hundred local governments in Japan, it was possible to predict a winning line that could withstand practical use with 95% reliability. It could be confirmed.

It is the figure which plotted the voting rate which is track record data, and a winning voter's rate, and fitted the approximate curve. It is a figure which illustrates the histogram of the actual number of votes obtained. It is a figure showing the probability density function and cumulative distribution function of a standard normal distribution. It is a figure which shows the structural example of the system which concerns on embodiment. It is a functional block diagram of an election prediction device concerning an embodiment. It is a flowchart of the winning line prediction process which concerns on embodiment. It is a figure which illustrates the input screen of the winning line prediction process which concerns on embodiment. It is a figure which illustrates the result display screen of the winning line prediction process which concerns on embodiment.

A system according to an embodiment of the present invention (hereinafter, this system) will be described with reference to the drawings.

Before describing the configuration and processing of the system in detail, the concept of winning line prediction according to the system will be described.
This prediction system has two stages: 1) prediction of the winning voter rate, and 2) approximation by the normal distribution of the number of candidate votes.

First, the prediction of the winning voter percentage is described.
The reason for the development of this system is that the applicant is the “competition rate (number of candidates) / constant number” of the election and “the ratio of the total number of votes of all the winners to the number of effective votes (winning vote rate)”. There is a place to pay attention to the relationship.
Taking Asahikawa City as an example, Table 1 shows the competition rate and the winning voter rate of post-war city council members.

When the competition rate is plotted on the horizontal axis and the vote rate is plotted on the vertical axis, there is a correlation between the two, and the graph in FIG. 1 shows an approximation curve.
Using the approximate curve as equation (1), the competition rate of the actual data in Table 1 as the independent variable x, and the logarithmic linear regression with the winner vote rate of the actual data as the dependent variable y, the predicted value of the winner vote rate is obtained. Thus, the coefficients a and b of the formula (1) for calculating the above were obtained, and a = 1.0017 and b = −0.5729 were obtained. Since this calculation method is publicly known, description thereof is omitted.

Since the determination coefficient at this time is R ² = 0.9787, it can be seen that there is a very high correlation.
The strong relationship between the competitive rate and the winning voter rate is not limited to Asahikawa City, but is also recognized in other cities. Each parameter in the above equation (1) is a≈1, and b is around -0.6. Value. It was -0.5785 when the value of b was calculated and the average value was calculated | required, respectively about 19 local governments of Hokkaido extracted arbitrarily. Therefore, I decided to use equation (2) as an approximate expression for calculating the winning voter rate, and calculated for local governments other than Hokkaido.

For example, in the 2007 Utsunomiya City Council election, the competition rate was 1.18, and the actual winners vote rate was 0.92.
When x = 1.18 was substituted into the above equation (2), y = 0.91 was obtained. The predicted value / actual value ratio was 0.99 (= 0.91 / 0.92), and a very good result was obtained. Similar results were obtained not only in Utsunomiya City but also in other municipalities.
As a result, it was found that if the values of the parameters a and b are appropriately determined in the municipal assembly election, Equation (1) gives a sufficiently accurate predicted value.

Next, an approximation by the normal distribution of the number of candidate votes will be described.
Considering the distribution of the votes of candidates in actual municipal assembly elections, there are generally few high-voting candidates and low-voting candidates, regardless of the local government or the year of the election, and candidates with intermediate votes. It turns out that there are many. Fig. 2 illustrates the Hakodate City Election Histogram in 1999, but this is not the only example. Basically, there are few high- and low-voting candidates and many intermediate-voting candidates. Show the distribution. Therefore, we decided to assume that “the distribution of votes for municipal election candidates can be approximated by a normal distribution”.

The characteristic of the normal distribution is that the percentage of the total deviation is calculated by how many times the standard deviation s is away from the central axis, and if you apply the fact that you can estimate the winning voter percentage as mentioned above, It is possible to estimate the boundary value between the voter and the winner.
FIG. 3 is a graph showing a probability density function and a cumulative distribution function of a standard normal distribution. For example, if the electoral voter rate is 85% in the A-city parliamentary elections (the lost voter voter rate is 15%), the cumulative distribution function corresponding to 0.15 on the vertical axis is calculated as z≈−1.004. . Since the central axis corresponds to the average value (μ) of the candidate vote, when converted to the original normal distribution (average μ, standard deviation s), Equation (3) is obtained. T is a prediction winning line to be obtained.

The above is a calculation example in which the winning voter percentage is 85%. In general, z is obtained by searching the standard normal distribution table. The general form is expressed by equation (4).

Next, consider the standard deviation s. As can be seen from the graph of FIG. 3, the vote distribution frequency decreases as the distance from the center value increases, and becomes almost 0 (less than 1% in total for both bases) when it exceeds 3 times s. However, in the election system, the lowest candidate may have zero votes, so if z = -5 where the cumulative distribution function value is substantially zero, then μ = 5s. . From this, the standard deviation used in this system is obtained by s = μ ÷ 5. T to be obtained can be calculated by equation (5).
In this way, the predicted value T of the winning line that the user of this system wants to know is calculated. Predictions with high accuracy can be realized by using a very simple formula and inputting a small amount of quantitative data.

Next, the configuration of the present embodiment will be described with reference to the system configuration diagram of FIG. 4 and the functional block diagram of the election prediction apparatus 2 of FIG.
As shown in FIG. 4, in the present system 1, the user terminal 3 accessing the election prediction device 2 via the Internet N obtains a predicted value of the winning line from the election prediction device 2.

The user terminal 3 is a terminal used by a user, that is, a candidacy candidate (including a prospective person) who intends to make use of the system, and related parties, or a news agency. This user terminal 3 is a so-called personal computer, mobile phone, PDA (Personal Digital
Assistant), etc., equipped with a keyboard and touch panel for inputting various data, a display for outputting the input data and data received via the network, etc., and transmitting and receiving data according to a predetermined protocol It can be performed.
In this system, when data is transmitted to the election prediction apparatus 2 or when a prediction result is received, it is performed via a Web page. Therefore, the user terminal 3 needs to have installed software (browser) for displaying a Web page described in HTML or the like on the screen.
Although there are two user terminals 3 in FIG. 4, the number is not limited.

The election prediction device 2 includes a CPU (Central Processing Unit), a computer program executed by the CPU, an internal memory such as a RAM (Random Access Memory) and a ROM (Read Only Memory) for storing the computer program and predetermined data, and a hard disk drive. The storage unit 4 and the processing unit 5 are configured by an external storage device such as the above. In addition, the communication interface unit 6, input means such as a mouse and keyboard, and output means such as a display and a printer are provided.
Although only one device is shown in FIG. 4, the processing is not necessarily performed by one device, and a plurality of information processing apparatuses may execute the processing in cooperation. For example, the performance data storage means 8 in the storage unit 4 may be provided in a hard disk built in the election prediction device 2 or an external hard disk connected by a cable, but a database server other than the election prediction device 2 is provided as a database server. It may be used as

The storage unit 4 includes user information storage means 7, performance data storage means 8, parameter storage means 9, program storage means 10, calculation data storage means 11, and other storage means. The other storage means stores a standard normal distribution table, default values when there is no user input, various constants, a screen data template to be transmitted to the user terminal 3, and the like.
The processing unit 5 includes a user information management unit 12, a regression equation calculation unit 13, a prediction request reception unit 14, a winner vote rate prediction value calculation unit 15, a winning line prediction value calculation unit 16, and a prediction result output. Means 17 and other processing means are included. However, the classification of each means of these processing units 5 is merely for convenience of explanation, and does not actually define the method of realization in this embodiment. In short, it is only necessary to realize the functions described here. In each means, the OS develops the computer program stored in the program storage means 10 on the memory, and the CPU executes it.

In the user information storage unit 7, information related to the user of this system is stored and managed by the user information management unit 12. In this system, it is desirable that the system can be used only by previously registered users in order to prevent interest-free use by persons unrelated to the election and for billing purposes. Therefore, the user information storage means 7 stores a user name, password, contact information, credit card information, etc. in association with the user ID.

The performance data storage means 8 is means for storing data relating to past local assembly elections. The local government name, the year of the election, the competition rate, and the winning voter rate are stored in association with each other.
The regression equation calculation means 13 performs statistical processing on the data acquired from the result data storage means 8 and approximate curve parameters (a and b in the expression (1)) used for calculating the predicted value of the winning voter percentage. Is calculated and stored in the parameter storage unit 9.
The approximate curve parameter calculation by the regression equation calculation means 13 is performed based on the actual data at an appropriate timing. It is important that prediction formulas can be used to calculate highly accurate prediction values, and collection of performance data is accumulated every time elections are held in various locations. If necessary, old data is excluded from the basic data for calculating the prediction formula. Thus, if the latest information is taken in and parameters are calculated, prediction with higher accuracy can be realized.
The collection of the actual data and the calculation process of the approximate curve are performed at an appropriate timing independently of the winning line prediction process that is the object of the present system.

The prediction request receiving unit 14 receives a prediction processing request from the user terminal 3, and passes received data such as the number of voters to the winner voter predicted rate calculation unit 15.
The winner vote rate predicted value calculation means 15 calculates the competition rate from the number of candidates received from the user terminal 3 and the legislator constant, and calculates the predicted value of the winner vote rate based on the parameters acquired from the parameter storage means 9. To do.
The winning line predicted value calculation means 16 calculates the predicted value of the winning line using a normal distribution. Values calculated in the course of these processes are written to the calculated data storage unit 11 each time and used in subsequent processes.
The prediction result output means 17 displays the predicted winning line on the graph, creates output data for displaying the expected number of votes inputted by the user on the screen, and transmits it to the user terminal 3.

Next, the process procedure of the winning line prediction process by the election prediction device 2 of the present embodiment will be described with reference to FIG.
The prediction request receiving unit 14 receives a prediction processing request from the user terminal 3 (step S1). User authentication is performed based on the received user ID and password. After searching the user information storage means 7 and confirming that a matching user ID and password are found, an input screen for input items is transmitted to the user terminal 3 (step S2).
The number of voters (I1), the number of legislators (I2), the number of candidates (I3), the estimated vote rate (I4), and the expected number of votes (I5) are transmitted from the user terminal 3 through the input screen illustrated in FIG. Then, this is received (step S3). It is desirable to check the numerical range as appropriate for the received numerical values. For example, when an impossible value such as a member of the Diet member constant is transmitted, screen data that prompts the user terminal 3 to input again is transmitted.
In FIG. 6, the description of error processing is omitted, but it goes without saying that predetermined processing is performed when user authentication fails or when a range error of input data and calculation data occurs. Yes.

The winner vote rate predicted value calculation means 15 calculates the predicted value (P) of the winner vote rate in the following procedure. First, the competition rate (Q) is calculated by dividing the number of candidates (I3) received from the user terminal 3 by the legislator constant (I2) (step S4). Next, the parameters a and b of the approximate curve are taken out from the parameter storage means 9, and the prediction formula for the winning voter rate is determined (step S5). By substituting the competition rate (Q) into this equation, the predicted value (P) is calculated (step S6). A value calculated in the course of processing such as the predicted value is written in the calculated data storage unit 11.

Subsequently, the winning line prediction by the winning line predicted value calculation means 16 is performed in the following procedure.
First, the candidate vote distribution is approximated by a normal distribution (step S7). The average (μ) of the normal distribution is calculated by the average value of the candidate votes, that is, the estimated number of votes (J) / the number of candidates (I3), and the standard deviation (s) of the normal distribution is approximated by μ / 5, The values of μ and s are written in the calculated data storage unit 11. Here, the estimated number of voters (J) is [the number of voters (I1) × the estimated vote rate (I4)].
Subsequently, a predicted winning line (T) is calculated using the standard normal distribution (step S8). That is, z of the cumulative frequency distribution table corresponding to (1-P) on the vertical axis is obtained from the standard normal distribution table, and T is calculated by the equation (5).

The prediction result output means 17 is used for screen display using a graph or color display so that the calculated predicted winning line (T) and the expected number of votes (I5) input by the user can be compared at a glance. Data is created and transmitted to the user terminal 3 (step S9). The user terminal 3 that has received the message displays it on the screen by the function of the web browser. FIG. 8 shows a screen display example.
The user may obtain a predicted winning line using this system every time there is a change in the number of candidates (I3) and the expected number of votes (I5). In addition, this system may be used, for example, by inputting various estimated vote rates (I4) in consideration of fluctuations in the vote rate due to the weather on the day of the vote, and estimating the number of votes to be targeted.

The present invention is characterized in that a prediction winning line is calculated when a predetermined item is input before a member of the Diet election, and the result is displayed in a graph or the like. Therefore, the above embodiment is merely an example, and various modifications can be considered within the scope of the claims.

For example, the model equation for regression analysis is not limited to the above equation (1). What is necessary is just to substitute the past performance data into various model formulas (6), and employ | adopt a prediction formula with high prediction precision. However, x and y are the variables described above. Further, a, b, c,... Are parameters accompanying the formula f.

Further, in the above embodiment, the user is required to input a numerical value. However, there is also data that can be grasped on the election prediction device 2 side. For example, the legislator constants and the number of voters are stored in a database in association with local government names. If the user registers the name of the local government that is scheduled to run as a user when registering, the system does not require the user to input the parliamentary constant and the number of voters.
The estimated vote rate can also be an optional user input item. If there is no user designation, a default value predetermined on the election prediction device 2 side may be used. Thus, the burden on the user is further reduced by reducing the essential user input items.

Further, in the above-described embodiment, the application data provider (ASP) form is taken into consideration, and the result data is based on the latest data. However, the computer program according to any one of claims 5 to 7 is sold on-line or in a package form, and input of items such as parliamentary constants on a user's personal computer, calculation of a predicted value, and display of a prediction result on a screen A series of processes may be executed. The latest performance data may be provided online or offline.

As a system provided in the form of an ASP or application package, it can be expected to be used by a candidate for a member of the Diet, its related parties, or a news agency.

1: Election prediction system,
2: Election prediction device, 3: User terminal,
4: storage unit, 5: processing unit, 6: communication interface unit,
7: user information storage means, 8: performance data storage means, 9: parameter storage means,
10: program storage means, 11: calculation data storage means,
12: user information management means, 13: regression equation calculation means, 14: prediction request receiving means,
15: Winner vote rate prediction value calculation means, 16: Winning line prediction value calculation means, 17: Prediction result output means,
N: Internet

Claims (7)

A winning line prediction system for realizing a pre-election prediction of a winning line in a congressional election by an election predicting device connected to a user terminal in a communicable manner,
The election prediction device is
Prediction request receiving means for receiving numerical data necessary for prediction of the winning line from the user terminal;
A winner vote rate predicted value calculation means for calculating a predicted winner vote rate based on a prediction formula calculated by a statistical method based on past performance data;
A winning line predicted value calculating means for calculating a predicted winning line by applying the calculated predicted winner percentage to a normal distribution,
A winning line prediction system comprising: a prediction result output means for transmitting the calculated predicted winning line to the user terminal. The numerical data received by the prediction request receiving means also includes the number of expected votes, and the prediction result output means visually associates the input number of expected votes with the calculated prediction winning line. The winning line prediction system according to claim 1, wherein screen display data is created and transmitted to the user terminal. 3. The winning voter rate predicted value calculation means uses a parameter calculated by performing regression analysis with the competition rate as an independent variable and the winner vote rate as a dependent variable. The winning line prediction system described in Crab. A computer program for causing a computer to operate as the election prediction apparatus according to any one of claims 1 to 3. A computer program for causing a computer having an input means and a screen display means to execute a pre-election prediction of a winning line in a legislator election,
In the computer,
Inputting numerical data necessary for prediction of a winning line through the input means;
Calculating a predicted winner percentage based on a prediction formula calculated by a statistical method based on past performance data;
Applying the calculated predicted winner percentage to a normal distribution to calculate a predicted winning line;
And a step of displaying the calculated predicted winning line on the screen display means. 6. The expected number of votes is also input via the input means, and the input number of expected votes and the calculated predicted winning line are displayed on the screen in a visually correlated manner. The computer program described. 7. The parameter calculated by performing regression analysis using the competition rate as an independent variable and the winner's vote rate as a dependent variable, for calculating the predicted winner vote rate. A computer program according to the above.

Images