JP2003529116A - Ad evolution with evolution algorithms - Google Patents

Abstract

(57) [Summary] Perform an ordered trial and error search for selective advertising characteristics using an evolutionary algorithm. Given an initial population of sample advertisements, the characteristics of each advertisement are encoded as a set of genes related to the writing members of the population. The effectiveness of each member is evaluated, for example, using the number of times an Internet user clicked on each advertisement. Members of the population generate a plurality of child advertisements inheriting features from the parent. Members of the population that are more effective than others are preferentially selected for offspring. Continued preferential selection of more effective parents increases the likelihood of producing children with a higher degree of effect. The inheritance of particular features or combinations of features in this evolutionary process provides the indication and validation of these advertising features that produce effective results.

Description

Detailed Description of the Invention

The present invention relates to the field of advertising, and in particular to the use of evolutionary algorithms in the generation and evolution of selective advertising.

In an environment of free competition, effective advertising is often the decisive factor in the commercial success or failure of a product. The demands on the share of viewer attention are so great that an entire industry, such as the Saray Vision industry, is funded primarily by advertisers. Web page providers receive advertising revenue by including advertising banners on their sites. Internet services and computing systems that use these services are provided free of charge to users who want their ads to appear continuously on their screens.

Traditionally, evolving the effectiveness of advertising campaigns is a time consuming, expensive and somewhat inefficient process. Prior to the advertisement being published, consumer feedback surveys are processed to determine viewer preferences and impressions, usually by questionnaires before and after viewing the proposed advertisement (s). In some cases, the proposed ads are published to a small test market for ongoing feedback. After the advertisement has been published, research is often conducted to determine the effectiveness of this advertisement on a random sample of possible viewers. The effectiveness of advertisements is typically assessed in terms of memory, attraction, and some behavior that occurs in response to the advertisement. The survey may also be conducted periodically to determine the effectiveness of ongoing advertising campaigns and may address detrimental factors associated with continuously repeating advertisements, such as viewer boredom or embarrassment.

Advertising characteristics such as content, color, behavior, placement, duration, etc. have a significant impact on the effectiveness of the advertisement. However, causality is not easily described or quantified. It is estimated that there is a correlation between a particular ad feature and ad effectiveness, and each ad developer uses those features that he or she believes are relevant to possible success. However, it is not possible to directly determine the validity of the estimated correlation, nor to determine the impact between the interrelationships. That is, if an ad is determined to be valid, then often, which particular feature or combination of features of the ad has the greatest impact on this success, which feature has the least impact, Even if other features are changed, it is not possible to determine which features have the greatest impact. Better ad development and the effectiveness of ad developers' beliefs are therefore a matter of speculation and speculation, despite the resources devoted to assessing and assessing the effectiveness of particular ads or particular characteristics. It appears to be.

It is an object of the present invention to provide a method and system for assessing the effect of a particular feature or combination of features on an advertisement. Another object of the present invention is to provide a method and system for improving the effectiveness of advertisement. Another object of the present invention is to provide a method and system for generating selective advertisement.

These and other objectives are achieved through the use of evolutionary algorithms that provide directed trial and error searching of selective advertising features. Give the initial population of sample advertisements,
Each advertising feature is encoded as a set of genes for each member of the population. The effectiveness of each member is evaluated using, for example, the number of clicks by the internet user in each advertisement. Members of the population generate multiple child advertisements, which inherit the characteristics from their parents. Members of the population that are more effective than others are preferentially selected for the offspring. Sequential preferential selection of parents who are more effective than others increases the likelihood of developing offspring with a higher degree of effect. The inheritance of specific features or combination of features in this evolutionary process provides direction and validation of these advertising features that produce effective results.

[0007]   The invention will be described in more detail, by way of example, with reference to the accompanying drawings.

FIG. 1 illustrates an example web page 100 that includes various advertisements 110-170.
Each ad 110-170 has characteristics that distinguish it from all other ads 110-170. Each of these advertisements competes for the viewer's attention, and the designer of each advertisement has selected features that he or she believes will capture the viewer's attention more than other features. For example, the designer of advertisement 110 may choose a heavy, heavy typeface and the designer of advertisement 120 may choose a lighter typeface. Heretofore, it has been difficult to ascertain whether heavy, thick typeface features are preferable to lighter typeface features. Advertising 110
Designers have also selected different words than the designers of the advertisement 120. Until now, it was difficult to ascertain whether the document content “Look Here!” (See here) in the advertisement 110 is preferable to the document “Buy Now” in the advertisement 120. In one way, if the advertisement 120 is determined to attract more viewers than the advertisement 110, it is the typeface features, the document content features, or a combination of these two features that make the advertisement 120 more effective. It was difficult to see if it was. The location of the advertisement 120 may also affect its effectiveness as compared to the advertisement 110. Similarly, the advertisement 160 has unique shape features that may enhance or reduce its effectiveness. The advertisement 170 has an evolving banner 170A and an associated FIG. 170B that may or may not affect its effect in terms of attracting the viewer's attention. Advertising 140 and 1
50 does not include the document content, and the advertisement 130 is the main figure 130A and the small document 13
With 0B.

The Internet environment offers a unique opportunity to evaluate the effectiveness of advertising. According to one aspect of the invention, the number of times an ad is selected by a viewer is used as a measure of the effectiveness of the ad, and the effect of a particular feature or combination of features is evaluated based on this criterion. By its nature, this measure of effectiveness is "noisy" in that it may not really be the measure of advertising appeal in all cases. The viewer may be looking for an airline ticket, for example. Advertising 170
May be the only advertisement on the web page 100 devoted to the ticket, and its choice by the viewer will not be related to its attention-giving ability. However, for millions of internet viewers a day, the number of times an ad is selected can be considered to be relevant to its attention-giving effect. For example, advertisement 12
If 0 is selected significantly more often than Ad 110, it makes sense to assume that there is something about the different characteristics of Ads 110 and 120 that produced this distinctly different result. If the ads 110 and 120 are each selected significantly more frequently than any of the other ads 130-170, then these better for features common to the ads 110 and 120, such as rectangular shape and text-only features. It makes sense to consider an ability to be something that resulted.

The Internet environment also provides a unique opportunity to evaluate selective advertising features.
A similar advertisement with different characteristics to the millions of possible viewers mentioned above,
It can be given on a weekly, daily or hourly basis. In many cases, the selective advertisement can be automatically provided to the possible viewers using, for example, rules or knowledge based systems or simple algorithms. According to another aspect of the invention, a variant of the advertisement is given to the possible viewers and the number of times each variant is selected is used as a measure of the effect for each variant. The following modification occurs depending on the criterion of the effect of the previous modification. For example, if one of the changed features is color, and blue ads show clearly more frequent choices than red ads, then differences in blue shadows may be evaluated next, and variations on different shades of red. There is no need to evaluate. As mentioned above, however, the correlation between a particular feature of the advertisement or a combination of features of the advertisement and the advertisement effect is difficult to determine. In the red / blue example, the poor performance may have been the use of a particular typeface in red advertising, and the use of a thicker typeface in red may be more effective than the use of the same thicker typeface in blue unknown. A class of algorithms, called evolutionary algorithms, may be particularly effective in determining the most effective combination of features that maximizes the effect, without requiring specific determination of the effect of each feature alone or in combination. know. According to the present invention, evolutionary algorithms are used to guide the generation and evolution of selective advertising features.

Evolutionary algorithms operate by an iterative, generative process. Evolutionary algorithms include genetic algorithms, mutation algorithms, etc. In a typical evolutionary algorithm, it is assumed that a particular attribute or gene is associated with the ability to perform a given task, and different combinations of genes result in different levels of effect of performing this task. Early evolution algorithms are particularly effective for problems where the relationship between the combination of attributes and the effect of doing the task does not have a closed form solution.

Using the progeny process, the specific combinations of genes that are most effective in performing a given task are determined using directed trial and error searches. The combination of genes or attributes is called a chromosome. In the genetic algorithm class of evolutionary algorithms, the reproductive-recombination cycle is used to breed offspring. Members of the population with different chromosomes pair and develop offspring. These children are
Typically, it has the attributes inherited from the parent members as some random combination of genes from each parent. In classical genetic algorithms, individuals who are more effective than others in performing a given task are given more opportunities to pair and develop offspring. That is, it gives individuals with suitable chromosomes more chances to develop offspring in the hope that they will inherit any gene that has their parents effectively performed a given task. Subsequent occurrences of parents are selected based on their priority with respect to their effectiveness in performing a given task. In this way, the number of children with attributes that are effective in performing a given task tends to increase with each generation. Asexual reproduction,
Examples of other methods of generating offspring such as mutations and the like are also used to generate offspring generations with increased likelihood of improved ability to perform a given task.

In the context of this disclosure, the population consists of member advertisements having chromosomes that reflect different advertisement characteristics. Some combinations of advertising features are more effective in attracting the viewer's attention than others. According to the present invention, it is considered that the effect of the child advertisement that attracts the viewer's attention is increased by generating the child from a member advertisement having a chromosome that is more effective in attracting the child's attention.

FIG. 2 illustrates an example ad evolution 201 according to the present invention. The child of ad 201 is shown as ad 211. The child advertisement 211 has the features of the parent advertisement 201, such as its shape, message content and typeface. Unlike the parent advertisement 211 in the size of the document,
Therefore, it can be called a mutation of the parent advertisement 211. Ad 212 shows a child ad that is a combination of features from two parent ads 201 and 291 and, in this example, assuming parent ad 291 has italicized features, child ad 212 receives from its parent 201 its The shape, message content, font format, and document size characteristics are inherited, and the italic characteristics are inherited from its parent 291.
As in natural evolution, some of the child's characteristics may be different from both parents and may be a mixture of the characteristics of each parent. For example, the ad 292 may have an elliptical shape, and the children 222 of the ad 292 (oval) and the ad 211 (rectangle) are shown in FIG. 2 as having a square shape. Advertisement 241 shows the inheritance of its message content from parent advertisement 293 and its square appearance inheritance from advertisement 231 as it was inherited from advertisement 222 through advertisement 232. Advertisement 251 is the original advertisement 2
Although it is a 01 series, it shows little if any of the features of advertisement 201, and advertisement 251 shows the inheritance of the graphic message content from advertisement 294 and the shape from advertisement 242. According to the present invention, the likelihood of inheriting each feature from generation to generation depends on the success rate of previous advertisements with this feature, as discussed below.

Many evolutionary algorithms are available that can be used in accordance with the present invention. The CRC suitability search algorithm has been found to be particularly effective for complex complex industry tasks. Rally J. J., published February 14, 1995. Eschelman and James D. US Pat. No. 5,390,283 by Shafer
"A method for optimizing the configuration of a selection and placement machine" shows the use of the CHC algorithm to determine a near-optimal assignment of components in a "selection and placement" machine, see this patent specification. Include here by. Compared to other evolutionary algorithms, the CHC algorithm is a genetic algorithm that uses "surviving the fittest" selection and uses only the better performing individuals of the parent or child to generate the next offspring. To counteract the adverse strain effects that such selective survival might introduce, the CHC algorithm avoids incest, breeding between individuals with very similar attributes. As will be apparent to those skilled in the art, each evolutionary algorithm presents pros and cons with respect to the scheme used to perform the iterative solution, and the particular choice of evolutionary algorithm is optimal for use in the present invention. For the sake of clarity and ease of understanding, the use of other evolutionary algorithms will be obvious to those skilled in the art in the context of the present disclosure, but will provide details of the invention using techniques common to the CHC algorithm.

According to the present invention, the chromosome is defined to include the characteristic outline of the advertisement. The feature may be any feature or attribute that the ad designer considers to be relevant to the attention-giving ability of the ad, and may include, for example, color, shape, texture, content, animation, and the like. The features may be an indirect representation of a set of features and may be the use of specific design rules and guidelines. FIG. 3A shows a message content 391,
An example of a set of features consisting of a text font 392, a text size 393, and whether the text is italicized 394 is shown. In FIG. 3B, eight member advertisements 301-308 are shown. Member advertisement 301-308,
Characterize using these features 391-394. FIG. 3F includes keys or mappings of individual features to codes used to encode the features. For example, the advertisement 301 includes the message "Buy Now!"
The message table of F is “0” related to the message 381 ′ “Buy Now!”.
3 shows the message region 391 of chromosome 301C corresponding to advertisement 301, which includes the encoding "00" in FIG. 3A. Advertisements 302 and 304 have the message "Look Here!" (See here). ) ”Is included in FIG.
Code 3 of "01" where F relates to message 381 '"Look Here!"
In order to indicate 81, it is encoded with "01" in the message area 391. Similarly, the text of the advertisement 301 is represented in the script typeface. FIG. 3F shows the code 382 of "010" corresponding to the font feature 381 'of "script". In this way, the font area 3 of the chromosome 301C corresponding to the advertisement 301
92 includes the code “010”. The font size 393 of the advertisement 301 is shown in FIG. 3F.
As defined in the mapping 393-393 ′ of FIG. The features of each of the eight example advertisements 301-308 are similarly encoded.

After showing the advertisements 301-308 to possible viewers, the effect criteria 301E-308
Determine E for each. In a preferred embodiment, the effectiveness metric is based on the number of times each advertisement was selected within a given time period and is normalized to the total number of selections during this period. In the example of FIG. 3A, the criteria for effectiveness is given as the total number of ads selected per 1000 total selections. Ads 304 and 302 are shown as having a highest (201) criteria 304E and a lowest (9) criteria 302E, respectively.

According to the present invention, chromosomes 301C-308C of each advertisement 301-308 are combined 350 in pairs to generate offspring chromosomes 311C-318C. Using the CHC algorithm, each child inherits all of the genes common to both parents and a random selection of different genes in each parent. In the example embodiment, each bit value of each region of the chromosome constitutes a gene. Chromosome 307C, 308C
Consider that the paired binding of sigma develops into progenitor chromosomes 317C and 318C. From left to right, parents 307C, 308C have the same first, fifth and eighth genes, thus each of the offspring 317C, 318C has the same first, fifth and eighth genes. A random number of different genes are exchanged, in this example the fourth gene 361 and the seventh gene 362 are exchanged in each offspring 317C, 318C. That is, chromosome 317C is the same as 307C except for the 4th and 7th genes, and chromosome 318C is also the same as 308C except for the 4th and 7th genes. Each of the underlined gene values in Figures 3B and 3D represents a randomly exchanged gene value.

The child chromosome values 311C-318C correspond to the new child advertisement 311-318.
That is, for example, the chromosome 317C (reference numeral 01-000-00-0) is shown in FIG. 3B.
As shown by advertisement 317 in, "Look Here!" Message (reference numeral 01), Ariel font (reference numeral 000), 8 point size (reference numeral 00).
) And non-italic (reference 0). Chromosome 318C (reference numeral 00-1)
10-11-0) "Buy" as indicated by advertisement 318 in FIG. 3B.
Now! Corresponds to message (00), Lucita font (110), 12 point size (11) and non-italic (0). Each of the other offspring chromosomes,
Similarly, decrypt these corresponding child ads.

The child advertisements 311 to 318 are shown to possible viewers, and the effect is evaluated in the same manner as the member advertisements 301 to 308. Normalized score 311E-31 corresponding to each child
8E is shown in the dotted box in FIG. 3B.

Original member advertisement 301-308 and newly generated child advertisement 311-3
From the set of 18, the best performing ad is selected as the parent for the next occurrence of the child. FIG. 3C shows a selection of the eight best performing advertisements for the next pair of pairs 351. The first advertisement in FIG. 3C is advertisement 315, which is 208
3C, the next ad in FIG. 3C is ad 304, which is 20
Have an effect rating of 1, and so on. Note that the lowest rated member in FIG. 3C has an effect rating of 73. All pre-members and offspring with efficacy ratings below 73 are not selected for offspring reproduction. In this way, it is estimated that the features represented by genes are associated with the effects of advertising, and that each generation develops from a pool of well-performing genes, thereby increasing the likelihood of developing high-performance offspring. .

The advertisements were arranged in FIG. 3C to give a high degree of diversification between the mating parent pairs. For example, advertisements 315 and 304 were paired because their chromosomes are practically different and have only three (2nd, 4th and 6th) gene values in common. As shown in FIG. 3D, progenitor chromosomes 321C-328C are generated from the members of FIG. 3C, as discussed with respect to FIG. 3B. Offspring chromosome 32
Show new child advertisements 321-328 corresponding to 1C-328C to possible viewers,
Evaluate for effectiveness as with previous advertisements 301-308, 311-318. The normalized scores 321E-328E corresponding to each child are shown in the dotted box in FIG. 3D.

The parent member advertisements 315, 304 ,. ． ． , 318, 311 and FIG. 3D.
From the set of newly-occurring child ads 321-328 of, select the best-performing ad as the parent for the next occurrence of the child. FIG. 3E shows a selection of eight best performing ads for the next pair of mating pairs. As can be seen, the average effect of this parent member in the third generation is substantially higher than each of the two previous generations. Also note that, as expected, the occurrence of advertisements with small size features is rare in this third generation. As seen in FIG. 3E, the smaller (8
The gene 368 that distinguishes between -10 point size text and larger (12-14 point) size text is actually larger (12-
14 points) Focus on a value of 1 corresponding to the size feature. Five of the eight high-performance chromosomes have a 14-point size feature (10), two have a 12-point size feature (11), and one has a 10-point size feature ( 01). Note also that the encoding of the size feature uses Gray code encoding in which adjacent values differ by only 1 bit. Such encoding helps ensure that the child is close in value to the parent for those features that have a specified sensation, such as size and intensity features.

FIG. 4 shows an example of a flow diagram for optimizing these attention-giving abilities for evolutionary generations of advertisements. At 410, the features to be evaluated and mutated are specified, and the method of encoding the features as a chromosomal gene is specified. At 420, an initial population of advertisements is defined, and at 430, the characteristics of these advertisements are encoded into chromosomes for future offspring. Alternatively, the desired features can be encoded first, and then an advertisement with these features can be automatically created or formed.

At 440, the effectiveness of each advertisement used is determined. As mentioned above, the common effect criterion for the effectiveness of an advertisement is based on the number of times the advertisement has been selected with respect to other information. Alternative criteria will be common to those skilled in the art in light of this disclosure.
For example, the time the user stays on the page containing the ad may be used as an indication that the ad has attracted the user's attention, instead of or in addition to the ad selection criteria. May be used. Similarly, the number of times a viewer purchased an item via a link through the ad, or the amount of dollars for such a purchase, could be used separately or in combination with the number of times the ad was selected. In a preferred embodiment, the purchase is given an effect value greater than a mere selection, and the value of the advertisement is based on the sum of the effect values. Regarding the intended intended audience,
The value of the effect criterion may include a weighting coefficient according to a demographic classification of a viewer who selects the advertisement. The weighting factors themselves may be controlled by a gene or set of genes, as well as other factors that control how demographics affect how ads are created.

Loops 450-472 go through the evolution process. At 450, child advertising features are generated from the features of current members of the population. In a preferred embodiment,
Use the CHC algorithm. As discussed with respect to FIG. 3C, the parent ad with the highest variety of features is selected and a child ad pair is generated.

A child with child advertising features occurs at 455. This advertisement generation may be an automated process, a manual process, or a combination of both. For example, font format changes may be made automatically, and text size or message content changes may require one or more direct replacements due to other limiting factors or artistic considerations. Each of the child ads is evaluated at 460 using the same effect criteria as used for the original member at 440. For example, child advertisements may be displayed on various web pages for one week, and the number of selections may be counted for each advertisement. In an alternative embodiment, space for one ad is allocated to selected web pages to reduce sampling variations. Each time the selected web page is accessed by a viewer, a different child advertisement is placed in the allocated space. The effectiveness of each child ad is measured by the number of times the ad is selected after a predetermined number of viewers have accessed a web page containing the ad.
The predetermined number of accesses is determined according to the desired degree of evaluation accuracy.
Obtaining additional samples reduces the noise associated with the measurement and improves the reliability of the measurement for comparison purposes. In comparison, obtaining additional samples requires additional evaluation time per sample generation. In the preferred embodiment, common industry trade-off analysis techniques and statistical sampling techniques are used to determine an appropriate assessment sample size and / or assessment period. The evaluation technology used is the result of the evaluation of the first-generation advertisement.
It is important to note that it must be comparable to the evaluation results of each of the previous generation advertisements.

The selection of the next generation parent advertisement is made at 470. Any child that has a better measure of validity than any of the parent ads replaces it and becomes the parent ad for the next generation, and the program loop loops back to 450 to generate a new child ad. 47
At 2, if the process converges or the process ends, for example by a timeout signal or a user interrupt, the evolution process stops. Optionally, at 476, the entire process may be repeated, searching for new evolutionary paths. In a preferred embodiment, mutations are introduced in all remaining members except the highest performing member and the entire process is repeated via 440. If the search stops at 476, the best performing ad of the remaining member ads is selected at 490 as the best solution found. In a preferred embodiment, the selected best performing advertisement or set of better performing advertisements is used for subsequent "product" advertisements without the burden of data collection and effectiveness evaluation. By recognizing the variable nature of viewer preferences and the detrimental effects of repetitive exposures, the evolutionary ad generation process of Figure 4 is repeated cyclically, and the identified better performing ads still perform better. And to improve the effectiveness of each advertisement, possibly by introducing changes to reduce the effects of boredom.

FIG. 5 shows an example of a block diagram of a system for providing evolving advertisements. Advertising 50
1 characterizes 510, forming chromosome 501C. Evaluate advertisement 501 550,
Criteria 55 for the effect associated with advertisement 501 and correspondingly with advertisement chromosome 501C
Give one. Many ads 501 are processed similarly. Evolutionary algorithm device 540
Collects the effect criteria 551 associated with each chromosome 501C,
11C, as described above, based on the effectiveness criteria of each of the advertisements. The ad generator 580, which may be a human, a machine, or a combination of both, may
1 is formed based on the characteristics of the next generation chromosome 511C. Next-generation advertising 511
Instead of the original advertisement 501, this is characterized 510 as described above.
And an evaluation of 550.

The evaluator 550 is a proposing device 55 that shows the advertisement 501 to one or more viewers.
2 and means for determining user response to the advertisement 501. In the preferred embodiment, the number of times the user selects the advertisement is counted 554. Optionally, other counters 556 and measuring devices may be coupled to the evaluator 550 to emphasize the quality or importance of the effect criteria, as described above. The count values from the counters 554 and 556 are
It is processed by an effect reference generator 558 to generate an effect reference 551.

The above merely illustrates the principles of the invention. Therefore, it will be apparent to those skilled in the art that various modifications may be made, without the explicit description or illustration herein, of using the principles of the invention and therefore falling within the spirit and scope thereof. For example, there are evolutionary processes that use the cloning process to give a distribution of members in a population in proportion to an effect criterion. That is, if, for example, the number of possible ads is fixed by the cost of creating each, and the number of ads used is large, then the better performing ads will be put in use more frequently than the worse performing ads. Should be. In this example, the rate at which an ad is duplicated in each generation of use for the ad is determined by the performance of the ad during the previous generation. Similarly, non-fixed length evolutionary algorithms may be used as well, and other evolutionary algorithmic applications for advertising genetics and evaluation are in the light of the present disclosure and will be appreciated by those skilled in the art from the invention as claimed. Would be obvious within the range.

A “computer program” should be understood as any software product stored on a computer-readable medium, downloadable over a network such as the Internet, or otherwise tradeable.

[Brief description of drawings]

FIG. 1 shows an example of a web page that includes various advertisements.

FIG. 2 illustrates an example of evolution of advertising according to the present invention.

FIG. 3 illustrates an example of evolution of an advertisement having features encoded as chromosomes of an evolutionary algorithm according to the present invention.

FIG. 4 shows an example of a flow chart for evaluating and evolving an advertisement according to the present invention.

Claims (15)

[Claims] 1. A method of developing a suitable advertising population, the steps of characterizing each member advertisement of the advertising population as a chromosome having a set of genes, and presenting each member advertisement to one or more viewers. Determining an effect criterion for each member advertisement in response to a relevant response to each member advertisement from the one or more viewers, each of which comprises a set of genes of one or more members of the advertisement population. Generating from the advertising population a plurality of child advertisements characterized as child chromosomes having a set of genes based on: displaying each child advertisement to one or more viewers; Determining a child effect criterion in response to an associated response from the one or more viewers to each child advertisement; Ad population, method characterized by comprising the effective reference of each member ad, and forming on the basis of the child effect reference of each child ad. 2. The method of claim 1, wherein the step of generating the plurality of child advertisements is also based on a performance criterion for member advertisements of each of the populations. 3. The method of claim 1, further comprising replacing the advertising population with the preferred advertising population, and repeating the steps of claim 1. 4. The method of claim 1, wherein presenting each of the member advertisements includes displaying the member advertisement as a selectable entity on a web page, the effect of each of the member advertisements. The method of claim 1, wherein determining a metric comprises incrementing a count value for the member advertisement whenever the selectable entity is selected. 5. The method of claim 4, wherein the step of determining a performance criterion for each member associates a second count for the member advertisement after a viewer selects the selectable entity. A method that also includes the step of incrementing every time a purchase is initiated. 6. The method of claim 1, wherein the step of generating the plurality of child advertisements determines a set of features corresponding to a chromosome of each child advertisement and each child advertisement has the corresponding feature. A method comprising: developing based on a set of. 7. In a computer program for developing a suitable advertising population,
When the computer program is executed on a computing device, it causes the computing device to show each member advertisement of an advertising population to one or more viewers, and an effect criterion relating to each member advertisement is displayed to the one or more viewers. To generate a plurality of child advertisement characteristics from the characteristics of the advertisement population, and to determine a plurality of child advertisements corresponding to the plurality of child advertisement characteristics by at least one of the plurality of child advertisement characteristics. Letting a person view and determine the advertising effectiveness criteria for each child advertisement according to the relevant reaction to each child advertisement from the one or more viewers, the suitable advertising population of each member advertisement A computer program having an executable form which is formed based on an effect standard and a child effect standard of each child advertisement. 8. The computer program according to claim 7, further causing the computing device to generate at least one of the plurality of child advertisements. 9. The computer program according to claim 7, wherein the computing device generates the plurality of child advertisement features based on an effect criterion of each member advertisement of the population. program. 10. The computer program according to claim 7, further comprising:
A computer program causing the computing device to replace the advertising population with the suitable advertising population, and to repeat the steps of claim 7. 11. The computer program according to claim 7, wherein the computing device provides each member advertisement as a selectable entity on a web page, and an effect criterion of each member advertisement is a count value regarding the member. A computer program, wherein the selectable entity is determined by incrementing it whenever it is selected. 12. The computer program according to claim 7, wherein the computing device further selects the effect criterion of each member advertisement, and the person who views the second count value regarding the member advertisement selects the selectable existence. A computer program characterized by deciding by incrementing whenever the relevant purchase is started after. 13. An evolutionary algorithm apparatus for providing a child advertisement based on an effect criterion for each parent advertisement based on the characteristics of the parent advertisement, and one or more viewers for the effect criterion for each parent advertisement. An evolutionary advertisement system comprising: an evaluation device that makes a decision based on the reaction from. 14. The evolutionary advertisement system according to claim 13, further comprising means for displaying each parent advertisement to one or more viewers. 15. The evolutionary advertisement system according to claim 13, further comprising a former that forms the child advertisement from child features based on the features of the parent advertisement.