CN111951038A - Feature processing method and device for advertisement monitoring data - Google Patents
Feature processing method and device for advertisement monitoring data Download PDFInfo
- Publication number
- CN111951038A CN111951038A CN202010655582.2A CN202010655582A CN111951038A CN 111951038 A CN111951038 A CN 111951038A CN 202010655582 A CN202010655582 A CN 202010655582A CN 111951038 A CN111951038 A CN 111951038A
- Authority
- CN
- China
- Prior art keywords
- feature
- model
- advertisement
- features
- feature processing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012544 monitoring process Methods 0.000 title claims abstract description 27
- 238000003672 processing method Methods 0.000 title claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 57
- 230000009466 transformation Effects 0.000 claims abstract description 9
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 238000009795 derivation Methods 0.000 claims abstract description 6
- 230000009467 reduction Effects 0.000 claims abstract description 6
- 230000005477 standard model Effects 0.000 claims description 25
- 238000013507 mapping Methods 0.000 claims description 15
- 230000001131 transforming effect Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 description 26
- 238000010801 machine learning Methods 0.000 description 13
- 238000000034 method Methods 0.000 description 13
- 230000000694 effects Effects 0.000 description 10
- 230000006399 behavior Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 230000008901 benefit Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000013145 classification model Methods 0.000 description 3
- 230000003796 beauty Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000003203 everyday effect Effects 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 229920004934 Dacron® Polymers 0.000 description 1
- 230000003321 amplification Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 239000008267 milk Substances 0.000 description 1
- 210000004080 milk Anatomy 0.000 description 1
- 235000013336 milk Nutrition 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 239000005020 polyethylene terephthalate Substances 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0242—Determining effectiveness of advertisements
- G06Q30/0245—Surveys
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Finance (AREA)
- General Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Software Systems (AREA)
- Accounting & Taxation (AREA)
- Evolutionary Biology (AREA)
- Entrepreneurship & Innovation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a feature processing method and a device of advertisement monitoring data, wherein the feature processing method comprises the following steps: the extraction step comprises: extracting basic original features from advertisement monitoring data; the processing steps are as follows: and performing derivation and transformation reduction on the basic original features to map the fine-grained feature dimension in the basic original features to the coarse-grained feature dimension.
Description
Technical Field
The present invention relates to a method and an apparatus for processing characteristics of advertisement monitoring data, and more particularly, to a method and an apparatus for processing characteristics of advertisement monitoring data for a mobile terminal.
Background
Machine learning is an algorithm that automatically improves the effect of itself by a machine. The most basic and important item in machine learning is feature extraction, and how to define features, how to process the features, and how to remove irrelevant data and redundant data, so as to improve the efficiency of machine learning and optimize the effect of machine learning all depend on feature engineering for extracting the features. Selecting the features important for the result is a very important and very complicated step in feature engineering, training a model on the basis of continuously selecting and converting the features, and measuring the quality of each group of selected features according to indexes calculated by the model.
In the advertisement marketing industry, different target crowds have different tendency degrees to different time, different terminals, different media, different types of commodities and the like, and the factors comprehensively influence the browsing and clicking behaviors of the user on the advertisement and finally influence the conversion effect. Therefore, it is important to find out features from the features which can really affect the division of the target population, and certain requirements on performance indexes can be met only by using models trained by the strong features as much as possible.
Different data characteristics in the mobile terminal advertisement monitoring data are different in selection and processing modes, data conversion can be really realized only by fully mining the meaning of each field, useful characteristics are derived and mapped, and finally the efficiency and the precision of a subsequent machine learning model are improved.
The most common way is to manually select the useful features for the result, and adjust the features through continuous trial and error, and finally train out the model. The more the types of features are, the more a person can be more carefully and comprehensively depicted. The common mode is to reserve the relevant features in the data, directly remove the irrelevant features, and perform some processing on some feature dimensions to become the features which can be directly used by the subsequent model. These features are used for training and final prediction of the machine learning model.
The prior art is not enough with the following points:
1. the selected characteristic dimension has a great number of values, which can cause dimension disaster;
2. too many features, more difficult model training, slow training speed and low precision, and the training, testing and storage expenses caused by the increase of the number of features are increased;
3. too many features cause the difference between the feature distribution of the training sample and the feature distribution of the predicted data to be too large, the result of prediction is poor, and many data cannot be classified.
Therefore, it is desirable to develop a feature processing method and apparatus for advertisement monitoring data of a mobile terminal that overcomes the above-mentioned drawbacks.
Disclosure of Invention
In view of the above problems, the present invention provides a method for processing characteristics of advertisement monitoring data, wherein the method comprises:
the extraction step comprises: extracting basic original features from advertisement monitoring data;
the processing steps are as follows: and performing derivation and transformation reduction on the basic original features to map fine-grained feature dimensions in the basic original features to coarse-grained feature dimensions.
The feature processing method described above, wherein the basic primitive features include: advertisement site ID characteristics, log type characteristics, timestamp characteristics, mobile device model characteristics, and region ID characteristics.
The feature processing method described above, wherein the processing step includes:
and a timestamp characteristic processing step: reducing the feature dimension of the timestamp feature through mapping to obtain a final timestamp feature;
mobile equipment model feature processing: associating brand information of the mobile equipment model according to the original model parameters of the mobile equipment returned by the trusted media;
advertisement site location ID feature processing step: and transforming the advertisement site location ID characteristics to obtain associated characteristics related to the advertisement site location ID characteristics.
In the above feature processing method, the step of processing the mobile device model features further includes:
a statistical step: counting the original model parameters of the mobile equipment returned by the trusted media;
selecting: selecting a standard model corresponding to the standard model parameter from the original model parameters according to a preset rule;
selecting: and associating brand information of the standard model according to the standard model.
The above feature processing method, wherein the associated features include: goods, brands, advertisers, and industry categories.
The invention also provides a feature processing device of the advertisement monitoring data, which comprises:
the extracting unit is used for extracting basic original characteristics from the advertisement monitoring data;
and the processing unit is used for deriving and transforming the basic original features to reduce the mapping of fine-grained feature dimensions in the basic original features to coarse-grained feature dimensions.
The feature processing apparatus described above, wherein the basic primitive features include: advertisement site ID characteristics, log type characteristics, timestamp characteristics, mobile device model characteristics, and region ID characteristics.
The above feature processing device, wherein the processing unit includes:
the time stamp feature processing module is used for reducing the feature dimension of the time stamp feature through mapping to obtain a final time stamp feature;
the mobile equipment model feature processing module is used for correlating brand information of the mobile equipment model according to the original model parameters of the mobile equipment returned by the trusted media;
and the advertisement site location ID feature processing module is used for transforming the advertisement site location ID feature to obtain the associated feature related to the advertisement site location ID feature.
In the above feature processing apparatus, the mobile device model feature processing module performs statistics on the original model parameters of the mobile device returned by the trusted media, selects a standard model corresponding to the standard model parameters from the original model parameters according to a preset rule, and associates brand information of the standard model according to the standard model.
The above feature processing apparatus, wherein the associated feature includes: goods, brands, advertisers, and industry categories.
In summary, compared with the prior art, the invention has the following effects: the invention tries to use the technical means of the feature engineering aiming at the mobile terminal advertisement monitoring data, and directly utilizes the feature relation mapping table to map some fine-grained feature dimensions to other feature dimensions, so that the feature dimensions have less feature space and are directly related to the original features, and the speed and the precision of a subsequent machine learning model can be greatly improved.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
FIG. 1 is a flow chart of a feature processing method of the present invention;
FIG. 2 is a flowchart illustrating the substeps of step S2 in FIG. 1;
FIG. 3 is a flowchart illustrating the substeps of step S22 in FIG. 2;
FIG. 4 is a schematic diagram of a feature processing apparatus according to the present invention.
Wherein the reference numerals are;
a name setting unit: 11
A name determination unit: 12
A marking unit: 13
A determination unit: 14
A name determination module: 121
A supplement determination module: 122
A first labeling module: 131
A second labeling module: 132
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As used herein, the terms "comprising," "including," "having," "containing," and the like are open-ended terms that mean including, but not limited to.
References to "a plurality" herein include "two" and "more than two".
Each advertiser can put a plurality of advertisements of different commodities in different time intervals, different regions, different media, different terminals and the like every day, the same advertiser can also have the same commodities with different brand lines, the number of the advertisements put in each day is very large, and if the advertisement site is simply used (the advertisement site uniquely identifies one advertisement), the number of the extracted characteristics is very large within a certain time range. In order to solve the balance problem of 'feature quantity' and 'model efficiency and accuracy', an original feature is tried to be mapped to other associated features through an advertisement point location association relation and a feature mapping function, the original feature is a point of a high-dimensional space, and the feature dimension after mapping is reduced, so that the effect of dimension reduction is achieved.
Referring to fig. 1-3, fig. 1 is a flow chart of a feature processing method according to the present invention; FIG. 2 is a flow chart of the substeps of FIG. 1; FIG. 3 is a flow chart of the substeps of FIG. 2. As shown in fig. 1 to 3, the feature processing method of the present invention includes: the method comprises the following steps:
extraction step S1: extracting basic original features from advertisement monitoring data;
processing step S2: and performing derivation and transformation reduction on the basic original features to map fine-grained feature dimensions in the basic original features to coarse-grained feature dimensions.
Wherein the basic primitive features include: advertisement site ID characteristics, log type characteristics, timestamp characteristics, mobile device model characteristics, and region ID characteristics.
Further, the processing step S2 includes:
timestamp feature processing step S21: reducing the feature dimension of the timestamp feature through mapping to obtain a final timestamp feature;
mobile device model feature processing step S22: associating brand information of the mobile equipment model according to the original model parameters of the mobile equipment returned by the trusted media;
advertisement site ID feature processing step S23: and transforming the advertisement site location ID characteristics to obtain associated characteristics related to the advertisement site location ID characteristics.
It should be noted that the present invention does not limit the sequence of step S21, step S22 and step S23.
Further, the mobile device model feature processing step further includes:
a statistical step S221: counting the original model parameters of the mobile equipment returned by the trusted media;
selection step S222: selecting a standard model corresponding to the standard model parameter from the original model parameters according to a preset rule;
selection step S223: and associating brand information of the standard model according to the standard model.
Wherein the associated features include: goods, brands, advertisers, and industry categories.
The present invention will be described in detail with reference to the following embodiments:
to extract features useful for subsequent machine learning from mobile-side advertisement monitoring data, the data needs to be sufficiently known.
Starting with the advertisement monitoring data of the mobile terminal, finding out which features promote the effect of the machine learning model, and listing five types of basic features based on the advertisement monitoring log of the mobile terminal as follows:
SPID (ad spot id feature): each advertisement point location corresponds to an id, and an advertisement is uniquely identified;
log _ type (log type feature): two values may be taken, imp or clk, imp representing exposure, i.e. the user has viewed the advertisement at that spot; clk represents a click, i.e., the user has clicked on the advertisement at that point;
timestamp (timestamp feature): the time stamp represents the time point when the user generates an advertisement behavior, and the time can describe the behavior activity characteristics of the user and is also an influence factor of the user attribute;
device _ model (mobile device model feature): the mobile equipment used by one user can also reflect certain attributes of the user, such as a business mobile phone, a music mobile phone, a female mobile phone, an old mobile phone and the like, and different mobile equipment models also reflect different user characteristics;
region _ id (region id feature): the advertisement monitoring log is provided with fields of region ids, each region id uniquely identifies a province, a city or a county and the like, and different region levels can indirectly reflect the attributes of the users.
The five types of basic data are directly collected from the original mobile terminal advertisement monitoring data without any processing. If the original features are directly used for training an algorithm of machine learning, certain effect requirements can be met, but from the business logic, the accuracy and the calculation speed of model calculation need to be improved, and therefore, the operations of derivation, transformation and the like on the features are needed.
The following describes the operations of performing different transformations on the above five basic features:
log _ type: only two values are taken, so that transformation is not needed;
region _ id: the province, the city or the county of the country are determined and can be enumerated, and the enumeration space is not large, so the region _ id does not need to be transformed, and can be directly used as the characteristic of the model;
time estimate: the time stamp refers to the total number of seconds from 00 minutes 00 seconds to 00 minutes 00 seconds at 1970, 01/00, and the feature space is particularly huge if we directly use the time stamp as a feature, so that it is not feasible to directly use the time stamp as a feature. The time stamps need to be mapped. Considering that a person's advertising behavior varies from monday to friday and weekend, and at certain days, at different time points, we map any one timestamp to the corresponding week and one of the 24 hours of each day in order to ensure that the feature space after mapping is small. Such as: 1591156734, the timestamp represents 2020-06-0311: 58:54, and we map this timestamp to 3_11, which represents 11 points on wednesday, so that the whole timestamp can be mapped to a feature space with size of 7 × 24, and a good dimension reduction effect is achieved;
device _ model: the existing mobile equipment has various models and brands, and different media are transmitted back to the same mobile equipment to be different models. Under the condition of the data quality, the models returned by the trusted media are counted, the model with the largest occurrence frequency is selected as the standard model of the mobile equipment, and the brand information of the mobile model is associated;
SPID: the basic feature of the advertisement site is the most important feature, and the tendency of each mobile device to browse and click different advertisements directly reflects the behavior tendency of the user. However, because each advertiser can put advertisements of different types of commodities on different media, the feature space corresponding to the advertisement site is particularly large, and if the features are directly used, the total number of the features of each model training can reach ten thousand to one hundred thousand. On one hand, if the number of the features is too large, the model training speed is slowed down; on the other hand, too many features cause that each feature has a certain influence on the result of prediction amplification, interference exists among the features, and the effect of the classifier trained by the final model is also poor.
Therefore, we need to convert the advertisement spots with the finest granularity to the feature dimensions with the coarse granularity, because any advertisement spot has its corresponding advertiser, the media placed by the advertiser and the commodity information corresponding to the advertisement, etc., and the advertisement spots with the finest granularity can be described by using some coarse-granularity dimensions through the advertisement spot association relationship updated every day, for example:
for example, an advertisement spot is a decimal number, the corresponding advertisement name is a publicity language related to milk powder, the advertiser corresponding to the advertisement is a dacron, the advertisement is put on a hundredth medium, the brand corresponding to the advertisement is love in other beauty, and the commodity is a harbor edition love in other beauty.
We have several benefits of mapping fine-grained base features to other associated feature dimensions:
1. the number of advertisement point positions with the finest granularity is extremely large, the feature space is extremely large, after the advertisement point positions are mapped to other associated feature dimensions, the feature space is greatly reduced, and the subsequent model training speed can be greatly improved;
2. if a model is trained by using the finest granularity of the advertisement spot A, the model uses the characteristic that the advertisement spot on the training sample match is A. When the model is used for prediction, a new mobile device may not have the advertisement spot a in the advertisement spot data set, and finally, an accurate classification is difficult to predict. However, if the basic features such as the advertisement spot a are mapped to other associated features, such as the category industry, brand, advertiser, etc. of the goods corresponding to the advertisement spot a, the associated features may play a role in the prediction link.
Examples are: both SPID _1 and SPID _2 are advertisements placed by brand B under advertiser a, and if only SPID is used as a feature, in these mobile devices of the training sample, the behavior is only over-represented under the spot of SPID _1, that is, in the model training, the model calculates the weight coefficient of SPID _ 1. However, in the prediction process, if a mobile device generates advertising behavior under SPID _2, but the feature does not appear in the training set, the feature cannot play any role in the final prediction; however, if we transform the features to map SPIDs to both advertisers and media, SPID _1 will be mapped to advertiser a and brand B during the training process, and during the model training process, we will calculate the weights of advertiser a and brand B, and in the final prediction process, a mobile device will generate ad behavior under SPID _2, and although SPID _2 does not appear in the training sample set, SPID _2 will also be mapped to advertiser a and brand B through the feature mapping, so this feature can be applied when predicting this mobile device, and since both spots are under one advertiser and brand, the mobile devices browsing and clicking these two spots should have similar attributes.
Therefore, the advertisement point positions are mapped to the characteristic dimensions of advertisers, brands, commodities, media and the like, the values of the characteristic dimensions are few, the characteristics are more concentrated, the predicted characteristic distribution of the mobile equipment is closer to the characteristic distribution of a training set in the prediction process, the assumption of independent and same distribution can be better met, and the accuracy of the obtained result is higher.
In addition, the characteristics processed by the method are tested on the mobile terminal gender classification model. Using the same seed population package, male seeds and female seeds (in the millions), the following two protocols were practiced:
the first scheme is that the five basic original characteristics without mapping change are used
1. Feature extraction: performing feature matching on each device _ id in the seed sample, wherein the number of features is in the order of one hundred thousand from the aspect of a feature matching result;
2. model training: after the characteristics are matched, inputting the samples with matched characteristics into a classification model for model training, wherein the training time is about 3 hours;
3. classification prediction and accuracy verification: from the final prediction results, the accuracy is about 60%.
Scheme two, the characteristics of the derivative transformation using the five basic original characteristics
1. Feature extraction: performing feature matching on each device _ id in the seed sample, wherein thousands of features can be matched from the result of the feature matching;
2. model training: after the characteristics are matched, inputting the samples with matched characteristics into a classification model for model training, wherein the training can be completed within 1 hour;
3. classification prediction and accuracy verification: from the final prediction results, the accuracy is about 76%.
It can be seen from the above two schemes that the characteristics of the derivative changes are used in the mobile terminal advertisement monitoring data, so that the training of the machine learning model can be shortened, and the prediction accuracy can be improved.
Referring to fig. 4, fig. 4 is a schematic structural diagram of a feature processing apparatus according to the present invention. As shown in fig. 4, the feature processing apparatus for advertisement monitoring data according to the present invention includes:
an extraction unit 11 that extracts basic original features from advertisement monitoring data;
and the processing unit 12 is used for deriving and transforming the basic original features to reduce and map the fine-grained feature dimensions in the basic original features to the coarse-grained feature dimensions.
Wherein the basic primitive features include: advertisement site ID characteristics, log type characteristics, timestamp characteristics, mobile device model characteristics, and region ID characteristics.
Further, the processing unit includes:
the timestamp feature processing module 121 is configured to obtain a final timestamp feature by reducing the feature dimension of the timestamp feature through mapping;
the mobile device model feature processing module 122 correlates brand information of the mobile device model according to the original model parameter of the mobile device returned by the trusted media;
and the advertisement site location ID feature processing module 123 transforms the advertisement site location ID feature to obtain an association feature related to the advertisement site location ID feature.
Furthermore, the mobile device model feature processing module 122 counts the original model parameters of the mobile device returned by the trusted media, selects a standard model corresponding to the standard model parameters from the original model parameters according to a preset rule, and associates brand information of the standard model according to the standard model.
Wherein the associated features include: goods, brands, advertisers, and industry categories.
In summary, the feature engineering is the most basic and important first part in the machine learning, and high-quality data can bring great promotion to subsequent models. The data and the characteristics determine the upper limit of machine learning, the model and the algorithm only approximate the upper limit, different characteristic processing modes are selected for different data and applied to the mobile terminal advertisement monitoring data, and the characteristic derivation and transformation not only meet the requirement on the training speed of the model, but also improve the accuracy of the model classifier. In the process of using the model for prediction, because the feature space is relatively concentrated, the predicted feature distribution is closer to the trained feature distribution, and the predicted result is more reliable.
Although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A feature processing method for advertisement monitoring data is characterized by comprising the following steps:
the extraction step comprises: extracting basic original features from advertisement monitoring data;
the processing steps are as follows: and performing derivation and transformation reduction on the basic original features to map fine-grained feature dimensions in the basic original features to coarse-grained feature dimensions.
2. The feature processing method of claim 1, wherein the base raw feature comprises: advertisement site ID characteristics, log type characteristics, timestamp characteristics, mobile device model characteristics, and region ID characteristics.
3. The feature processing method of claim 2, wherein the processing step comprises:
and a timestamp characteristic processing step: reducing the feature dimension of the timestamp feature through mapping to obtain a final timestamp feature;
mobile equipment model feature processing: associating brand information of the mobile equipment model according to the original model parameters of the mobile equipment returned by the trusted media;
advertisement site location ID feature processing step: and transforming the advertisement site location ID characteristics to obtain associated characteristics related to the advertisement site location ID characteristics.
4. The feature processing method of claim 3, wherein the mobile device model feature processing step further comprises:
a statistical step: counting the original model parameters of the mobile equipment returned by the trusted media;
selecting: selecting a standard model corresponding to the standard model parameter from the original model parameters according to a preset rule;
selecting: and associating brand information of the standard model according to the standard model.
5. The feature processing method of claim 3, wherein the associating features comprises: goods, brands, advertisers, and industry categories.
6. A feature processing apparatus for advertisement monitoring data, comprising:
the extracting unit is used for extracting basic original characteristics from the advertisement monitoring data;
and the processing unit is used for deriving and transforming the basic original features to reduce the mapping of fine-grained feature dimensions in the basic original features to coarse-grained feature dimensions.
7. The feature processing apparatus of claim 6, wherein the base raw feature comprises: advertisement site ID characteristics, log type characteristics, timestamp characteristics, mobile device model characteristics, and region ID characteristics.
8. The feature processing apparatus of claim 7, wherein the processing unit comprises:
the time stamp feature processing module is used for reducing the feature dimension of the time stamp feature through mapping to obtain a final time stamp feature;
the mobile equipment model feature processing module is used for correlating brand information of the mobile equipment model according to the original model parameters of the mobile equipment returned by the trusted media;
and the advertisement site location ID feature processing module is used for transforming the advertisement site location ID feature to obtain the associated feature related to the advertisement site location ID feature.
9. The feature processing apparatus according to claim 8, wherein the mobile device model feature processing module performs statistics on the original model parameters of the mobile device returned by the trusted media, selects a standard model corresponding to the standard model parameters from the original model parameters according to a preset rule, and associates brand information of the standard model according to the standard model.
10. The feature processing apparatus of claim 8, wherein the associated feature comprises: goods, brands, advertisers, and industry categories.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010655582.2A CN111951038A (en) | 2020-07-09 | 2020-07-09 | Feature processing method and device for advertisement monitoring data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010655582.2A CN111951038A (en) | 2020-07-09 | 2020-07-09 | Feature processing method and device for advertisement monitoring data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111951038A true CN111951038A (en) | 2020-11-17 |
Family
ID=73339993
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010655582.2A Pending CN111951038A (en) | 2020-07-09 | 2020-07-09 | Feature processing method and device for advertisement monitoring data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111951038A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105260365A (en) * | 2014-06-04 | 2016-01-20 | 中国移动通信集团宁夏有限公司 | Terminal information processing method and device |
| CN111191860A (en) * | 2020-04-14 | 2020-05-22 | 北京热云科技有限公司 | Prediction method based on ensemble learning, prediction system and readable storage medium |
-
2020
- 2020-07-09 CN CN202010655582.2A patent/CN111951038A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105260365A (en) * | 2014-06-04 | 2016-01-20 | 中国移动通信集团宁夏有限公司 | Terminal information processing method and device |
| CN111191860A (en) * | 2020-04-14 | 2020-05-22 | 北京热云科技有限公司 | Prediction method based on ensemble learning, prediction system and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Florysiak et al. | The information content of ico white papers | |
| US20190019213A1 (en) | Predicting the effectiveness of a marketing campaign prior to deployment | |
| US20190236497A1 (en) | System and method for automated model selection for key performance indicator forecasting | |
| CN103514178A (en) | Searching and sorting method and device based on click rate | |
| CN103154991A (en) | Credit risk mining | |
| CN111178624A (en) | Method for predicting new product demand | |
| WO2021102133A1 (en) | Methods, systems, apparatus and articles of manufacture for receipt decoding | |
| CN116452261B (en) | Advertisement delivery data processing method based on cross-border E-commerce service platform | |
| EP2939196A1 (en) | Techniques for measuring video profit | |
| US20220277331A1 (en) | Systems and methods for procurement cost forecasting | |
| CN106251215A (en) | A kind of financial budget analyzes system | |
| JP2020149468A (en) | Product management system and product management method | |
| CN105359172A (en) | Calculating a probability of a business being delinquent | |
| CN116611888A (en) | Dynamic bidding method, storage medium and electronic device for bidding advertisement | |
| CN105303447A (en) | Method and device for carrying out credit rating through network information | |
| CN108009850A (en) | One kind marketing advertisement scheme matching process and system | |
| CN116304811B (en) | Dynamic sample weight adjustment method and system based on focus loss function | |
| CN111951038A (en) | Feature processing method and device for advertisement monitoring data | |
| CN115563176A (en) | Electronic commerce data processing system and method | |
| CN115018529A (en) | Financial advertisement generation method, device, equipment and storage medium | |
| JP2005100221A (en) | Investment judgement support information providing device and method | |
| WO2015002879A1 (en) | Generating a simulted invoice | |
| Ballingall et al. | DEPARTMENT OF ECONOMICS AND FINANCE SCHOOL OF BUSINESS AND ECONOMICS UNIVERSITY OF CANTERBURY CHRISTCHURCH, NEW ZEALAND | |
| Sinclair et al. | Differences in Early GDP Component Estimates Between Recession and Expansions | |
| Cokis et al. | Reading through the Lines: Price-setting Indicators from Earnings Calls |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |