Advertisement delivery data evaluation system based on multi-channel information integration

Abstract

The invention relates to the technical field of evaluation optimization, and discloses an advertisement delivery data evaluation system based on multi-channel information integration, and the system comprises a data standardization and fusion module which is responsible for receiving multi-source heterogeneous original data, carrying out the cleaning through a built-in data connector and an analysis engine, and carrying out the analysis of the multi-source heterogeneous original data; and forcibly converting all data into a unified atomic event model containing a unique device identifier, a standard timestamp and standard latitude and longitude coordinates. The technical effects of scientifically stripping interference factors and quantifying an online conversion net contribution value brought by offline advertisement putting in a causal inference mode are achieved by constructing an exposure group, generating a scientific control group by applying a tendency score matching algorithm and finally calculating a conversion improvement rate; the defect that the actual contribution of offline advertisements to online conversion cannot be accurately attributed and quantified due to the lack of reliable cross-screen tracking and scientific attribution methods is overcome.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of evaluation optimization, in particular to an advertisement delivery data evaluation system based on multi-channel information integration. BACKGROUND

[0002] The advertisement effect is the influence of advertisement activities or advertisement works on consumers; the narrow-sense advertisement effect refers to the economic effect achieved by the advertisement, and the degree to which the advertisement achieves the target is the communication effect and the sales effect generally included; in a broad sense, the advertisement effect includes the psychological effect and the social effect; the psychological effect is the influence degree of the advertisement on the psychological cognition, emotion and will of the audience, and is the concentrated embodiment of the communication function, economic function, education function and social function of the advertisement; the social effect of the advertisement is the influence of the advertisement on the social morality, cultural education, ethics and environment; a good social effect can also bring good economic benefits to the enterprise; the evaluation of the advertisement effect generally refers to the evaluation of the economic effect of the advertisement; the evaluation of the advertisement effect is to investigate the contact situation of consumers for various media such as newspapers, magazines, radio stations, televisions and outdoor advertisements.

[0003] For example, the Chinese invention application with the publication number CN110264241A discloses an advertisement delivery evaluation optimization system based on big data, which comprises an advertisement delivery data acquisition module, an advertisement delivery data arrangement module, a database, a central processing unit, a display screen, an advertisement delivery effect analysis module, an advertisement delivery evaluation module and a display screen; the advertisement delivery data acquisition module is used for acquiring data information of advertisement delivery; the data information includes click area information, browsing volume information, click volume information, population quantity information and account information of the person who clicks; the account information records the age information of the person who clicks.

[0004] According to the above-mentioned comparative documents, the present application provides an advertisement delivery data evaluation system based on multi-channel information integration, which has the following disadvantages in the prior art:

[0005] The people flow data, the advertisement exposure log and the user positioning event source from different systems, and the fundamental differences of the data structure, time unit and geographic coordinates cannot be integrated into a unified view for correlation analysis due to the lack of a standardized data fusion model in the prior art.

[0006] The geographic position of the outdoor screen is fixed and the playing period is different, while the audience population is dynamically changing in the urban space; the prior art cannot realize the fair comparison of the advertisement effect across screens and regions because it cannot map the discrete advertisement exposure events and the continuous population flow data on a unified space-time reference.

[0007] The anonymous audience of offline advertisement exposure and the online conversion behavior occurring on personal devices cannot be associated due to the lack of reliable cross-device identification and tracking methods in the prior art, so that the actual contribution of outdoor advertising to online conversion cannot be accurately attributed and quantified.

[0008] To this end, the present application proposes an advertisement delivery data evaluation system based on multi-channel information integration to solve the problems raised above. SUMMARY

[0009] In view of the deficiencies of the prior art, the present application provides an advertisement delivery data evaluation system based on multi-channel information integration to solve the problems raised in the background art.

[0010] To achieve the above purpose, the present application is implemented by the following technical solution: an advertisement delivery data evaluation system based on multi-channel information integration, comprising:

[0011] A data standardization and fusion module is responsible for receiving multi-source heterogeneous raw data, cleaning through the built-in data connector and analysis engine, forcibly converting all data into a unified atomic event model containing a unique device identifier, a standard timestamp and a standard latitude and longitude coordinate, and finally outputting a standard atomic event stream;

[0012] A spatio-temporal data fusion and indexing module is used to receive the output standard atomic event stream, map each standard atomic event to a unique spatio-temporal coordinate through spatial gridding and temporal slicing technology, and construct a spatio-temporal indexing database supporting efficient query;

[0013] An incremental conversion attribution analysis module is based on the constructed database, defines an exposure group according to the advertisement activity, generates a corresponding control group using a propensity score matching algorithm, and finally calculates and compares the conversion rate to obtain the conversion improvement rate of the quantified advertisement effect;

[0014] The incremental conversion attribution analysis module further comprises:

[0015] An exposure group defining unit is used to receive a geographical grid set and a time slice set of an advertisement activity, and based on the spatio-temporal indexing data output by the spatio-temporal data fusion and indexing module, extract all unique device identifiers that appear in the geographical grid set of the advertisement activity within the time slice set of the advertisement activity to form an exposure group;

[0016] A control group generating unit is used to receive the exposure group generated by the exposure group defining unit, and execute a propensity score matching algorithm to generate a control group, the algorithm first calculates the propensity score for the exposure group and candidate devices not exposed to the advertisement activity through a logistic regression model, the logistic regression model is:

[0017]

[0018] wherein p(X) is the probability of the device belonging to the exposure group, e is the base of the natural logarithm, X is a feature vector composed of preset user portrait covariates X i , β i are regression coefficients obtained through model training;

[0019] The algorithm is used for each device in the exposure group, and the device with the most similar matching propensity score in the candidate device is matched. The set of all matched devices constitutes the control group.

[0020] A conversion attribution calculation unit is configured to receive the generated exposure group and the control group generated by the control group generation unit, count the number of converted devices in a preset attribution period, and calculate the conversion lift rate by the following method:

[0021]

[0022] wherein Lift is the final calculated conversion lift rate, N C,E is the number of converted devices in the exposure group, |P E | is the total number of devices in the exposure group, N C,C is the number of converted devices in the control group, |P C | is the total number of devices in the control group.

[0023] A statistical significance test unit is configured to receive the calculated conversion lift rate and conversion data of each group, calculate a Z statistic using a two-sample proportion Z test to determine the statistical significance of the conversion lift rate, and the calculation formula of the Z statistic is:

[0024]

[0025] wherein Z is the calculated Z statistic, CR E is the conversion rate of the exposure group, CR C is the conversion rate of the control group, is the fusion conversion rate;

[0026] When the absolute value of the Z statistic is greater than the critical value corresponding to the preset significance level, it is determined that the conversion lift rate has statistical significance.

[0027] A data application and visualization module is configured to receive the output conversion lift rate and related statistical test results, and present the quantitative evaluation indicators to the end user through an application programming interface or a visualization dashboard.

[0028] Preferably, the data standardization and fusion module maps different source device identifiers to a unique device identifier in the system through an internal device ID atlas.

[0029] Preferably, the data standardization and fusion module converts non-precise location information such as base station ID or IP address into a standard latitude and longitude coordinate position by calling a geocoding service.

[0030] Preferably, the spatio-temporal data fusion and indexing module divides the geographic space into the geographic grid using the Geohash algorithm and divides the time axis into the time slices at fixed intervals.

[0031] Preferably, the incremental conversion attribution analysis module defines the exposure group in the following manner: extracting a set of all unique device identifiers that appear in the set of geographic grids of the advertising activity within the set of time slices of the advertising activity;

[0032] The incremental conversion attribution analysis module generates the control group in the following manner: from devices that have not been exposed to the advertising activity, filtering out a set of devices that have similarities with the exposure group in terms of pre-set user portrait covariates through a propensity score matching algorithm.

[0033] Preferably, the conversion lift rate is calculated in the following manner: first, calculating the conversion rates of the exposure group and the control group within a pre-set attribution period, and then dividing the difference between the conversion rate of the exposure group and the conversion rate of the control group by the conversion rate of the control group.

[0034] Preferably, the incremental conversion attribution analysis module performs a statistical significance test on the conversion lift rate to determine whether the conversion lift caused by the advertising delivery has statistical significance.

[0035] Preferably, the data standardization and fusion module further comprises:

[0036] A data access and analysis unit is configured to receive multi-source heterogeneous raw data through a pre-set data connector, call a parsing engine to extract fields from the raw data, and generate raw records containing original device identifiers, original time stamps, and original location information.

[0037] A data validity checking unit is configured to receive the generated raw records, execute a validity checking function based on each raw record, and output valid raw records that pass the validity checking. The judgment condition of the validity checking function is:

[0038] F valid (R raw )=(ID raw ≠NULL)∧IsParsable(Traw ) ∧ IsGeocodable(L raw ),

[0039] wherein, F valid is a validity check function, R raw is a single raw record, ID raw is a raw device identifier contained in the raw record, T raw is a raw timestamp contained in the raw record, L raw is raw location information contained in the raw record, IsParsable function is used to determine whether the format of this timestamp can be parsed by the system, and IsGeocodable function is used to determine whether this location information can be converted into geographic coordinates;

[0040] a core field standardization unit, configured to receive the valid raw records output, and call the following conversion functions to perform standardization conversion on the fields in the valid raw records, so as to generate a standard device identifier, a standard timestamp and a standard longitude and latitude coordinate:

[0041] a device identifier mapping function M(ID raw )→ID std , configured to map the raw device identifier to the standard device identifier unique in the system by querying a built-in device ID atlas;

[0042] a timestamp conversion function C(T raw )→T std , configured to uniformly convert the parseable raw timestamp into the standard timestamp in units of milliseconds of world coordinated time;

[0043] a geographic location conversion function G(L raw )→L std , configured to uniformly convert the geocodable raw location information into the standard longitude and latitude coordinate in WGS-84 coordinate system;

[0044] a standard event generation unit, configured to receive the generated standard device identifier, standard timestamp and standard longitude and latitude coordinate, encapsulate the standard fields into a preset unified atomic event model, and finally collect all generated unified atomic event models to constitute the standard atomic event stream.

[0045] Preferably, the space-time data fusion and indexing module further comprises:

[0046] a standard event receiving unit, configured to receive the standard atomic event stream, and distribute the standard atomic events one by one, wherein each of the standard atomic events contains a standard timestamp and a standard longitude and latitude coordinate;

[0047] a space dimension mapping unit, configured to receive the distributed standard atomic event, and invoke a space coding function to process the standard longitude and latitude coordinates contained in the standard atomic event, to generate a unique geographical grid ID, the space coding function being:

[0048] ID grid =GeohashEncode(L std ,P geo ),

[0049] wherein ID grid is the generated geographical grid ID, GeohashEncode is a standard Geohash algorithm function for encoding longitude and latitude coordinates, L std is the input standard longitude and latitude coordinates, and P geo is a preset geographical grid encoding parameter;

[0050] a time dimension mapping unit, configured to receive the standard atomic event to which the geographical grid ID is appended after processing, and invoke a time slicing function to process the standard timestamp contained in the standard atomic event, to generate a unique time slice ID, the time slicing function being:

[0051]

[0052] wherein ID slot is the generated time slice ID, is a floor function, T std is the input standard timestamp, and Δt is a preset fixed time length of a single time slice;

[0053] a space-time event indexing unit, configured to receive the standard atomic event to which the geographical grid ID and the time slice ID are appended after processing, and store a device identifier in the standard atomic event into an index database pointed by a composite key composed of the geographical grid ID and the time slice ID, to finally form the space-time indexing data.

[0054] Preferably, the data application and visualization module further comprises:

[0055] an attribution result receiving unit, configured to receive a conversion lift rate, a Z statistic and related conversion data output by the incremental conversion attribution analysis module, and encapsulate the received data into a structured attribution result object;

[0056] an evaluation index analysis and enhancement unit, configured to receive the encapsulated attribution result object, invoke a function to analyze and enhance calculation of data in the attribution result object, to generate an enhanced evaluation index containing a confidence interval:

[0057] The p value corresponding to the Z statistic is calculated first:

[0058] p = 2 x (1 - Φ (|Z|) ),

[0059] where p is the calculated p value, |Z| is the absolute value of the |Z| statistic, and Φ is the cumulative distribution function of the standard normal distribution;

[0060] The confidence interval of the difference in conversion rates between the exposure group and the control group is calculated:

[0061]

[0062] where CI is the calculated confidence interval, CR E and CR C represent the conversion rates of the exposure group and the control group, |P E | and |P C | are the total number of devices in the exposure group and the control group, z α / 2 is the critical value of the standard normal distribution at the preset confidence level z α / 2 ;

[0063] The terminal output generation unit receives the generated enhanced evaluation indicators and performs format processing according to a preset output target, which includes:

[0064] If the output target is an application programming interface, the enhanced evaluation indicators are serialized into a JSON format response body;

[0065] If the output target is a visual dashboard, the enhanced evaluation indicators are converted into a visual chart component, and a judgment conclusion is generated according to the comparison result of the p value and a preset significance level a, and the generation condition of the judgment conclusion is:

[0066]

[0067] where Conclusion is the final judgment conclusion presented to the user.

[0068] The data standardization and fusion module fundamentally solves the initial obstacle of multi-source heterogeneous data that cannot be fused and analyzed due to different standards by establishing a unified atomic event model. The advantage is to provide a single, reliable, and high-quality data entry for the system. By forcibly converting all disorganized raw data into standardized and consistent standard event streams, it ensures the data interoperability and consistency of all subsequent analysis links, and is a solid foundation for cross-channel data evaluation;

[0069] The spatio-temporal data fusion and index module ingeniously solves the industry problem that the dynamic changing crowd and the static advertising screen are not easy to accurately align by constructing a unified and discretized spatio-temporal coordinate system.

[0070] The incremental conversion attribution analysis module adopts a causal inference method based on a control experiment to provide a scientific and rigorous advertising effect measurement capability, and overcomes the limitation of a traditional attribution model that can only reveal correlation.

[0071] The data application and visualization module, as the final output layer of the system, is the "last mile" connecting complex data science and business decision-making.

[0072] The present application provides an advertising data evaluation system based on multi-channel information integration.

[0073] 1、The present application converts all received multi-source heterogeneous data into a unified atomic event model, thereby converting data of different sources, formats and definitions into a unified, regular and directly analyzable standard event stream.

[0074] 2、The present application accurately maps all standard atomic events into a unified and discretized spatio-temporal coordinate system through spatial gridding and time slicing technology, thereby accurately matching discrete advertising exposure events and continuous crowd flow data on a unified spatio-temporal reference.

[0075] 3、The present application constructs an exposure group, uses a tendency score matching algorithm to generate a scientific control group, and finally calculates the conversion promotion rate, so as to scientifically strip off interference factors and quantify the net contribution value of online conversion brought by offline advertising in a causal inference manner, compared with the prior art technical solution which relies on correlation analysis or lacks a control group, solves the problem that the actual contribution of offline advertising to online conversion cannot be accurately attributed and quantified due to the lack of reliable cross-screen tracking and scientific attribution method. BRIEF DESCRIPTION OF DRAWINGS

[0076] Figure 1 The system framework diagram of the present application. DETAILED DESCRIPTION

[0077] In order to enable personnel in the art to understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings of the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, other embodiments obtained by those of ordinary skill in the art without creative labor should be within the scope of protection of the present application.

[0078] The present application will be described in detail below in conjunction with the drawings:

[0079] Embodiment one: smart city outdoor digital media advertising effect evaluation

[0080] Please refer to the attached Figure 1 The present embodiment aims to evaluate the actual conversion promotion effect of outdoor digital media advertising of a coffee brand in the core business district of a city based on online applet coupon redemption behavior.

[0081] Data standardization and fusion:

[0082] Receive data: the data standardization and fusion module receives three types of heterogeneous raw data:

[0083] 1. User signaling data provided by a telecom operator, anonymized, containing device identifier, timestamp and base station location;

[0084] 2. Advertising screen media provider advertising play log, containing screen ID, advertising material ID and play time;

[0085] 3. Coupon redemption records provided by the coffee brand applet backend, containing user unique identifier, redemption time and coupon ID;

[0086] Standardization processing: the units in the module execute in turn:

[0087] The first data validity checking unit eliminates records lacking necessary information; next, the core field standardization unit uniformly maps the operator's device identifier and applet OpenID into a unique device identifier within the system, converts all times into UTC millisecond-level timestamps, and converts base station locations into WGS-84 standard latitude and longitude coordinates through a geocoding service; finally, the standard event generation unit outputs three types of standard atomic event streams: LOCATION PING , DOOH EXPOSURE , and ONLINE CONVERSION .

[0088] Spatiotemporal data fusion and indexing:

[0089] Spatiotemporal mapping: The spatiotemporal data fusion and indexing module receives the above standard atomic event streams. The module uses the Geohash algorithm with a precision of 7 to grid the city and time-slice it at 5-minute intervals;

[0090] LOCATION PING and DOOH EXPOSURE events are mapped into unique combinations and stored in the spatiotemporal index database;

[0091] Incremental conversion attribution analysis:

[0092] Circle the group: After the advertising campaign ends, the incremental conversion attribution analysis module starts. The exposure group circle defining unit extracts all matching device identifiers from the spatiotemporal index database according to the advertising plan to form the exposure group;

[0093] Generate control group: The control group generation unit selects a large number of devices that do not appear in the exposure geographical grid and time slice, uses the user's activity range entropy, historical business circle visit frequency, and online consumption preference label as covariates, calculates the propensity score through a logistic regression model, and matches the device with the closest score to form a control group similar in image to the exposure group;

[0094] Calculate and test: Within the 7-day attribution period after the advertisement ends, the conversion attribution calculation unit counts the number of applet coupons received by users:

[0095] Assuming that the exposure group has 100,000 people and 1,200 people receive, and the control group has 100,000 people and 500 people receive, the conversion improvement rate is calculated to be 140%; the post-statistical significance test unit calculates the Z statistic, and if the absolute value is greater than 1.96 (corresponding to a 95% confidence level), it is determined that this placement has brought statistically significant conversion improvement;

[0096] Data application and visualization:

[0097] Results presentation: the data application and visualization module receives the results of a 140% conversion rate increase and a significant test pass; this indicator is displayed in chart form on the end user's dashboard, with confidence intervals and p-values, while the map component displays the geographic distribution of the exposed population in the form of a heat map, providing decision-making basis for the brand to optimize future delivery points.

[0098] Example Two: New Retail Smart Store Precise Drainage Evaluation

[0099] This example aims to evaluate the digital guide screen advertising at the entrance of a large shopping center, and the drainage effect on the store traffic and member registration behavior of the specific brand;

[0100] Scenario setting: the shopping center is a sports brand store that has placed new product advertisements on the entrance screen;

[0101] Data sources:

[0102] 1. Customer flow data provided by Wi-Fi probes or Bluetooth beacon systems deployed by the shopping center;

[0103] 2. Ad play logs of the entrance large screen;

[0104] 3. Member registration system backend data and in-store customer flow counter data of the sports brand store;

[0105] Exposure definition: customers whose devices are captured by Wi-Fi probes in the vicinity of the entrance large screen during the ad play period.

[0106] Conversion definition:

[0107] 1. Within 1 hour after entering, the device is recognized by the customer flow counter or Wi-Fi probe at the entrance of the sports brand store;

[0108] 2. Within 24 hours after entering, complete brand online member registration;

[0109] Analysis process: the system circumscribes the exposure group that sees the advertisement at the entrance, and matches the control group from other entrance customers with similar portraits, calculates the conversion rate increase by comparing the store visit rate and member registration rate;

[0110] Beneficial effects: the shopping center can provide the brand with quantitative drainage effect data reports to prove the value of the advertising position; the brand can evaluate the actual effectiveness of offline advertising in attracting potential customers to the store and expanding private domain traffic accordingly.

[0111] Example Three: Precise Reach Evaluation of City Public Service Information

[0112] This example aims to evaluate the effect of the city government's public service announcement on the electronic screen of the bus shelter on the use rate of online reservation service by citizens;

[0113] Scenario setting: The city government rolls out the notice on the electronic screen of the smart bus shelter in the city, guiding citizens to make online reservations for residence permits;

[0114] Data sources:

[0115] 1. City population flow signaling data provided by telecom operators;

[0116] 2. The play log of each bus shelter electronic screen provided by the bus group;

[0117] 3. The online reservation record of residence permit handling in the municipal affairs service APP background;

[0118] Exposure definition: Citizens whose device signals appear in the specific geographic grid around the bus shelter during the exposure period;

[0119] Conversion definition: Within 3 days after exposure, successfully submit a residence permit handling reservation through the municipal affairs service APP;

[0120] Analysis process: The system identifies the exposure group that contacts the publicity information at the bus shelter, and matches the control group from the citizens with similar demographic characteristics in the city but not through the bus shelter. By comparing the online reservation rate, the conversion rate is calculated;

[0121] Beneficial effect: The city government can scientifically evaluate the social benefits of this public information placement, and according to the differences in conversion rate brought by bus shelters in different areas, identify areas with better publicity effect or demand, and optimize future public resource allocation and information release strategy.

[0122] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. An advertising placement data evaluation system based on multi-channel information integration, characterized in that, The advertising data evaluation system includes: The data standardization and fusion module is responsible for receiving multi-source heterogeneous raw data, cleaning it through the built-in data connector and parsing engine, forcibly converting all data into a unified atomic event model containing a unique device identifier, standard timestamp and standard latitude and longitude coordinates, and finally outputting a standard atomic event stream. The spatiotemporal data fusion and indexing module is used to receive the output standard atomic event stream. Through spatial gridding and temporal slicing technology, each standard atomic event is mapped to a unique spatiotemporal coordinate to build a spatiotemporal index database that supports efficient querying. The incremental conversion attribution analysis module, based on the constructed database, defines the exposure group according to the advertising campaign definition, generates the corresponding control group using a propensity score matching algorithm, and finally calculates and compares the conversion rate to obtain the conversion improvement rate of the advertising effect. The incremental conversion attribution analysis module further includes: The exposure group delineation unit is used to receive the geographic grid set and time slice set of the advertising campaign, and based on the spatiotemporal index data output by the spatiotemporal data fusion and indexing module, extract all unique device identifiers that appear in the geographic grid set of the advertising campaign within the time slice set of the advertising campaign, and form an exposure group. The control group generation unit receives the exposure group generated by the exposure group selection unit and executes a propensity score matching algorithm to generate a control group. The algorithm first calculates a propensity score for the exposure group and candidate devices that have not been exposed to the advertising campaign using a logistic regression model. The logistic regression model is as follows: Where p(X) is the probability that the device belongs to the exposure group, e is the base of the natural logarithm, and X is the probability of a device belonging to the exposure group based on a preset user profile covariate X. i The eigenvectors formed by β0 and β i These are the regression coefficients obtained through model training; The algorithm then describes the control group for each device in the exposure group, which consists of the candidate devices with the closest matching tendency score. The conversion attribution calculation unit is used to receive the generated exposure group and the control group generated by the control group generation unit, count the number of conversion devices within a preset attribution period, and then calculate the conversion improvement rate in the following manner: Where Lift is the final calculated conversion improvement rate, and N C，E |P represents the number of devices that underwent conversion within the exposure group. E | represents the total number of devices in the exposure group, N C,C |P represents the number of devices that underwent transformation in the control group. C | represents the total number of devices in the control group; The statistical significance testing unit receives the calculated conversion improvement rate and the conversion data of each group, and uses a two-sample proportion Z-test to calculate the Z-statistic to determine the statistical significance of the conversion improvement rate. The formula for calculating the Z-statistic is as follows: Where Z is the calculated Z-statistic, CR E For the conversion rate of the exposure group, CR C The conversion rate of the control group is shown. For fusion conversion rate; When the absolute value of the Z statistic is greater than the critical value corresponding to the preset significance level, the conversion improvement rate is determined to be statistically significant. The data application and visualization module is used to receive the output conversion improvement rate and related statistical test results, and present the quantitative evaluation indicators to the end user through the application programming interface or visualization dashboard.

2. The advertising placement data evaluation system based on multi-channel information integration according to claim 1, characterized in that, The data standardization and fusion module uses a built-in device ID map to uniformly map device identifiers from different sources to unique device identifiers within the system.

3. The advertising placement data evaluation system based on multi-channel information integration according to claim 1, characterized in that, The data standardization and fusion module converts imprecise location information such as base station IDs or IP addresses into standard latitude and longitude coordinates by calling a geocoding service.

4. The advertising placement data evaluation system based on multi-channel information integration according to claim 1, characterized in that, The spatiotemporal data fusion and indexing module uses the Geohash algorithm to divide the geographic space into geographic grids and divide the time axis into time slices at fixed intervals.

5. The advertising placement data evaluation system based on multi-channel information integration according to claim 1, characterized in that, The incremental conversion attribution analysis module identifies the exposure group by extracting the set of all unique device identifiers that appear in the geographic grid set of the advertising campaign within the time slice set of the advertising campaign. The incremental conversion attribution analysis module generates the control group by using a propensity score matching algorithm to select a set of devices that are similar to the exposure group in terms of preset user profile covariates from devices that have not been exposed to the advertising campaign.

6. The advertising placement data evaluation system based on multi-channel information integration according to claim 5, characterized in that, The conversion improvement rate is calculated as follows: first, the conversion rates of the exposure group and the control group within a preset attribution period are calculated, and then the difference between the conversion rates of the exposure group and the control group is divided by the conversion rate of the control group.

7. The advertising placement data evaluation system based on multi-channel information integration according to claim 6, characterized in that, The incremental conversion attribution analysis module is used to perform a statistical significance test on the conversion improvement rate to determine whether the conversion improvement brought about by the advertising campaign is statistically significant.

8. The advertising placement data evaluation system based on multi-channel information integration according to claim 1, characterized in that, The data standardization and fusion module further includes: The data access and parsing unit is used to receive multi-source heterogeneous raw data through a preset data connector, call the parsing engine to extract fields from the raw data, and generate a raw record containing the original device identifier, the original timestamp, and the original location information. A data validity verification unit is used to receive the generated original records, execute a validity verification function based on each original record, and output valid original records that pass the validity verification. The validity verification function has the following criteria: F valid (R raw )=(ID raw ≠NULL)∧IsParsable(T raw )∧IsGeocodable(L raw ), Among them, F valid For the validity check function, R raw For a single original record, ID raw T is the original device identifier contained in the original record. raw L is the original timestamp contained in the original record. raw The IsParsable function is used to determine whether the format of this timestamp can be parsed by the system, and the IsGeocodable function is used to determine whether this location information can be converted into geographic coordinates. The core field standardization unit receives the output valid original record and calls the following transformation function to perform standardization transformation on the fields in the valid original record to generate standard device identifiers, standard timestamps, and standard latitude and longitude coordinates: Device identifier mapping function M(ID) raw )→ID std This is used to map the original device identifier to a unique standard device identifier within the system by querying the built-in device ID graph; Timestamp conversion function C(T) raw →T std This is used to uniformly convert the parsable original timestamps into the standard timestamps in Coordinated Universal Time (UTC) millisecond units; Geographic location transformation function G(L) raw )→L std This is used to uniformly convert the geocoded original location information into the standard latitude and longitude coordinates of the WGS-84 coordinate system; The standard event generation unit is used to receive the generated standard device identifier, standard timestamp, and standard latitude and longitude coordinates, and encapsulate the standard fields into a preset unified atomic event model. Finally, all generated unified atomic event models are collected to form the standard atomic event stream.

9. The advertising placement data evaluation system based on multi-channel information integration according to claim 1, characterized in that, The spatiotemporal data fusion and indexing module further includes: A standard event receiving unit is used to receive the standard atomic event stream and distribute the standard atomic events one by one. Each standard atomic event contains a standard timestamp and standard latitude and longitude coordinates. A spatial dimension mapping unit is used to receive the distributed standard atomic events and call a spatial encoding function to process the standard latitude and longitude coordinates contained in the standard atomic events to generate a unique geographic grid ID. The spatial encoding function is: ID grid =GeohashEncode(L std ,P geo ), Among them, ID grid For the generated geographic grid ID, GeohashEncode is the standard Geohash algorithm function that encodes latitude and longitude coordinates. std P is the input standard latitude and longitude coordinates. geo These are the preset geographic grid coding parameters; The time dimension mapping unit is used to receive the standard atomic event with the geographic grid ID appended after processing, and to call the time slicing function to process the standard timestamp contained in the standard atomic event to generate a unique time slice ID. The time slicing function is as follows: Among them, ID slot For the generated time slice ID, T is the floor function. std The input is the standard timestamp, and Δt is the preset fixed duration of a single time slice; The spatiotemporal event indexing unit is used to receive the standard atomic event after processing, which is simultaneously appended with the geographic grid ID and the time slice ID, and to store the device identifier in the standard atomic event into the index database pointed to by the composite key formed by the geographic grid ID and the time slice ID, thus ultimately forming the spatiotemporal index data.

10. The advertising placement data evaluation system based on multi-channel information integration according to claim 1, characterized in that, The data application and visualization module further includes: The attribution result receiving unit is used to receive the conversion improvement rate, Z statistic and related conversion data output by the incremental conversion attribution analysis module, and to encapsulate the received data into a structured attribution result object. The evaluation metric parsing and enhancement unit is used to receive the encapsulated attribution result object, and call the following function to parse and enhance the data in the attribution result object to generate enhanced evaluation metrics containing confidence intervals: First, calculate the p-value corresponding to the Z-statistic: ρ=2×(1-Φ(|Z|)), Where ρ is the calculated p-value, |Z| is the absolute value of the |Z| statistic, and Φ is the cumulative distribution function of the standard normal distribution; Calculate the confidence interval for the difference in conversion rates between the exposure group and the control group: Where CI is the calculated confidence interval, and CR E With CR C |P represents the conversion rate between the exposure group and the control group. E |and|P C | represents the total number of devices in the exposure group and the control group, z α / 2 To achieve the preset information level z α / 2 The critical value of the standard normal distribution under the given conditions; A terminal output generation unit is configured to receive the generated enhanced evaluation metrics and perform formatting processing according to a preset output target. The formatting processing includes: If the output target is an application programming interface (API), the enhanced evaluation metrics are serialized into a JSON-formatted response body. If the output target is a visual dashboard, the enhanced evaluation index is converted into a visual chart component. Based on the comparison between the p-value and the preset significance level α, a judgment conclusion is generated. The conditions for generating the judgment conclusion are as follows: Here, Conclusion refers to the final judgment presented to the user.

Images