KR20230125913A - Causal inference analysis system and control method thereof - Google Patents

Abstract

The present invention relates to a causal reasoning analysis system and a control method thereof. A control method of a causal reasoning analysis system according to the present invention includes the steps of selecting a reference time point and an end time point corresponding to a specific event from collected data; selecting a plurality of subject groups that are roughly classified in relation to the specific event among data from the reference point in time to the end point in time; extracting a plurality of sampling target groups corresponding to a part of each of the plurality of selected target target groups; performing a random extraction homogeneity test on the basis of data prior to the reference point in time with respect to the extracted sampled subject groups and determining whether or not they pass; and performing a causal inference analysis related to the specific event based on data corresponding to the corresponding sampled subject groups when the random extraction homogeneity test is passed as a result of the determination.

Description

Causal inference analysis system and its control method {CAUSAL INFERENCE ANALYSIS SYSTEM AND CONTROL METHOD THEREOF}

The present invention relates to a causal reasoning analysis system and a control method thereof, and more particularly, to a system and a control method for performing a causal reasoning analysis after confirming homogeneity between two groups used for causal reasoning.

Nowadays, a lot of data is being collected due to the diversification of the Internet environment and data collection tools.

However, the data collected in this way does not mean much in itself, and it is necessary to know how to 'correctly' analyze the hidden context created by the data.

In particular, if causal inference is possible from data, it can be of great help in the business field.

Here, cause and effect are cause and effect. Inference is to draw conclusions by inferring and estimating based on 'existing facts'. In other words, causal inference is about two facts, "Does a causal relationship really exist?" It is a methodology for judging by evaluating "Are the two facts a cause and an effect, respectively?"

Here, 'causation' is different from 'correlation'.

As described above, the causal relationship means a relationship between cause and effect, and the correlation only means that there is a relationship between each other, but does not indicate a cause and effect.

There are various attempts to infer causal relationships.

The surest way to make causal inference is to compare the effect of the response according to treatment in the data obtained through a completely randomized controlled trial (RCT) from the design stage of the experiment.

For example, in terms of marketing, after running an advertisement, it is possible to analyze the extent to which the result of purchase is caused by the advertisement by analyzing the purchase (purchase rate, purchase amount, etc.) between the buyer who saw the advertisement and the buyer who did not see the advertisement. .

However, depending on the content of the experiment, such as medical experiments, random experiments cannot be conducted due to issues such as ethics and equity, or causal analysis is required using campaign data that has already been conducted in the past.

For such ex post causal inference, several tools using machine learning have recently been proposed.

However, in order to ensure reliability in the results of ex post causal relationship inference, the sampling of the subject must be appropriately performed.

For example, in the previous example, if there is a bias attribute between customers who viewed the advertisement and those who did not, other than whether or not they watched the advertisement (for example, if the age group of customers who viewed the advertisement is different from the age group of customers who did not view the advertisement). , the analysis of the purchase effect according to the advertisement is unreliable.

To do this, random sampling must be performed on objects with various attributes. Even in this case, if the number of attributes is small, unobserved attribute variables that can affect whether or not to view an advertisement (T) and purchase response (Y) Since the probability of existence is high, it cannot be considered that proper random sampling has been performed, and the results using the analysis tool are also unreliable.

In other words, the analysis result can be trusted only when homogeneity is recognized between the comparison objects except for whether or not the treatment to be analyzed (advertisement exposure in the example above) is recognized. there was no

Patent Publication No. 10-2009-0018806

The present invention has been devised to solve the above-mentioned conventional problems, and its purpose is causal reasoning that enables causal inference analysis to be performed by confirming homogeneity between comparison objects through a separate algorithm even when the number of attributes of comparison objects is small. It is to provide an analysis system and its control method.

In order to achieve the above object, the control method of the causal reasoning analysis system according to the present invention includes the steps of selecting a reference time point and an end time point corresponding to a specific event from collected data in the control method performed by the causal reasoning analysis system. and; selecting a plurality of subject groups that are roughly classified in relation to the specific event among data from the reference point in time to the end point in time; extracting a plurality of sampling target groups corresponding to a part of each of the plurality of selected target target groups; performing a random extraction homogeneity test on the basis of data prior to the reference point in time with respect to the extracted sampled subject groups and determining whether or not they pass; The method may include performing causal inference analysis related to the specific event based on data corresponding to the corresponding sampled subject groups when the random extraction homogeneity test is passed as a result of the determination.

Here, the specific event is a specific campaign targeting customers, and two target groups may be selected according to whether or not they have been exposed to the campaign.

Here, after calculating the analysis period from the reference time point to the end time point, a comparison end time point corresponding to a previous past time point corresponding to the analysis period from the reference time point is selected, and A random sampling homogeneity test may be performed based on data from the reference point in time to the comparison end point in time.

Here, the random extraction homogeneity test performed may be an A/A test using the same result as that of the specific event.

Here, if the determination result does not pass the random extraction homogeneity test, re-extracting a plurality of sampled subject groups corresponding to a part of each of the selected plurality of subject groups; Further comprising the step of performing a random extraction homogeneity test on the basis of the data prior to the reference point in time for each of the re-extracted plurality of sampled subject groups and determining whether the random extraction homogeneity test is passed as a result of the determination. Causal inference analysis related to the specific event may be performed based on data corresponding to the corresponding sampled subject groups.

Here, as a result of the determination, if the random extraction homogeneity test is not passed a predetermined number of consecutive times, the causal reasoning analysis for the specific event may be stopped.

In addition, the causal reasoning analysis system according to the present invention in order to achieve the above object, the analysis period selection unit for selecting a reference point in time and an end point corresponding to a specific event in the collected data; a group selector for selecting a plurality of subject groups that are roughly classified in relation to the specific event among data from the reference point in time to the end point in time; a sample group extraction unit for extracting a plurality of sampled subject groups corresponding to portions of each of the plurality of subject groups selected by the group selection unit; a sample verifying unit for performing a random extraction homogeneity test based on data prior to the reference point in time with respect to the sampled subject groups extracted from the sample group extracting unit and determining whether or not they pass; A causal reasoning analysis processing unit for performing a causal reasoning analysis related to the specific event based on data corresponding to the corresponding sampled subject groups when the random extraction homogeneity test is passed as a result of the determination of the sample checking unit may be included.

Here, the specific event is a specific campaign targeting customers, and the group selector may select two target groups according to whether or not they have been exposed to the campaign.

Here, the sample verifying unit calculates an analysis period from the reference time point to the end time point, selects a comparison end time point corresponding to a previous past time point by the analysis period from the reference time point, and the sample group extraction unit Random extraction homogeneity test may be performed on the basis of data from the reference time point to the comparison end point for each of the sampled subject groups extracted from .

Here, the random extraction homogeneity test performed by the sample verification unit may be an A/A test using the same result as the result of the specific event.

Here, the sample group extraction unit re-extracts a plurality of sampling target groups corresponding to a part of each of the plurality of target groups selected by the group selection unit when the random extraction homogeneity test is not passed as a result of the determination of the sample verifying unit, The sample verifier may perform a random extraction homogeneity test on the basis of data prior to the reference point in time with respect to the re-extracted plurality of sampled subject groups to determine whether or not to pass.

Here, the causal reasoning analysis processing unit performs causal reasoning for a specific event when a sampling target group that has passed the random extraction homogeneity test is not found until the sample group extraction unit and the sample verification unit are repeatedly performed more than a predetermined number of times. analysis can be discontinued.

As described above, according to the present invention, when the number of attributes of the object to be performed causal inference analysis is small or the amount of data itself is small so that the result of causal inference analysis is reliable, the homogeneity test for the extracted sampling data By performing in advance, the reliability of causal inference analysis can be ensured.

In particular, by performing a homogeneity test between each sampling data group related to a specific campaign through an A/A test using the same or similar past results as the response result of a specific campaign to data before the start of a specific campaign related to causal inference analysis The reliability of the homogeneity judgment itself can be increased.

1 is a functional block diagram of a causal inference analysis system according to an embodiment of the present invention;
2 and 3 are views for explaining data extracted or used by the causal inference analysis system of FIG. 1,
4 is an example of a result using a causal reasoning analysis tool;
5 is a control flow diagram of a causal inference analysis system according to an embodiment of the present invention.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

Hereinafter, each embodiment according to the present invention is only one example to aid understanding of the present invention, and the present invention is not limited to these embodiments. In particular, the present invention may be composed of at least one or more combinations of individual components, individual functions, or individual steps included in each embodiment.

In particular, although alphabets such as '(a)' are included in some claims of the claims for convenience, these alphabets do not prescribe the order of each step.

The causal reasoning analysis system 100 according to an embodiment of the present invention performs a function of performing causal reasoning analysis by analyzing pre-stored data and referring to a response to a specific event.

In particular, in this embodiment, it is assumed that the causal reasoning analysis system 100 has previously stored necessary data and performs causal reasoning analysis using the previously stored data.

Pre-stored data may include, for example, customer reactions to various events. For example, the causal reasoning analysis system 100 pays a specific coupon for marketing or exposes a specific advertisement to online customers, It is possible to analyze the effect of the specific coupon or specific advertisement on the purchase of customers by comparing the purchase results of customers exposed to the specific coupon or specific advertisement with the purchase results of customers who are not exposed to the specific coupon or advertisement.

An example of a specific functional block of the causal reasoning analysis system 100 is as shown in FIG. 1 .

As shown in the figure, the causal reasoning analysis system 100 includes an analysis period selection unit 110, a group selection unit 120, a sample group extraction unit 130, a sample verification unit 140, and a causal reasoning analysis processing unit. It may be configured to include (150).

The analysis period selection unit 110 performs a function of selecting a reference time point and an end time point corresponding to a specific event in the collected data.

As mentioned above, data for causal reasoning analysis is collected in advance, and the collected data may be stored in the causal reasoning analysis system 100's own storage (not shown), or a separate database server (not shown). shown), the analysis period selection unit 110 may be configured to include an interface capable of accessing its own storage or a separate database server.

Here, the specific event may correspond to various campaigns in the field of marketing conducted at a specific time in the past.

In this embodiment, it is taken as an example that the specific event is a coupon providing event that was performed as part of an online event during a specific period in the past.

For example, the causal reasoning analysis system 100 may provide a coupon to some of the customers accessing the online shopping mall and notify the fact through a pop-up window or the like.

The group selector 120 performs a function of selecting a plurality of subject groups that are roughly classified with respect to a specific event among data from a reference point in time to an end point in time.

In particular, the group selection unit 120 may select two target groups according to whether or not they have been exposed to a specific event (for example, a specific campaign).

For example, the group selection unit 120, when a coupon is provided to some customers during a specific period, the first coupon provided during the period during which the coupon is provided (ie, the period from the reference point to the end point). It is possible to select a group and a second group to which coupons are not provided.

In this embodiment, selecting two contrast groups is an example.

The sample group extraction unit 130 performs a function of extracting a plurality of sampled subject groups corresponding to a portion of each of the plurality of subject groups selected by the group selection unit 120 .

For example, the sample group extractor 130 may extract a 1-1 group that is a part of the first group and a 2-1 group that is a part of the second group.

Accordingly, for example, if the number of customers belonging to the first group is 1000, the 1-1 group may be 300 customers, which is a part of the 1000 customers. This is also the case of the second group and the 2-1 group.

A random extraction homogeneity test may be performed by the sample verifying unit 140 as will be described later on the sampling target group extracted by the sample group extracting unit 130. In this case, the sample group extraction unit 130 may re-extract a plurality of sampled subject groups corresponding to a portion of each of the plurality of subject groups selected by the group selection unit 120, and this process may be repeated a predetermined number of times. can

Meanwhile, the sample verification unit 140 performs a random extraction homogeneity test on the basis of data prior to the reference point in time for each of the sampled subject groups extracted by the sample group extraction unit 130, and determines whether or not they pass.

Here, the random extraction homogeneity test is to test whether homogeneity, such as random extraction, is maintained between predetermined contrast groups, and is to determine whether or not biased extraction has been performed.

For example, when a portion of people belonging to a specific group is extracted, the sample size must be large and random sampling (i.e., random sampling) must be performed in order for the extracted portion to be representative of the specific group. It should be homogeneous.

That is, the higher the homogeneity of the population, the smaller the sampling error.

Therefore, the homogeneity test of the extracted population is very important. In the present proposed invention, the sample verification unit 140 determines whether the groups of the sampled subjects all belong to a homogeneous group based on data before the reference point in time for each other.

In particular, the sample verifying unit 140 calculates the analysis period from the reference time point to the end time point, selects the comparison end time point corresponding to the previous past time point by the analysis period from the reference time point, and selects the sample group extractor ( 130), a random extraction homogeneity test may be performed for each sampled subject group based on data from the reference time point to the comparison end point.

For example, if the analysis period from the reference time point to the end time point is 1 week, the sample verifying unit 140 uses the data of the previous week from the reference time point to the sampling target group extracted by the sample group extraction unit 130. It is possible to perform a random extraction homogeneity test for each other.

At this time, the random extraction homogeneity test performed by the sample verifying unit 140 may be an A/A test using the same result as that of a specific event.

For example, when the result of a specific event is "whether a product is purchased or not", the sample verifying unit 140 uses the data of one week before this from the reference point in time to determine whether or not a product has been purchased for each sample group. A homogeneity test can be performed.

In this case, whether or not the homogeneity test is passed between the 1-1 group and the 2-1 group may be determined according to, for example, whether the "product purchase ratio" within each group is similar within a predetermined range.

In addition, as mentioned above, the sample verifying unit 140, even when the sampled subject groups are re-extracted by the sample group extracting unit 130, randomizes the re-extracted sampled subject groups based on data prior to the reference point in time. The extraction homogeneity test may be performed again to determine whether or not it passes, and this process may be repeated a predetermined number of times.

For reference, FIGS. 2 and 3 show examples of data extracted and used by the group selection unit 120 and the sample group extraction unit 130 of the causal reasoning analysis system 100 according to an embodiment of the present invention. .

First, referring to FIG. 2 , a start time and an end time of a specific event are shown, and progress data after the event is shown.

In addition, FIG. 2 shows data for comparison of homogeneity between groups, which, as shown, corresponds to data in the past from the start of the event.

That is, at least a part of post-event progress data is used for actual causal inference analysis processing, and at least a part of data for comparison of homogeneity between groups is used to determine appropriateness as data for causal inference analysis.

For convenience, the post-event data of FIG. 2 will be referred to as first region data, and the data for comparing homogeneity between groups will be referred to as second region data.

Figure 3 shows the data for comparison of homogeneity between these groups in more detail.

3(a) shows the first group and the second group in the data for comparison of homogeneity between groups.

Here, as mentioned in the previous example, the first group is past data (data before the reference point) corresponding to the target group to which coupons were provided, and the second group is past data (standard) corresponding to the target group to which coupons were not provided. data prior to the point in time).

3(b) shows the 1-1 group data and the 2-1 group data obtained by sampling some of the first group data and the second group data.

On the other hand, the causal reasoning analysis processing unit 150 performs a causal reasoning analysis related to a specific event based on data corresponding to the corresponding sampling subject groups when the random extraction homogeneity test is passed as a result of the determination of the sample confirmation unit.

For example, as shown in FIG. 3(b), if the homogeneity test is passed by comparing the homogeneity of the 1-1 group data and the 2-1 group data sampled from data before the reference point, causal inference analysis The processing unit 150 performs causal inference analysis on the first area data of the subject belonging to the 1-1 group and the first area data of the subject belonging to the 2-1 group.

If there are two groups that have passed the random extraction homogeneity test, the causal inference analysis process performed after that directly compares the difference between the two groups in response to a specific event (Y value: response rate, purchase amount, number of purchases, etc.) Various well-known methods such as a method of obtaining an average treatment effect (ATE) or a method of estimating a conditional average treatment effect (CATE) using a meta-learner may be used.

For example, in the case of performance analysis using marketing campaign result data, CATE is estimated from the extracted dataset, and performance comparison by customer group, comparison of influence on response (Y) by customer attribute (X), or uplift analysis of target customer group with optimal cost offer is performed. can see.

For reference, FIG. 4 is an example of such result analysis.

In particular, the causal reasoning analysis processing unit 150 is configured to perform the sample group extraction unit 130 and the sample verification unit 140 repeatedly until a predetermined number of times or more is performed, when the sampling subject group that has passed the random extraction homogeneity test is not found. You can discontinue causal inference analysis for that particular event.

For example, if the homogeneity test is not passed five times or more consecutively as a result of the test performed by the sample verifying unit 140, further processing may be stopped and an error message may be output.

Hereinafter, the overall control flow of the causal reasoning analysis system 100 according to an embodiment of the present invention will be described with reference to FIG. 5 .

The causal reasoning analysis system 100 selects a specific campaign start time and end time with reference to pre-stored data (step S1).

Subsequently, the causal reasoning analysis system 100 selects a contrast group using data between the above-described specific campaign time and end time.

For example, customers exposed to a campaign of providing coupons may be selected as a first group, and customers who are not exposed to a campaign may be selected as a second group (step S3).

Thereafter, the causal reasoning analysis system 100 selects a sampling target group from the first group and the second group (step S5), for example, selects some of the customers belonging to the first group as the 1-1 group, Some of the customers belonging to the second group are selected as the 2-1 group.

In addition, the causal reasoning analysis system 100 performs a homogeneity test between sampled subject groups using the result data before the start of the campaign (step S7).

For example, an A/A test is performed on the data stored before the start of the campaign for the subjects belonging to the 1-1 group (coupon provision time) and the data stored before the start of the campaign for the subjects belonging to the 2-1 group. It is to determine whether the subjects belonging to the 1-1 group and the 2-1 group are the same by performing.

For example, if the subjects belonging to the 1-1 group and the subjects belonging to the 2-1 group showed the same level of purchasing patterns in the past, the causal reasoning analysis system 100 determines that the homogeneity between the groups is maintained. can

At this time, instead of the purchase pattern, homogeneity may be determined based on the same result as the result of the campaign to make causal inference. Homogeneity between the 1-1 group and the 2-1 group is limited only when the purchase amount of the subjects belonging to the 1-1 group and the 2-1 group in the past (ie, before the time of coupon provision) was similar. can be judged to be maintained.

When it is determined that homogeneity between groups is maintained (step S9), the causal reasoning analysis system 100 performs a causal reasoning analysis (step S11), that is, the attributes of subjects belonging to the 1-1 group and the 2-1 group Causal inference analysis may be performed using the attributes of the target audience and the purchase amount generated by the target audience from the start of the campaign to the end of the campaign.

For example, if the attributes (X) of each customer include age, gender, and address, each customer's age, gender, address, campaign exposure (applied) (T), product purchase amount (reaction result Y ) can be considered to perform causal inference analysis.

Here, since performing causal inference analysis using the attributes (X) of each subject, whether a campaign is applied (T), and the response result (Y) per se corresponds to a known technique, a detailed description thereof will be omitted.

If it is determined that the homogeneity between groups is not maintained (step S9), the causal reasoning analysis system 100 re-selects the sampling subject group (step S15) if the predetermined number of iterations is not exceeded (step S13), and performs a homogeneity test for this group. (Step S7), and this process may be repeated until the predetermined number of times is exceeded (Step S13).

If the inter-group homogeneity test is not passed until the number of iterations exceeds the preset number, the causal reasoning analysis system 100 generates an error and ends the causal reasoning analysis (step S17).

Meanwhile, it goes without saying that the process of performing each of the above-described embodiments may be performed by a program or application stored in a predetermined recording medium (for example, computer-readable). Here, the recording medium includes all of an electronic recording medium such as RAM (Random Access Memory), a magnetic recording medium such as a hard disk, an optical recording medium such as a CD (Compact Disk), and the like.

At this time, the program stored in the recording medium may be executed on hardware such as a computer or smart phone to perform each of the above-described embodiments. In particular, at least one of the functional blocks of the causal reasoning analysis system according to the present invention described above may be implemented by such a program or application.

In addition, the present invention is not limited to the specific embodiment described above, but can be implemented by various modifications and variations within the scope of the present invention. It will be apparent that such variations and modifications are included in the present invention provided they come within the scope of the appended claims.

100: causal inference analysis system 110: analysis period selection unit
120: group selection unit 130: sample group extraction unit
140: sample verification unit 150: causal reasoning analysis processing unit

Claims (14)

In the control method performed by the causal reasoning analysis system,
(a) selecting a reference time point and an end time point corresponding to a specific event from the collected data;
(b) selecting a plurality of subject groups that are broadly classified in relation to the specific event among data from the reference point in time to the end point in time;
(c) extracting a plurality of sampled subject groups corresponding to a portion of each of the plurality of subject groups selected in step (b);
(d) performing a random extraction homogeneity test based on data prior to the reference point in time with respect to the sampled subject groups extracted in step (c) and determining whether or not they pass;
(e) performing a causal inference analysis related to the specific event based on data corresponding to the corresponding sampled subject groups when the random extraction homogeneity test is passed as a result of the determination in step (d). Control method of causal reasoning analysis system. According to claim 1,
The specific event is a specific campaign targeting customers,
The control method of the causal reasoning analysis system, characterized in that in the step (b), two target groups are selected according to whether they are exposed to the campaign. According to claim 1,
In the step (d), after calculating the analysis period from the reference time point to the end time point, a comparison end time point corresponding to a previous past time point by the analysis period from the reference time point is selected, and in the step (c) A control method of a causal inference analysis system, characterized in that for performing a random extraction homogeneity test based on data from the reference time point to the comparison end point for each of the sampled subject groups extracted from. According to claim 1,
The control method of the causal reasoning analysis system, characterized in that the random extraction homogeneity test performed in step (d) is an A / A test using the same result as the result for the specific event. According to claim 1,
(f) re-extracting a plurality of sampled subject groups corresponding to a part of each of the plurality of subject groups selected in step (b) when the random extraction homogeneity test is not passed as a result of the determination in step (d); and ;
(f) performing a random extraction homogeneity test on the basis of data prior to the reference point in time with respect to the re-extracted plurality of sampled subject groups to determine whether they pass,
In the step (e), when the random extraction homogeneity test is passed as a result of the determination in step (f), causal inference analysis related to the specific event is performed based on data corresponding to the corresponding sampled subject groups. Control method of reasoning analysis system. According to claim 5,
In the step (e), if the random extraction homogeneity test is not passed consecutively for more than a predetermined number of times as a result of the determination in the step (f), the control of the causal reasoning analysis system is characterized in that to stop the causal reasoning analysis for the specific event. method. A computer-readable recording medium recording a program for executing the method of any one of claims 1 to 6. An application program stored in a computer readable recording medium to be combined with hardware to execute the method of any one of claims 1 to 6. an analysis period selector for selecting a reference time point and an end time point corresponding to a specific event in the collected data;
a group selector for selecting a plurality of subject groups that are broadly classified in relation to the specific event among data from the reference point in time to the end point in time;
a sample group extraction unit for extracting a plurality of sampled subject groups corresponding to portions of each of the plurality of subject groups selected by the group selection unit;
a sample verifying unit for performing a random extraction homogeneity test based on data prior to the reference point in time with respect to the sampled subject groups extracted from the sample group extracting unit and determining whether or not they pass;
Causal inference analysis processing unit for performing causal inference analysis related to the specific event based on data corresponding to the corresponding sampling subject groups when the random extraction homogeneity test is passed as a result of the determination of the sample confirmation unit. analysis system. According to claim 9,
The specific event is a specific campaign targeting customers,
The causal reasoning analysis system, characterized in that the group selection unit selects two target groups according to whether or not they are exposed to the campaign. According to claim 9,
The sample verifying unit calculates the analysis period from the reference time point to the end time point, selects a comparison end time point corresponding to a previous past time point by the analysis period from the reference time point, and extracts the data from the sample group extraction unit. A causal inference analysis system, characterized in that for performing a random extraction homogeneity test based on data from the reference time point to the comparison end point for each of the sampled subject groups. According to claim 9,
Causal inference analysis system, characterized in that the random extraction homogeneity test performed by the sample verification unit is an A / A test using the same result as the result for the specific event. According to claim 9,
The sample group extraction unit re-extracts a plurality of sampling target groups corresponding to a part of each of the plurality of target groups selected by the group selection unit when the random extraction homogeneity test is not passed as a result of the determination of the sample verifying unit,
The sample verification unit performs a random extraction homogeneity test based on data before the reference point in time with respect to the re-extracted plurality of sampled subject groups to determine whether or not they pass. According to claim 13,
The causal inference analysis processing unit performs causal inference analysis on a specific event when no sampled subject group that has passed the random extraction homogeneity test is found until the sample group extraction unit and the sample verification unit are repeatedly performed more than a predetermined number of times. A causal inference analysis system, characterized in that to stop.