CN111126614B - Attribution method, attribution device and storage medium - Google Patents
Attribution method, attribution device and storage medium Download PDFInfo
- Publication number
- CN111126614B CN111126614B CN201811293288.0A CN201811293288A CN111126614B CN 111126614 B CN111126614 B CN 111126614B CN 201811293288 A CN201811293288 A CN 201811293288A CN 111126614 B CN111126614 B CN 111126614B
- Authority
- CN
- China
- Prior art keywords
- channel
- channels
- attribution
- target
- training set
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000010801 machine learning Methods 0.000 claims abstract description 51
- 238000006243 chemical reaction Methods 0.000 claims description 87
- 238000013527 convolutional neural network Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 6
- 230000007704 transition Effects 0.000 description 6
- 230000006399 behavior Effects 0.000 description 4
- 230000001960 triggered effect Effects 0.000 description 4
- 230000009466 transformation Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The invention provides an attribution method, an attribution device and a storage medium. The method comprises the following steps: determining the identification of a target channel; inputting the identification of the target channel into a machine learning model to obtain a characteristic weight of the target channel, wherein the characteristic weight is used for representing attribution results of the target channel; the machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths. The invention improves the accuracy of attribution results.
Description
Technical Field
The present invention relates to the field of internet technologies, and in particular, to an attribution method, an attribution device, and a storage medium.
Background
In the field of information delivery, information can be delivered through a variety of channels.
In the prior art, when information is put in through multiple channels, attribution results of different putting channels, namely contribution degrees of the different putting channels to information conversion, need to be determined. Here, the conversion may be, for example, downloading, consultation, purchase, or the like. Currently, in determining attribution results for each of a plurality of channels, only user browsing paths (i.e., conversion paths) where conversion has occurred are considered, and user browsing paths (i.e., unconverted paths) where no conversion has occurred are not considered.
Therefore, in the prior art, there is a problem that the attribution result is inaccurate.
Disclosure of Invention
The embodiment of the invention provides an attribution method, an attribution device and a storage medium, which are used for solving the problem of inaccurate attribution results in the prior art.
In a first aspect, the present invention provides an attribution method, comprising:
determining the identification of a target channel;
inputting the identification of the target channel into a machine learning model to obtain a characteristic weight of the target channel, wherein the characteristic weight is used for representing attribution results of the target channel;
the machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths.
In one possible implementation, the method further comprises: determining negative examples in the training set according to channels in the unconverted path within the target duration range, and determining positive examples in the training set according to channels in the converted path within the target duration range;
and training the machine learning model according to the training set.
In one possible implementation, the determining a negative example in the training set according to the channels in the unconverted path within the target duration range includes:
and randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set.
In one possible implementation, the specific number is a preset number, or the specific number is determined according to the total number of channels in the positive example.
In one possible implementation, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes:
randomly extracting a specific number of channels in the last channel of all unconverted paths in the target duration range as negative examples of the training set.
The determining the positive example in the training set according to the channel in the conversion path in the target duration range comprises the following steps:
and taking the last channel of all conversion paths in the target duration range as the positive example of the training set.
In one possible implementation, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes:
and randomly extracting a specific number of all channels of all unconverted paths in the target duration range as negative examples of the training set.
The determining the positive example in the training set according to the channel in the transition path in the target duration range comprises the following steps:
and taking all channels of all conversion paths in the target duration range as positive examples of the training set.
In one possible implementation, the method further comprises:
and carrying out range conversion on the characteristic weights of the target channel according to the relation between the numerical range of the characteristic weights and the numerical range of the attribution results to obtain the attribution results of the target channel.
In a second aspect, the present invention provides an attribution apparatus comprising:
the determining module is used for determining the identification of the target channel;
the obtaining module is used for inputting the identification of the target channel determined by the determining module into a machine learning model to obtain the characteristic weight of the target channel, wherein the characteristic weight is used for representing the attribution result of the target channel;
the machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths.
In one possible implementation, the apparatus further includes a training module to:
determining negative examples in the training set according to channels in the unconverted path within the target duration range, and determining positive examples in the training set according to channels in the converted path within the target duration range;
and training the machine learning model according to the training set.
In one possible implementation, the training module is configured to determine a negative example in a training set according to channels in an unconverted path within the target duration range, and specifically includes:
and randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set.
In one possible implementation, the specific number is a preset number, or the specific number is determined according to the total number of channels in the positive example.
In one possible implementation, the training module is configured to randomly extract, as a negative example in the training set, a specific number of channels in unconverted paths within the target duration range, where the training module specifically includes:
randomly extracting a specific number of channels in the last channel of all unconverted paths in the target duration range as negative examples of the training set.
The training module is used for determining the positive examples in the training set according to the channels in the conversion path in the target duration range, and specifically comprises the following steps:
and taking the last channel of all conversion paths in the target duration range as the positive example of the training set.
In one possible implementation, the training module is configured to randomly extract, as a negative example in the training set, a specific number of channels in unconverted paths within the target duration range, where the training module specifically includes:
and randomly extracting a specific number of all channels of all unconverted paths in the target duration range as negative examples of the training set.
The training module is used for determining the positive examples in the training set according to the channels in the conversion path in the target duration range, and specifically comprises the following steps:
and taking all channels of all conversion paths in the target duration range as positive examples of the training set.
In one possible implementation, the apparatus further includes: and the conversion module is used for carrying out range conversion on the characteristic weights of the target channel according to the relation between the numerical range of the characteristic weights and the numerical range of the attribution results to obtain the attribution results of the target channel.
In a third aspect, the present invention provides an attribution apparatus, comprising:
a processor and a memory for storing computer instructions; the processor executing the computer instructions to perform the method of any of the first aspects above.
In a fourth aspect, the present invention provides a computer readable storage medium, which when executed by a processor of an attribution apparatus, enables the attribution apparatus to perform the method of any of the first aspects above.
According to the attribution method, the attribution device and the storage medium, the characteristic weight of the target channel is obtained by determining the identification of the target channel and inputting the identification of the target channel into the machine learning model, and the characteristic weight is used for representing the attribution result of the target channel, wherein the machine learning model is a model obtained by training according to the unconverted path and the channels in the conversion path within the target duration range, so that the conversion path and the channels in the unconverted path can be considered when the attribution result of the channel is determined, the situation that only the channels in the conversion path are considered is avoided, the determined attribution result only represents the number of users obtaining information through the channels and carrying out information conversion, but cannot represent the occupation ratio of the number of users obtaining information through the channels and carrying out information conversion, and the problem of inaccurate attribution result is caused, and the accuracy of attribution result is improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments or the description of the prior art will be briefly described below, it will be obvious that the drawings in the following description are some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort to a person skilled in the art.
Fig. 1 is a schematic diagram of an application scenario of an attribution method provided by an embodiment of the present invention;
fig. 2 is a schematic flow chart of a first embodiment of a attribution method provided in an embodiment of the present invention;
fig. 3 is a schematic flow chart of a second embodiment of a attribution method provided by an embodiment of the present invention;
fig. 4 is a schematic flow chart of a third embodiment of a attribution method provided by an embodiment of the present invention;
fig. 5 is a schematic structural diagram of a attribution device according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of a attribution device according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 is an application scenario schematic diagram of an attribution method provided by an embodiment of the present invention. As shown in fig. 1, the application scenario may include a terminal, a server, and an attribution apparatus. The information delivery party can deliver information of the same popularization subject through different channels provided by the terminal, wherein the channels specifically refer to popularization channels of the information, such as WeChat, microblog, mailbox and the like, which can be considered as popularization channels. It should be noted that the information of different promotion channels corresponding to the same promotion subject may be different.
The user can acquire information of one promotion topic through a plurality of channels, for example, the user can acquire the information of the promotion topic provided by the channel 1 in a period of time, then acquire the information of the promotion topic provided by the channel 2, and then acquire the information of the promotion topic provided by the channel 3. At this time, channel 1- > channel 2- > channel 3 may be considered as a user browsing path in the period of time, and further, when information conversion is triggered in channel 3, channel 1- > channel 2- > channel 3 may be considered as a conversion path in the period of time; when no information conversion is triggered in channel 3, channel 1- > channel 2- > channel 3 can be considered an unconverted path during that period.
Here, the information conversion refers to a conversion behavior triggered by information, which can bring corresponding value to the information delivery party. For example, for a seller of an e-commerce web site, the transformation behavior may be referred to as a deal; for advertisers who are devoted to promoting applications, the transformation behavior may be referred to as downloading; for some consultation type companies, the transformation behavior may be referred to as consultation.
Wherein the server may collect events of users viewing information provided by different channels. Further, the unconverted path and the channels in the converted path in the period of time can be determined based on the events of the information provided by the different channels, which are collected by the server, of the user in the period of time. The unconverted path and the channel in the converted path in the period of time can be specifically the unconverted path and the channel in the converted path of the information of one or more promotion topics put in by one information putting party in the period of time.
The attribution device can train to obtain a machine learning model which can be used for obtaining the characteristic weight of the channel according to the identification of the channel based on the unconverted path in the period of time and the channel in the converted path, and the characteristic weight of the channel can represent attribution results of the channel. Further, the attribution device may determine an identification of a channel (i.e., a target channel) for which an attribution result is to be determined, and input the identification of the target channel to the machine learning model, thereby obtaining a feature weight of the target channel.
In fig. 1, the machine learning model is trained by the attribution apparatus, or the machine learning model may not be trained by the attribution apparatus.
It should be noted that, the present invention is not limited to this, and the user may collect events of information provided by different channels by the same server or by different servers.
Fig. 2 is a schematic flow chart of a first embodiment of a attribution method according to an embodiment of the present invention. As shown in fig. 2, the method of the present embodiment may include:
step 201, determining the identification of the target channel.
In this step, the target channel may specifically refer to a channel for which an attribution result is to be determined. The identification of the target channel may be, for example, the name of the target channel. Alternatively, the identification of the target channel may be determined by user input, for example, the user may input the identification of the target channel, or the user may select the target channel; alternatively, the identification of the target channel may be obtained from other devices, e.g., the identification of the target channel sent by other devices may be received.
Step 202, inputting the identification of the target channel into a machine learning model to obtain the characteristic weight of the target channel, wherein the characteristic weight is used for representing the attribution result of the target channel.
In this step, the machine learning model is a model obtained by training according to the unconverted path and the channels in the converted path within the target duration range. Alternatively, the target duration range may specifically identify a last duration range, such as the last 10 days, the last month, etc. The machine learning model is a model obtained by training according to channels in the unconverted path and channels in the converted path within the target duration range, so that the model can learn the capability of determining the characteristic weight of the channel according to the identification of the channel. Alternatively, the machine learning model may specifically be a (Neural Network, NN) Neural Network model, such as a convolutional Neural Network (Convolutional Neural Network, CNN) model.
Here, since the model is obtained by training the machine learning model according to the channel in the unconverted path and the channel in the converted path in the target range, the attribution result of the channel represented by the feature weight of the channel obtained by training the machine learning model can be considered.
Since the information delivery party typically pays for the act of the user obtaining information delivered by the information delivery party through a channel, not only the channel in the conversion path but also the channel in the non-conversion path need to be considered in determining the attribution result of a channel based on consideration of return on investment.
If only the channel in the conversion path is considered in determining the attribution result, and not the channel in the non-conversion path, only a partial investment of the information delivery party for the channel information delivery is considered, which is an investment of the conversion path for the channel, and other partial investments of the information delivery party for the channel are not considered, which are investments of the non-conversion path for the channel, so the attribution result is inaccurate. It can be seen that if only the channel in the conversion path is considered, the determined attribution result only represents the number of users who obtain information through the channel and perform information conversion, but cannot represent the ratio of the number of users who obtain information through the channel and perform information conversion to the number of users who obtain information through the channel. For example, assuming that the number of times channel 1 is included in the conversion path is 4, the number of times channel 2 is 5, the number of times channel 1 is included in the non-conversion path is 2, the number of times channel 2 is also 5, the degree of contribution of channel 2 to information conversion is certainly greater than channel 1 if only the conversion path is considered, and the degree of contribution of channel 2 to information conversion is not necessarily greater than channel 1 if both the conversion path and the non-conversion path are considered. For example, assuming that the larger the attribution result of the channel means the greater the degree of contribution, and the attribution result of the channel is equal to the number of times in the conversion path minus the number of times in the non-conversion path by 0.5 times, the attribution result of the channel 1 is equal to 3, and the attribution result of the channel 2 is equal to 2.5.
According to the attribution method provided by the embodiment, the characteristic weight of the target channel is obtained by determining the identification of the target channel and inputting the identification of the target channel into the machine learning model, and the characteristic weight is used for representing the attribution result of the target channel, wherein the machine learning model is a model obtained by training according to the unconverted path and the channels in the conversion path within the target duration range, so that the conversion path and the channels in the unconverted path can be considered when the attribution result of the channel is determined, the situation that only the channels in the conversion path are considered is avoided, the determined attribution result only represents the number of users obtaining information through the channels and carrying out information conversion, but cannot represent the ratio of the number of users obtaining information through the channels and carrying out information conversion to the number of users obtaining information through the channels is avoided, the problem that the attribution result is inaccurate is caused, and the attribution result accuracy is improved.
Fig. 3 is a schematic flow chart of a second embodiment of a attribution method according to an embodiment of the present invention. This embodiment generally describes an alternative implementation of training a machine learning model based on the embodiment shown in fig. 2. As shown in fig. 3, the method of the present embodiment may include:
step 301, determining negative examples in the training set according to channels in the unconverted path within the target duration range.
In this step, negative examples may also be referred to as negative samples (negative samples). Alternatively, channels in the unconverted paths within the target duration range may be used as negative examples in the training set.
Considering that the information conversion can bring corresponding value to the information input by the information investor, in order to avoid the problem that the number of negative examples is too large, the importance of the duty ratio in determining the attribution result is too large, and the value brought by the information conversion to the investor cannot be highlighted, the number of the negative examples can be limited. Thus, optionally, step 301 may specifically include: and randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set. Optionally, the specific number is a preset number; alternatively, the specific number may be determined according to the total number of channels in the positive example in the training set, and may be, for example, 10% of the total number of channels.
Step 302, determining the positive examples in the training set according to the channels in the transition path in the target duration range.
In this step, negative examples may also be referred to as negative samples (negative samples). Optionally, the channel in the transition path in the target duration range may be used as a positive example in the training set.
It is contemplated that although the number of channels in the conversion path may be multiple, the channel triggering the conversion of information is typically the last channel in the conversion path. Optionally, step 302 may specifically include: and taking the last channel of all conversion paths in the target duration range as the positive example of the training set. Further, in order to ensure consistency between negative examples and positive examples, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes: randomly extracting a specific number of channels in the last channel of all unconverted paths in the target duration range as negative examples of the training set.
It is contemplated that although the number of channels in the conversion path may be multiple, the channel from which the user first obtains information is typically the first channel in the conversion path. Optionally, step 302 may specifically include: and taking the first channel of all conversion paths in the target duration range as a positive example of the training set. Further, in order to ensure consistency between negative examples and positive examples, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes: and randomly extracting a specific number of channels in the first channels of all unconverted paths in the target duration range as negative examples of the training set.
Considering that the number of channels in the conversion path can be multiple, the user continuously obtains information through the multiple channels in the conversion path, and finally information conversion is triggered. Optionally, step 302 may specifically include: and taking all channels of all conversion paths in the target duration range as positive examples of the training set. Further, in order to ensure consistency between negative examples and positive examples, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes: and randomly extracting a specific number of all channels of all unconverted paths in the target duration range as negative examples of the training set.
Optionally, the feature weight of the target channel is the attribution result of the target channel.
It should be noted that there is no restriction on the sequence between the step 302 and the step 301.
And step 303, training the machine learning model according to the training set.
In this step, since the positive examples are determined according to the channels in the conversion path within the target duration range, after the machine learning model is trained according to the positive examples, the trained machine learning model can consider the positive influence of the number of users who obtain information and perform information conversion through one channel on the attribution result of the channel, that is, the greater the number of users who obtain information and perform information conversion through one channel, the greater the contribution of the channel to information conversion.
Because the negative examples are determined according to the channels in the non-conversion path within the target duration range, after the machine learning model is trained according to the negative examples, the trained machine learning model can consider the negative influence on the attribution result of the channel due to the proportion of the number of users who acquire information through one channel and perform information conversion to the number of users who acquire information through the channel, namely, the larger the proportion of the number of users who acquire information through one channel and perform information conversion to the number of users who acquire information through the channel, the smaller the contribution of the channel to information conversion.
According to the attribution method provided by the embodiment, the negative examples in the training set are determined according to the channels in the unconverted path within the target duration range, the positive examples in the training set are determined according to the channels in the converting path within the target duration range, and the machine learning model is trained according to the training set, so that the characteristic weights of the target channels determined based on the trained machine learning model can be considered, the channels in the converting path and the unconverted path are avoided, only the channels in the converting path are considered, the determined attribution result only represents the number of users obtaining information through the channels and generating information conversion, the proportion of the number of users obtaining information through the channels and generating information conversion to the number of users obtaining information through the channels cannot be represented, the problem of inaccurate attribution results is caused, and the attribution result accuracy is improved.
Fig. 4 is a schematic flow chart of a third embodiment of a attribution method provided by an embodiment of the present invention. This embodiment mainly describes an alternative implementation of deriving attribution results from feature weights based on the embodiment shown in fig. 2. As shown in fig. 4, the method of the present embodiment may include:
step 401, determining an identification of a target channel.
It should be noted that, step 401 is similar to step 201, and will not be described herein.
Step 402, inputting the identification of the target channel into a machine learning model to obtain a feature weight of the target channel, wherein the feature weight is used for representing attribution results of the target channel.
The machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths.
It should be noted that, step 402 is similar to step 202, and will not be described herein.
Step 403, performing range conversion on the feature weight of the target channel according to the relationship between the numerical range of the feature weight and the numerical range of the attribution result, so as to obtain the attribution result of the target channel.
In the step, when the characteristic weight of the target channel cannot be directly used as the attribution result of the target channel, the range conversion can be carried out on the characteristic weight of the target channel to obtain the attribution result of the target channel. For example, assuming that the numerical range of the feature weight is 0 to 1 and the numerical range of the attribution result is 0 to 100, the result obtained by multiplying the feature weight of the target channel by 100 may be used as the attribution result of the target channel.
According to the attribution method provided by the embodiment, the attribution result of the target channel is obtained by converting the range of the characteristic weight of the target channel according to the relation between the range of the characteristic weight and the range of the attribution result, so that when the characteristic weight of the target channel cannot be directly used as the attribution result of the target channel, the attribution result of the target channel is determined according to the characteristic weight of the target channel.
Fig. 5 is a schematic structural diagram of a attribution device provided by an embodiment of the present invention, and the device provided by the embodiment may be applied to the above method embodiment. As shown in fig. 5, the apparatus of this embodiment may include: a determining module 51 and a deriving module 52. Wherein,
a determining module 51, configured to determine an identifier of the target channel;
an obtaining module 52, configured to input the identification of the target channel determined by the determining module 51 to a machine learning model, and obtain a feature weight of the target channel, where the feature weight is used to represent an attribution result of the target channel;
the machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths.
In a possible implementation, the apparatus further comprises a training module 53 for:
determining negative examples in the training set according to channels in the unconverted path within the target duration range, and determining positive examples in the training set according to channels in the converted path within the target duration range;
and training the machine learning model according to the training set.
In one possible implementation, the training module 53 is configured to determine a negative example in the training set according to channels in the unconverted path within the target duration range, and specifically includes:
and randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set.
In one possible implementation, the specific number is a preset number, or the specific number is determined according to the total number of channels in the positive example.
In one possible implementation, the training module 53 is configured to randomly extract, as negative examples in the training set, a specific number of channels in the unconverted paths within the target duration range, where the specific number of channels includes:
randomly extracting a specific number of channels in the last channel of all unconverted paths in the target duration range as negative examples of the training set.
The training module 53 is configured to determine a positive example in the training set according to the channel in the conversion path in the target duration range, and specifically includes:
and taking the last channel of all conversion paths in the target duration range as the positive example of the training set.
In one possible implementation, the training module 53 is configured to randomly extract, as negative examples in the training set, a specific number of channels in the unconverted paths within the target duration range, where the specific number of channels includes:
and randomly extracting a specific number of all channels of all unconverted paths in the target duration range as negative examples of the training set.
The training module 53 is configured to determine a positive example in the training set according to a channel in the transition path within the target duration range, and specifically includes:
and taking all channels of all conversion paths in the target duration range as positive examples of the training set.
In one possible implementation, the apparatus further includes: and the conversion module 54 is configured to perform range conversion on the feature weight of the target channel according to a relationship between a numerical range of the feature weight and a numerical range of the attribution result, so as to obtain the attribution result of the target channel.
The device of the present embodiment may be used to implement the technical solution of the embodiment shown in the foregoing method, and its implementation principle and technical effects are similar, and are not repeated here.
Fig. 6 is a schematic structural diagram of a attribution device according to an embodiment of the present invention, as shown in fig. 6, the attribution device may include: a processor 61 and a memory 62 for storing computer instructions.
Wherein processor 61 executes the computer instructions to perform the method of:
determining the identification of a target channel;
inputting the identification of the target channel into a machine learning model to obtain a characteristic weight of the target channel, wherein the characteristic weight is used for representing attribution results of the target channel;
the machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths.
In one possible implementation, the method further comprises: determining negative examples in the training set according to channels in the unconverted path within the target duration range, and determining positive examples in the training set according to channels in the converted path within the target duration range;
and training the machine learning model according to the training set.
In one possible implementation, the determining a negative example in the training set according to the channels in the unconverted path within the target duration range includes:
and randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set.
In one possible implementation, the specific number is a preset number, or the specific number is determined according to the total number of channels in the positive example.
In one possible implementation, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes:
randomly extracting a specific number of channels in the last channel of all unconverted paths in the target duration range as negative examples of the training set.
The determining the positive example in the training set according to the channel in the conversion path in the target duration range comprises the following steps:
and taking the last channel of all conversion paths in the target duration range as the positive example of the training set.
In one possible implementation, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes:
and randomly extracting a specific number of all channels of all unconverted paths in the target duration range as negative examples of the training set.
The determining the positive example in the training set according to the channel in the transition path in the target duration range comprises the following steps:
and taking all channels of all conversion paths in the target duration range as positive examples of the training set.
In one possible implementation, the method further comprises:
and carrying out range conversion on the characteristic weights of the target channel according to the relation between the numerical range of the characteristic weights and the numerical range of the attribution results to obtain the attribution results of the target channel.
Embodiments of the present invention also provide a computer-readable storage medium, which when executed by a processor of an attribution apparatus, enables the attribution apparatus to perform an attribution method, the method comprising:
determining the identification of a target channel;
inputting the identification of the target channel into a machine learning model to obtain a characteristic weight of the target channel, wherein the characteristic weight is used for representing attribution results of the target channel;
the machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths.
In one possible implementation, the method further comprises: determining negative examples in the training set according to channels in the unconverted path within the target duration range, and determining positive examples in the training set according to channels in the converted path within the target duration range;
and training the machine learning model according to the training set.
In one possible implementation, the determining a negative example in the training set according to the channels in the unconverted path within the target duration range includes:
and randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set.
In one possible implementation, the specific number is a preset number, or the specific number is determined according to the total number of channels in the positive example.
In one possible implementation, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes:
randomly extracting a specific number of channels in the last channel of all unconverted paths in the target duration range as negative examples of the training set.
The determining the positive example in the training set according to the channel in the conversion path in the target duration range comprises the following steps:
and taking the last channel of all conversion paths in the target duration range as the positive example of the training set.
In one possible implementation, the randomly extracting a specific number of channels in the unconverted paths within the target duration range as negative examples in the training set includes:
and randomly extracting a specific number of all channels of all unconverted paths in the target duration range as negative examples of the training set.
The determining the positive example in the training set according to the channel in the transition path in the target duration range comprises the following steps:
and taking all channels of all conversion paths in the target duration range as positive examples of the training set.
In one possible implementation, the method further comprises:
and carrying out range conversion on the characteristic weights of the target channel according to the relation between the numerical range of the characteristic weights and the numerical range of the attribution results to obtain the attribution results of the target channel.
Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the method embodiments described above may be performed by hardware associated with program instructions. The foregoing program may be stored in a computer readable storage medium. The program, when executed, performs steps including the method embodiments described above; and the aforementioned storage medium includes: various media that can store program code, such as ROM, RAM, magnetic or optical disks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the 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 scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (6)
1. A attribution method, comprising:
determining the identification of a target channel;
inputting the identification of the target channel into a machine learning model to obtain a characteristic weight of the target channel, wherein the characteristic weight is used for representing attribution results of the target channel;
the machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths; the machine learning model is a convolutional neural network model;
the method further comprises the steps of:
randomly extracting a specific number of channels in the last channel of all unconverted paths in the target duration range as negative examples of the training set, and taking the last channel of all converted paths in the target duration range as positive examples of the training set; training a machine learning model according to the training set;
or,
randomly extracting a specific number of all channels of all unconverted paths in the target duration range as negative examples of the training set, and taking all channels of all converted paths in the target duration range as positive examples of the training set; and training the machine learning model according to the training set.
2. The method of claim 1, wherein the specific number is a preset number or the specific number is determined according to the total number of channels in the positive example.
3. The method according to claim 1, wherein the method further comprises:
and carrying out range conversion on the characteristic weights of the target channel according to the relation between the numerical range of the characteristic weights and the numerical range of the attribution results to obtain the attribution results of the target channel.
4. An attribution apparatus, comprising:
the determining module is used for determining the identification of the target channel;
the obtaining module is used for inputting the identification of the target channel determined by the determining module into a machine learning model to obtain the characteristic weight of the target channel, wherein the characteristic weight is used for representing the attribution result of the target channel; the machine learning model is a convolutional neural network model;
the machine learning model is a model obtained by training according to unconverted paths in a target duration range and channels in converted paths;
the training module is used for randomly extracting a specific number of channels in the last channel of all unconverted paths in the target duration range as negative examples of the training set, and taking the last channel of all converted paths in the target duration range as positive examples of the training set; training a machine learning model according to the training set; or randomly extracting a specific number of all channels of all unconverted paths in the target duration range as negative examples of the training set, and taking all channels of all converted paths in the target duration range as positive examples of the training set; and training the machine learning model according to the training set.
5. An attribution apparatus, comprising:
a processor and a memory for storing computer instructions; the processor executing the computer instructions to perform the method of any of claims 1-3.
6. A computer readable storage medium, characterized in that instructions in the storage medium, when executed by a processor of attribution means, enable attribution means to perform the method of any of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811293288.0A CN111126614B (en) | 2018-11-01 | 2018-11-01 | Attribution method, attribution device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811293288.0A CN111126614B (en) | 2018-11-01 | 2018-11-01 | Attribution method, attribution device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111126614A CN111126614A (en) | 2020-05-08 |
| CN111126614B true CN111126614B (en) | 2024-01-16 |
Family
ID=70494171
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811293288.0A Active CN111126614B (en) | 2018-11-01 | 2018-11-01 | Attribution method, attribution device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111126614B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112200618B (en) * | 2020-10-29 | 2022-05-17 | 度小满科技(北京)有限公司 | Message channel attribution method, device and system |
Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103562946A (en) * | 2011-05-27 | 2014-02-05 | 谷歌公司 | Multiple attribution models with return on ad spend |
| CN103605745A (en) * | 2013-11-20 | 2014-02-26 | 北京国双科技有限公司 | Method, device and system for processing conversion paths |
| CN103733217A (en) * | 2011-04-11 | 2014-04-16 | 谷歌公司 | Illustrating cross channel conversion paths |
| CN104133837A (en) * | 2014-06-24 | 2014-11-05 | 上海交通大学 | Internet information putting channel optimizing system based on distributed computing |
| CN105069654A (en) * | 2015-08-07 | 2015-11-18 | 新一站保险代理有限公司 | User identification based website real-time/non-real-time marketing investment method and system |
| CN105160545A (en) * | 2015-08-07 | 2015-12-16 | 百度在线网络技术（北京）有限公司 | Delivered information pattern determination method and device |
| CN105408891A (en) * | 2013-06-03 | 2016-03-16 | 机械地带有限公司 | Systems and methods for multi-user multi-lingual communications |
| CN105447730A (en) * | 2015-12-25 | 2016-03-30 | 腾讯科技（深圳）有限公司 | Target user orientation method and device |
| CN105960656A (en) * | 2013-12-05 | 2016-09-21 | 谷歌公司 | Methods and systems for creating data-driven attribution model for assigning attribution credit to multiple events |
| CN106529721A (en) * | 2016-11-08 | 2017-03-22 | 安徽大学 | Advertisement click-through rate predication system based on deep characteristic extraction and predication method thereof |
| CN107346496A (en) * | 2016-05-05 | 2017-11-14 | 腾讯科技（北京）有限公司 | Targeted customer's orientation method and device |
| DE102016210886A1 (en) * | 2016-06-17 | 2017-12-21 | Robert Bosch Gmbh | Concept for controlling a traffic within a parking lot |
| CN107563781A (en) * | 2016-06-30 | 2018-01-09 | 阿里巴巴集团控股有限公司 | A kind of information launches effect attribution method and device |
| CN107741967A (en) * | 2017-10-09 | 2018-02-27 | 北京京东尚科信息技术有限公司 | Method, apparatus and electronic equipment for behavioral data processing |
| CN108171528A (en) * | 2016-12-07 | 2018-06-15 | 阿里巴巴集团控股有限公司 | A kind of attribution method and attribution system |
| CN108280670A (en) * | 2017-01-06 | 2018-07-13 | 腾讯科技（深圳）有限公司 | Seed crowd method of diffusion, device and information jettison system |
| CN108320170A (en) * | 2017-01-16 | 2018-07-24 | 北京京东尚科信息技术有限公司 | Improve the method, apparatus and electronic device of the conversion ratio of the article in shopping cart |
| CN108564398A (en) * | 2018-03-29 | 2018-09-21 | 北京酷云互动科技有限公司 | A kind of appraisal procedure, system and the storage medium of channel ad conversion rates |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| NZ502019A (en) * | 1999-12-22 | 2002-10-25 | Compudigm Int Ltd | Method for generating a Venn diagram with each area of circle being proportional to the ratio of the number of elements in the circle |
| US10387909B2 (en) * | 2016-01-25 | 2019-08-20 | Adobe Inc. | Simulation-based evaluation of a marketing channel attribution model |
| US11222268B2 (en) * | 2017-03-09 | 2022-01-11 | Adobe Inc. | Determining algorithmic multi-channel media attribution based on discrete-time survival modeling |
-
2018
- 2018-11-01 CN CN201811293288.0A patent/CN111126614B/en active Active
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103733217A (en) * | 2011-04-11 | 2014-04-16 | 谷歌公司 | Illustrating cross channel conversion paths |
| CN103562946A (en) * | 2011-05-27 | 2014-02-05 | 谷歌公司 | Multiple attribution models with return on ad spend |
| CN105408891A (en) * | 2013-06-03 | 2016-03-16 | 机械地带有限公司 | Systems and methods for multi-user multi-lingual communications |
| CN103605745A (en) * | 2013-11-20 | 2014-02-26 | 北京国双科技有限公司 | Method, device and system for processing conversion paths |
| CN105960656A (en) * | 2013-12-05 | 2016-09-21 | 谷歌公司 | Methods and systems for creating data-driven attribution model for assigning attribution credit to multiple events |
| CN104133837A (en) * | 2014-06-24 | 2014-11-05 | 上海交通大学 | Internet information putting channel optimizing system based on distributed computing |
| CN105069654A (en) * | 2015-08-07 | 2015-11-18 | 新一站保险代理有限公司 | User identification based website real-time/non-real-time marketing investment method and system |
| CN105160545A (en) * | 2015-08-07 | 2015-12-16 | 百度在线网络技术（北京）有限公司 | Delivered information pattern determination method and device |
| CN105447730A (en) * | 2015-12-25 | 2016-03-30 | 腾讯科技（深圳）有限公司 | Target user orientation method and device |
| CN107346496A (en) * | 2016-05-05 | 2017-11-14 | 腾讯科技（北京）有限公司 | Targeted customer's orientation method and device |
| DE102016210886A1 (en) * | 2016-06-17 | 2017-12-21 | Robert Bosch Gmbh | Concept for controlling a traffic within a parking lot |
| CN107563781A (en) * | 2016-06-30 | 2018-01-09 | 阿里巴巴集团控股有限公司 | A kind of information launches effect attribution method and device |
| CN106529721A (en) * | 2016-11-08 | 2017-03-22 | 安徽大学 | Advertisement click-through rate predication system based on deep characteristic extraction and predication method thereof |
| CN108171528A (en) * | 2016-12-07 | 2018-06-15 | 阿里巴巴集团控股有限公司 | A kind of attribution method and attribution system |
| CN108280670A (en) * | 2017-01-06 | 2018-07-13 | 腾讯科技（深圳）有限公司 | Seed crowd method of diffusion, device and information jettison system |
| CN108320170A (en) * | 2017-01-16 | 2018-07-24 | 北京京东尚科信息技术有限公司 | Improve the method, apparatus and electronic device of the conversion ratio of the article in shopping cart |
| CN107741967A (en) * | 2017-10-09 | 2018-02-27 | 北京京东尚科信息技术有限公司 | Method, apparatus and electronic equipment for behavioral data processing |
| CN108564398A (en) * | 2018-03-29 | 2018-09-21 | 北京酷云互动科技有限公司 | A kind of appraisal procedure, system and the storage medium of channel ad conversion rates |
Non-Patent Citations (1)
| Title |
|---|
| 用户电子服务渠道偏好研究与应用;蔡丽华;黄婷;蔡丽平;;中国电力企业管理(26);61-63 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111126614A (en) | 2020-05-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2019072107A1 (en) | Prediction of spending power | |
| CN108109011A (en) | A kind of anti-cheat method of advertisement and computing device | |
| CN105373887B (en) | Quality evaluation method and system for terminal application | |
| CN105654200A (en) | Deep learning-based advertisement click-through rate prediction method and device | |
| CN103647800A (en) | Method and system of recommending application resources | |
| JP2018503158A (en) | Data mining method and apparatus based on social platform | |
| CN110896488B (en) | Recommendation method for live broadcast room and related equipment | |
| US11226991B2 (en) | Interest tag determining method, computer device, and storage medium | |
| US11281723B2 (en) | Widget recommendation for an online event using co-occurrence matrix | |
| CN104376083B (en) | It is a kind of that method is recommended based on concern relation and the figure of multi-user's behavior | |
| CN109062914A (en) | User's recommended method and device, storage medium and server | |
| CN110472154A (en) | A kind of resource supplying method, apparatus, electronic equipment and readable storage medium storing program for executing | |
| CN105893484A (en) | Microblog Spammer recognition method based on text characteristics and behavior characteristics | |
| TWI674517B (en) | Information interaction method and device | |
| CN106921795A (en) | A kind of contact data management method and its system | |
| CN108920479B (en) | Cross-information-source account recommendation method for two micro terminals | |
| CN103700004A (en) | Method and device for pushing microblog advertising service information | |
| CN111126614B (en) | Attribution method, attribution device and storage medium | |
| CN106909567A (en) | Data processing method and device | |
| CN111294620A (en) | Video recommendation method and device | |
| CN106257507A (en) | The methods of risk assessment of user behavior and device | |
| CN111064996B (en) | Method, system and storage medium for identifying user watching video content preference | |
| CN103312584B (en) | A kind of method and apparatus to release news in Web Community | |
| CN107451247A (en) | user identification method and device | |
| CN104268189B (en) | Evaluate the method and device of application |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |