CN111461760A - Price interval estimation method and device, electronic equipment and storage medium - Google Patents
Price interval estimation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN111461760A CN111461760A CN202010131079.7A CN202010131079A CN111461760A CN 111461760 A CN111461760 A CN 111461760A CN 202010131079 A CN202010131079 A CN 202010131079A CN 111461760 A CN111461760 A CN 111461760A
- Authority
- CN
- China
- Prior art keywords
- price
- price interval
- distribution
- preset
- commodity
- 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
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
- G06Q30/0278—Product appraisal
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Accounting & Taxation (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Engineering & Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the application discloses a price interval estimation method, a price interval estimation device, electronic equipment and a storage medium, wherein the price interval estimation method comprises the following steps: splicing the historical price distribution of the user and the supply distribution of merchants within a preset range of an appointed position into an input matrix; inputting the input matrix into a price interval estimation model, and outputting an estimated price interval by the price interval estimation model based on the input matrix; the step of outputting the estimated price interval by the price interval estimation model based on the input matrix comprises the following steps: performing convolution processing on the input matrix to obtain a certain number of characteristic graphs; performing spatial pyramid pooling on the certain number of feature maps respectively to obtain a certain number of first feature vectors, wherein the scale of the first feature vectors is a preset scale; and carrying out full connection processing on the certain number of first feature vectors to obtain a starting value and an ending value of a price interval. The embodiment of the application improves the estimation accuracy of the price interval.
Description
Technical Field
The embodiment of the application relates to the technical field of internet, in particular to a price interval estimation method and device, electronic equipment and a storage medium.
Background
In traditional traffic conversion scenes such as searching, advertisement recommendation and the like, the problem of estimation of personalized price gear of a user is faced.
In the prior art, when a house source price interval is estimated, the price of a house source is estimated by combining information such as room attributes, seasons, festivals and the like, a hierarchical structure is applied to a prediction model, the probability of the house being booked is estimated through a classification model, and then the price interval of the house is predicted through a logistic regression model according to the estimated probability. In the estimation mode, the preset probability and the preset price interval of the house belong to two estimation targets, and the sample space and the characteristic semantic space of the two estimation targets are inconsistent, so that the problem of error cascade propagation is brought, and the accuracy of the estimated price interval is low.
Disclosure of Invention
The embodiment of the application provides a price interval estimation method and device, electronic equipment and a storage medium, which are beneficial to improving the accuracy of price interval estimation.
In order to solve the above problem, in a first aspect, an embodiment of the present application provides a method for estimating a price interval, including:
splicing the historical price distribution of the user and the supply distribution of merchants within a preset range of an appointed position into an input matrix;
inputting the input matrix into a price interval estimation model, and outputting an estimated price interval by the price interval estimation model based on the input matrix;
wherein the step of outputting the estimated price interval based on the input matrix by the price interval estimation model comprises:
performing convolution processing on the input matrix to obtain a certain number of characteristic graphs;
performing spatial pyramid pooling on the certain number of feature maps respectively to obtain a certain number of first feature vectors, wherein the scale of the first feature vectors is a preset scale;
and carrying out full connection processing on the certain number of first feature vectors to obtain a starting value and an ending value of a price interval.
In a second aspect, an embodiment of the present application provides an estimation apparatus for a price interval, including:
the input characteristic acquisition module is used for splicing the historical price distribution of the user and the supply distribution of merchants within a preset range of the designated position into an input matrix;
the price interval estimation module is used for inputting the input matrix into a price interval estimation model, and outputting an estimated price interval by the price interval estimation model based on the input matrix;
wherein, the price interval estimation module comprises:
the convolution processing unit is used for carrying out convolution processing on the input matrix to obtain a certain number of characteristic graphs;
the pooling processing unit is used for respectively performing space pyramid pooling on the certain number of feature maps to obtain a certain number of first feature vectors, and the scale of each first feature vector is a preset scale;
and the price interval estimation unit is used for carrying out full-connection processing on the certain number of first characteristic vectors to obtain a starting value and a terminating value of the price interval.
In a third aspect, an embodiment of the present application further provides an electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the method for estimating the price interval according to the embodiment of the present application when executing the computer program.
In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to perform the steps of the estimation method for price intervals disclosed in the embodiments of the present application.
The price interval estimation method, the device, the electronic equipment and the storage medium provided by the embodiment of the application are characterized in that the historical price distribution of a user and the supply distribution of merchants in a specified position range are spliced into an input matrix, the input matrix is input into a price interval estimation model, the estimated price interval is output by the price interval estimation model based on the input matrix, the price interval estimation model carries out convolution processing on the input matrix to obtain a certain number of characteristic graphs, space pyramid pooling processing is respectively carried out on the certain number of characteristic graphs to obtain a certain number of first characteristic vectors, full connection processing is carried out on the certain number of first characteristic vectors to obtain the initial value and the final value of the price interval, so that the price interval can be estimated directly according to the historical data of the user and the supply information of merchants in the specified position range without adopting a hierarchical structure of reservation probability estimation and price interval estimation, therefore, the problem of error cascade propagation in the hierarchical structure is solved, and the estimation accuracy of the price interval is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be 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 only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
Fig. 1 is a flowchart of a method for estimating a price interval according to an embodiment of the present application;
FIG. 2 is a schematic diagram of spatial pyramid pooling in the embodiment of the present application with different scales;
fig. 3 is a schematic structural diagram of an estimation apparatus for price range according to a second embodiment of the present application;
fig. 4 is a schematic structural diagram of an electronic device according to a third embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. 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 application.
Example one
As shown in fig. 1, the method for estimating a price interval provided in this embodiment includes: step 110 to step 140.
And step 110, splicing the historical price distribution of the user and the supply distribution of merchants within a preset range of the designated position into an input matrix.
The historical price distribution of the user is the order quantity distribution of the order price of the consumption order of the user in a preset time range in a preset price interval, and the merchant supply distribution is the commodity quantity distribution of the commodity price of the merchant in a preset position range in the preset price interval. The designated location may be a location designated by the user or a current location located, for example, the user may input an address, and the address is used as the designated location, or a map is displayed, and a location selected by the user on the map is used as the designated location, or the current location located by the user terminal is obtained and the current location is used as the designated location, and of course, other manners of obtaining the designated location are available, which is not limited herein. The order quantity distribution is the order quantity corresponding to each price gear in a preset price interval, and the commodity quantity distribution is the commodity quantity corresponding to each price gear in the preset price interval. For example, in the field of take-away, the merchant is a take-away merchant, the preset price interval may be (0,200), each price gear may be increased by 1 yuan relative to the previous price gear, that is, 1 yuan, 2 yuan, 3 yuan, 4 yuan may be counted, and the order quantity or the commodity quantity corresponding to each price gear may be counted up to 200 yuan, so as to obtain the user historical price distribution and the merchant supply distribution.
The user historical price distribution and the merchant supply distribution can be respectively expressed as a vector, the user historical price distribution and the merchant supply distribution are spliced in parallel to form an input matrix, and the input matrix is used as the input characteristic of the price interval estimation model. For example, if the preset price interval is (0,8), the user historical price distribution is [0,0,0,3,4,5,3,2,1], the merchant supply distribution is [0,0,6,9,10,8,7,5,6], the user historical price distribution and the merchant supply distribution are merged into the following matrix:
in an embodiment of the application, before the splicing the user historical price distribution and the merchant supply distribution within the preset range of the designated location into the input matrix, the method further includes: acquiring historical consumption data within a preset time range of a user, and counting the order quantity distribution of the historical consumption data within a preset price interval to be used as the historical price distribution of the user; determining merchants within a preset range of a specified position, and counting the commodity quantity distribution of the commodity price of the merchants within a preset price interval according to the commodities of the merchants and the corresponding commodity price to serve as the supply distribution of the merchants; the starting value of the preset price interval is smaller than or equal to the starting value of the price interval output by the price interval estimation model, and the ending value of the preset price interval is larger than or equal to the ending value of the price interval output by the price interval estimation model. The range of the preset price interval is larger than the range of the estimated price interval, namely, a smaller price interval is estimated according to the historical price distribution of the user and the supply distribution of the merchant in a larger preset price interval. For example, the preset price interval may be (0,200), and the estimated price interval may be (0,20), (0,30), or (10, 40). The preset time range is a time range closer to the current time, such as the last three months.
In an embodiment of the application, the counting order quantity distribution of the historical consumption data in a preset price interval as the user historical price distribution includes: determining orders in the historical consumption data and corresponding order prices, and counting order quantity distribution of the order prices in a preset price interval according to the orders and the corresponding order prices to serve as first price distribution; and/or determining the commodity class in the historical consumption data, determining the commodity class with the largest order quantity as the class to be counted, and counting the order quantity distribution of the order price in the preset price interval of the class to be counted as a second price distribution; and taking the first price distribution and/or the second price distribution as the user historical price distribution.
When determining the historical price distribution of the user, the order quantity distribution in the whole preset price interval can be used as the historical price distribution of the user, the order quantity distribution in the preset price interval under the commodity class with the largest order quantity can also be used as the historical price distribution of the user, and the order quantity distribution in the whole preset price interval and the order quantity distribution in the preset price interval under the commodity class with the largest order quantity can also be used as the historical price distribution of the user together, namely, the first price distribution and the second price distribution are used as the historical price distribution of the user together. When determining the to-be-counted item, in addition to the above-mentioned item with the largest order number as the to-be-counted item, the item with the largest click amount of the user may also be used as the to-be-counted item.
In an embodiment of the application, the counting, according to the commodity of the merchant and the corresponding commodity price, the distribution of the quantity of the commodity of the merchant in a preset price interval as the supply distribution of the merchant includes: selecting a first preset number of commodities from each merchant as a first commodity to be counted, and counting the commodity number distribution of the commodity price of the first commodity to be counted in a preset price interval as a first supply distribution; determining brand merchants in the merchants, selecting a first preset number of commodities from each brand merchant as a second commodity to be counted, counting commodity number distribution of commodity prices of the second commodity to be counted in a preset price interval, and taking the commodity number distribution as second supply distribution; determining commodity types in the historical consumption data, determining the commodity type with the largest order number as a to-be-counted commodity type, selecting a second preset number of commodities under the to-be-counted commodity type from the commodities of the merchant to serve as a third to-be-counted commodity, and counting the commodity number distribution of the commodity price of the third to-be-counted commodity in a preset price interval to serve as a third supply distribution; at least one of the first offer distribution, the second offer distribution, and the third offer distribution is designated as the merchant offer distribution.
In determining the merchant offer distribution, one offer distribution, namely a first offer distribution, a second offer distribution, and a third offer distribution, may be determined for each of the merchant, the brand merchant, and the category of merchandise, with at least one of the three offer distributions being the merchant offer distribution. The first preset number is the number of commodities selected in each merchant, the second preset number is the number of commodities selected in the category to be counted, the first preset number is smaller than the second preset number, the first preset number may be 100, for example, and the second preset number may be 500, for example. In the field of takeaway, the item may be a package.
For example, the preset price interval may be (1,200), each price gear is increased by 1 element from the previous price gear, the first price distribution, the second price distribution, the first supply distribution, the second supply distribution and the third supply distribution are respectively 200-dimensional vectors, the first price distribution may be used as a user history price distribution, the first supply distribution, the second supply distribution and the third supply distribution may be used as a merchant supply distribution, so that the input vector includes four 200-dimensional vectors, and the four vectors are spliced into a 4 × 200-dimensional input matrix.
And 120, inputting the input matrix into a price interval estimation model, and outputting an estimated price interval by the price interval estimation model based on the input matrix.
The price interval estimation model is used for estimating the price interval based on the input matrix, and can comprise an input layer, a convolution layer, a pooling layer, a full-connection layer and an output layer. The convolutional layer is a multi-scale convolutional layer, a certain number of convolutional kernels with different widths are adopted for convolution processing, and the pooling layer converts a feature graph output by the convolutional layer into a feature with a preset scale in a spatial pyramid pooling mode.
The step of outputting the estimated price interval based on the input matrix by the price interval estimation model may include steps 121 to 123.
And step 121, performing convolution processing on the input matrix to obtain a certain number of characteristic graphs.
The certain number is the number of the adopted convolution kernels, and can be preset according to requirements. When convolution processing is carried out, a certain number of convolution kernels are adopted for convolution processing to obtain a certain number of feature maps, and different features can be extracted by adopting a plurality of convolution kernels for processing.
In an embodiment of the present application, the convolving the input matrix to obtain a certain number of feature maps includes: and performing convolution processing on the input matrix by adopting a certain number of convolution kernels with different widths to obtain a certain number of characteristic graphs.
After the input matrix is input into the price interval estimation model, convolution layers in the price interval estimation model respectively carry out convolution processing on the input matrix by adopting a certain number of convolution kernels with different widths so as to extract a certain number of different characteristics, and each convolution kernel extracts a characteristic diagram, so that a certain number of characteristic diagrams can be extracted through a certain number of convolution kernels.
For example, in the field of takeaway, the preset number may be 4, the sizes of the convolution kernels used may be 4 × 5, 4 × 10,4 × 15 and 4 × 20, respectively, and the 4 convolution kernels are used to perform convolution processing on the input matrix of 4 × 200-dimension and 200-dimension, respectively, to obtain 4 groups of feature maps of different scales, so that features of different price ranges can be extracted.
And step 122, performing spatial pyramid pooling on the certain number of feature maps respectively to obtain a certain number of first feature vectors, wherein the scale of the first feature vectors is a preset scale.
And the pooling layer of the price interval estimation model adopts a spatial pyramid pooling mode to convert a certain number of different scale characteristic diagrams into first characteristic vectors with preset scales respectively to obtain a certain number of first characteristic vectors.
The space pyramid pooling operation divides a feature map of any size by using scales of different sizes. As shown in fig. 2, an input feature map is divided by using three scales with different sizes, that is, a first division manner is to divide a feature map into 16 blocks, a second division manner is to divide a feature map into 4 blocks, a third division manner is to set a feature map as 1 block, and finally, 16+4+ 1-21 blocks (bins) are obtained in total, and a feature is extracted from each of the 21 blocks, so that 21-dimensional feature vectors (i.e., feature vectors with fixed scales) can be extracted.
In the embodiment of the present application, each feature map may be divided into 32 blocks, 16 blocks, 8 blocks, 4 blocks, 2 blocks, and 2 blocks, respectively, for 64 blocks, so as to extract a 64-dimensional first feature vector. The specific dividing manner may be determined according to the requirement, and is not limited herein.
And 123, performing full connection processing on the certain number of first feature vectors to obtain a starting value and an ending value of the price interval.
After a certain number of first feature vectors are obtained by performing spatial pyramid pooling on a certain number of feature maps, a full-connection layer with fixed dimensionality can be input, full-connection processing is performed on the certain number of first feature vectors through the full-connection layer of the price interval estimation model, and therefore the initial value and the final value of the price interval are fitted to obtain an estimated price interval. The estimated price interval can be used in a man-machine interaction mode, for example, in a dialogue ordering scene, accurate problem guidance, especially more accurate price guidance can enable a user to quickly find a package suitable for the user, and the time of the user is saved.
In an embodiment of the application, the performing full concatenation processing on the certain number of first feature vectors to obtain a start value and an end value of a price interval optionally includes: splicing a certain number of first eigenvectors into second eigenvectors; and carrying out full connection processing on the second feature vector to obtain a starting value and an ending value of a price interval.
Splicing a certain number of first eigenvectors into one vector to obtain a second eigenvector, inputting the second eigenvector into a full-connection layer of the price interval estimation model, and performing full-connection processing on the second eigenvector through the full-connection layer to enable the output of the full-connection layer to be two values, namely the initial value and the final value of the price interval. The full-connection layer can comprise multiple layers of full-connection, and the number of parameters of the first layer of full-connection layer in the multiple layers of full-connection is the product of the preset number and the preset scale.
In an embodiment of the present application, the loss function corresponding to the price interval estimation model is represented as follows:
wherein, theta is the network parameter of the price interval estimation model, N is the number of input samples of each iteration, xiFor the input matrix corresponding to the ith sample, bxiIs the starting value of the price interval corresponding to the ith sample, exiIs the end value of the price interval corresponding to the ith sample, b (x)i(ii) a Theta) is the estimated value of the estimation model of the price interval to the initial value of the price interval, e (x)i(ii) a Theta) are respectively the predicted values of the price interval prediction model to the price interval termination value, lambda is a preset coefficient, phi (theta) is a regularization term, and theta*The adjusted network parameters are iterated for one time. The loss function is a loss function when the estimated target is a price interval. argmin represents the value of the corresponding variable θ when the latter expression takes the minimum value.
When the price interval estimation model is trained, a sample is prepared, clicking data and order data of each price gear in a preset price interval by a user and merchant data in a preset range of a corresponding position are collected, historical price distribution of the user and supply distribution of merchants are counted, and an initial value and a final value of the corresponding price interval are marked to serve as training samples. When the price interval estimation model is trained, samples with the preset number of samples are input in an iteration mode, the estimated value of the price interval is obtained according to each sample, back propagation is conducted, and the adjusted network parameters are determined.
In one embodiment of the present application, the method may further include: and outputting the estimated click rate and/or conversion rate of the price interval when the estimated price interval is output by the price interval estimation model based on the input matrix.
When the price interval is estimated, the click rate corresponding to the price interval and the price interval can be estimated at the same time, or the conversion rate corresponding to the price interval and the price interval can be estimated at the same time, or the click rate and the conversion rate corresponding to the price interval and the price interval can be estimated at the same time, and of course, other estimation indexes can be added. When the price interval and the click rate corresponding to the price interval are estimated at the same time, the result output by the full connection layer is three values, namely the starting value and the ending value of the price interval and the click rate corresponding to the price interval. When the price interval and the conversion rate corresponding to the price interval are estimated at the same time, the result output by the full connection layer is three values, namely the initial value and the final value of the price interval and the conversion rate corresponding to the price interval. When the price interval and the click rate and the conversion rate corresponding to the price interval are estimated at the same time, the result output by the full connection layer is four values, namely the initial value and the final value of the price interval, and the click rate and the conversion rate corresponding to the price interval. The same sample can be adopted when the click rate and/or the conversion rate are estimated at the same time of estimating the price interval, the price interval and the click rate and/or the conversion rate are estimated at the same time, the sample space is constant, the characteristic scale is consistent, a hierarchical cascade structure does not exist, the problem of error cascade propagation caused by the hierarchical cascade structure is solved, the estimation accuracy of the price interval is improved, and the estimation accuracy of the click rate and/or the conversion rate is improved.
When the estimated target is a price interval and the click rate corresponding to the price interval, the loss function corresponding to the price interval estimation model is expressed as follows:
wherein, theta is the network parameter of the price interval estimation model, N is the number of input samples of each iteration, xiFor the input matrix corresponding to the ith sample, bxiIs the starting value of the price interval corresponding to the ith sample, exiIs the end value of the price interval corresponding to the ith sample, b (x)i(ii) a Theta) is the estimated value of the estimation model of the price interval to the initial value of the price interval, e (x)iTheta) estimation of the end value of the price interval for the price interval estimation model, L1As a function of click rate lossλ is a predetermined coefficient, φ (θ) is a regularization term, θ*The adjusted network parameters are iterated for one time.
Wherein, L1May be a distance loss function and may be expressed asWherein, cxiClick rate of price interval corresponding to ith sample, c (x)i(ii) a Theta) is an estimated value of the click rate of the price interval.
When the estimated target is a price interval and the conversion rate corresponding to the price interval, the loss function corresponding to the price interval estimation model is expressed as follows:
wherein, theta is the network parameter of the price interval estimation model, N is the number of input samples of each iteration, xiFor the input matrix corresponding to the ith sample, bxiIs the starting value of the price interval corresponding to the ith sample, exiIs the end value of the price interval corresponding to the ith sample, b (x)i(ii) a Theta) is the estimated value of the estimation model of the price interval to the initial value of the price interval, e (x)iTheta) estimation of the end value of the price interval for the price interval estimation model, L2For the conversion loss function, λ is a predetermined coefficient, φ (θ) is a regularization term, θ*The adjusted network parameters are iterated for one time. Wherein the conversion loss function may be a logarithmic loss function.
When the estimated target is a price interval, the click rate and the conversion rate corresponding to the price interval, the loss function corresponding to the price interval estimation model is expressed as follows:
the parameters in the above formula are as above.
The price interval estimation method provided by the embodiment of the application comprises the steps of splicing historical price distribution of a user and supply distribution of merchants within a preset range of a designated position into an input matrix, inputting the input matrix into a price interval estimation model, outputting an estimated price interval by the price interval estimation model based on the input matrix, namely performing convolution processing on the input matrix by the price interval estimation model to obtain a certain number of characteristic diagrams, performing space pyramid pooling processing on the certain number of characteristic diagrams respectively to obtain a certain number of first characteristic vectors, performing full connection processing on the certain number of first characteristic vectors to obtain an initial value and a final value of the price interval, so that the price interval can be estimated directly according to historical data of the user and the supply information of the merchants within the range of the designated position without adopting a hierarchical structure of reservation probability estimation and price interval estimation, therefore, the problem of error cascade propagation in the hierarchical structure is solved, and the estimation accuracy of the price interval is improved.
Example two
In the estimation apparatus for price interval according to the present embodiment, as shown in fig. 3, the estimation apparatus 300 for price interval includes:
the input characteristic acquisition module 310 is configured to splice the user historical price distribution and the merchant supply distribution within a preset range of a specified location into an input matrix;
a price interval estimation module 320, configured to input the input matrix into a price interval estimation model, and output an estimated price interval by the price interval estimation model based on the input matrix;
the price interval estimation module 320 includes:
a convolution processing unit 321, configured to perform convolution processing on the input matrix to obtain a certain number of feature maps;
a pooling processing unit 322, configured to perform spatial pyramid pooling on the certain number of feature maps respectively to obtain a certain number of first feature vectors, where a scale of the first feature vectors is a preset scale;
a price interval estimation unit 323, configured to perform full concatenation processing on the certain number of first feature vectors to estimate a start value and an end value of a price interval.
Optionally, the price interval estimation module is specifically configured to:
and inputting the input matrix into a price interval estimation model, and outputting an estimated price interval and an estimated click rate and/or conversion rate of the price interval by the price interval estimation model based on the input matrix.
Optionally, the loss function corresponding to the price interval estimation model is represented as follows:
wherein, theta is the network parameter of the price interval estimation model, N is the number of input samples of each iteration, xiFor the input matrix corresponding to the ith sample, bxiIs the starting value of the price interval corresponding to the ith sample, exiIs the end value of the price interval corresponding to the ith sample, b (x)i(ii) a Theta) is the estimated value of the estimation model of the price interval to the initial value of the price interval, e (x)i(ii) a Theta) are respectively the predicted values of the price interval prediction model to the price interval termination value, lambda is a preset coefficient, phi (theta) is a regularization term, and theta*The adjusted network parameters are iterated for one time.
Optionally, the price interval estimation unit includes:
the characteristic splicing subunit is used for splicing a certain number of first characteristic vectors into second characteristic vectors;
and the price interval estimation subunit is used for carrying out full connection processing on the second feature vector to obtain an initial value and a final value of the price interval.
Optionally, the apparatus further comprises:
the user distribution counting module is used for acquiring historical consumption data within a preset time range of a user, and counting the order quantity distribution of the historical consumption data within a preset price interval to serve as the historical price distribution of the user;
the merchant distribution counting module is used for determining merchants within a preset range of the designated position, counting the commodity quantity distribution of the commodity price of the merchants within a preset price interval according to the commodities of the merchants and the corresponding commodity price, and taking the commodity quantity distribution as the supply distribution of the merchants;
the starting value of the preset price interval is smaller than or equal to the starting value of the price interval, and the ending value of the preset price interval is larger than or equal to the ending value of the price interval.
Optionally, the user distribution statistics module is specifically configured to:
determining orders in the historical consumption data and corresponding order prices, and counting order quantity distribution of the order prices in a preset price interval according to the orders and the corresponding order prices to serve as first price distribution;
determining commodity types in the historical consumption data, determining the commodity type with the largest order quantity as a to-be-counted type, and counting the order quantity distribution of the order price in a preset price interval under the to-be-counted type as second price distribution;
and taking the first price distribution and/or the second price distribution as the user historical price distribution.
Optionally, the merchant distribution statistics module is specifically configured to:
selecting a first preset number of commodities from each merchant as a first commodity to be counted, and counting the commodity number distribution of the commodity price of the first commodity to be counted in a preset price interval as a first supply distribution;
determining brand merchants in the merchants, selecting a first preset number of commodities from each brand merchant as a second commodity to be counted, counting commodity number distribution of commodity prices of the second commodity to be counted in a preset price interval, and taking the commodity number distribution as second supply distribution;
determining commodity types in the historical consumption data, determining the commodity type with the largest order number as a to-be-counted commodity type, selecting a second preset number of commodities under the to-be-counted commodity type from the commodities of the merchant to serve as a third to-be-counted commodity, and counting the commodity number distribution of the commodity price of the third to-be-counted commodity in a preset price interval to serve as a third supply distribution;
at least one of the first offer distribution, the second offer distribution, and the third offer distribution is designated as the merchant offer distribution.
Optionally, the convolution processing unit is specifically configured to:
and performing convolution processing on the input matrix by adopting a certain number of convolution kernels with different widths to obtain a certain number of characteristic graphs.
The price interval estimation device provided in the embodiment of the present application is used to implement each step of the price interval estimation method described in the first embodiment of the present application, and the specific implementation of each module of the device refers to the corresponding step, which is not described herein again.
The price interval estimation device provided by the embodiment of the application splices the user historical price distribution and the merchant supply distribution in the appointed position preset range into an input matrix through an input characteristic acquisition module, the price interval estimation module inputs the input matrix into a one-price interval estimation model, the price interval estimation model outputs the estimated price interval based on the input matrix, the price interval estimation module specifically performs convolution processing on the input matrix through a convolution processing unit to obtain a certain number of characteristic graphs, a pooling processing unit performs space pyramid pooling processing on the certain number of characteristic graphs respectively to obtain a certain number of estimated first characteristic vectors, the price interval unit performs full connection processing on the certain number of first characteristic vectors to obtain the initial value and the final value of the price interval, and therefore the price interval can be estimated directly according to the user historical data and the merchant supply information in the appointed position range, and a hierarchical structure of booking probability estimation and price interval estimation is not needed, so that the problem of error cascade propagation in the hierarchical structure is solved, and the estimation accuracy of the price interval is improved.
EXAMPLE III
Embodiments of the present application also provide an electronic device, as shown in fig. 4, the electronic device 400 may include one or more processors 410 and one or more memories 420 connected to the processors 410. Electronic device 400 may also include input interface 430 and output interface 440 for communicating with another apparatus or system. Program code executed by processor 410 may be stored in memory 420.
The processor 410 in the electronic device 400 calls the program code stored in the memory 420 to perform the estimation method of the price interval in the above-described embodiment.
The above elements in the above electronic device may be connected to each other by a bus, such as one of a data bus, an address bus, a control bus, an expansion bus, and a local bus, or any combination thereof.
The embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the method for estimating a price interval according to the first embodiment of the present application.
The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
The method, the device, the electronic device and the storage medium for estimating the price interval provided by the embodiment of the application are introduced in detail, a specific example is applied in the text to explain the principle and the implementation of the application, and the description of the embodiment is only used for helping to understand the method and the core idea of the application; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. With this understanding in mind, the above-described technical solutions may be embodied in the form of a software product, which can be stored in a computer-readable storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments.
Claims (11)
1. A method for estimating a price interval, comprising:
splicing the historical price distribution of the user and the supply distribution of merchants within a preset range of an appointed position into an input matrix;
inputting the input matrix into a price interval estimation model, and outputting an estimated price interval by the price interval estimation model based on the input matrix;
wherein the step of outputting the estimated price interval based on the input matrix by the price interval estimation model comprises:
performing convolution processing on the input matrix to obtain a certain number of characteristic graphs;
performing spatial pyramid pooling on the certain number of feature maps respectively to obtain a certain number of first feature vectors, wherein the scale of the first feature vectors is a preset scale;
and carrying out full connection processing on the certain number of first feature vectors to obtain a starting value and an ending value of a price interval.
2. The method of claim 1, further comprising:
and outputting the estimated click rate and/or conversion rate of the price interval when the estimated price interval is output by the price interval estimation model based on the input matrix.
3. The method of claim 1, wherein the loss function corresponding to the price interval estimation model is represented as follows:
wherein, theta is the network parameter of the price interval estimation model, N is the number of input samples of each iteration, xiFor the input matrix corresponding to the ith sample, bxiIs the starting value of the price interval corresponding to the ith sample, exiIs the end value of the price interval corresponding to the ith sample, b (x)i(ii) a Theta) is the estimated value of the estimation model of the price interval to the initial value of the price interval, e (x)i(ii) a Theta) are respectively the predicted values of the price interval prediction model to the price interval termination value, lambda is a preset coefficient, phi (theta) is a regularization term, and theta*The adjusted network parameters are iterated for one time.
4. The method of claim 1, wherein the fully concatenating the quantity of first eigenvectors to obtain a start value and an end value of a price interval comprises:
splicing a certain number of first eigenvectors into second eigenvectors;
and carrying out full connection processing on the second feature vector to obtain a starting value and an ending value of a price interval.
5. The method of claim 1, wherein before the step of combining the user historical price distribution and the merchant offer distribution within the preset range of the designated location into the input matrix, the method further comprises:
acquiring historical consumption data within a preset time range of a user, and counting the order quantity distribution of the historical consumption data within a preset price interval to be used as the historical price distribution of the user;
determining merchants within a preset range of a specified position, and counting the commodity quantity distribution of the commodity price of the merchants within a preset price interval according to the commodities of the merchants and the corresponding commodity price to serve as the supply distribution of the merchants;
the starting value of the preset price interval is smaller than or equal to the starting value of the price interval, and the ending value of the preset price interval is larger than or equal to the ending value of the price interval.
6. The method according to claim 5, wherein the counting, as the user historical price distribution, the order quantity distribution of the historical consumption data within a preset price interval comprises:
determining orders in the historical consumption data and corresponding order prices, and counting order quantity distribution of the order prices in a preset price interval according to the orders and the corresponding order prices to serve as first price distribution;
determining commodity types in the historical consumption data, determining the commodity type with the largest order quantity as a to-be-counted type, and counting the order quantity distribution of the order price in a preset price interval under the to-be-counted type as second price distribution;
and taking the first price distribution and/or the second price distribution as the user historical price distribution.
7. The method according to claim 5, wherein the step of counting a distribution of the quantity of the commodities of the merchant in a preset price interval according to the commodities of the merchant and corresponding commodity prices as a merchant supply distribution comprises:
selecting a first preset number of commodities from each merchant as a first commodity to be counted, and counting the commodity number distribution of the commodity price of the first commodity to be counted in a preset price interval as a first supply distribution;
determining brand merchants in the merchants, selecting a first preset number of commodities from each brand merchant as a second commodity to be counted, counting commodity number distribution of commodity prices of the second commodity to be counted in a preset price interval, and taking the commodity number distribution as second supply distribution;
determining commodity types in the historical consumption data, determining the commodity type with the largest order number as a to-be-counted commodity type, selecting a second preset number of commodities under the to-be-counted commodity type from the commodities of the merchant to serve as a third to-be-counted commodity, and counting the commodity number distribution of the commodity price of the third to-be-counted commodity in a preset price interval to serve as a third supply distribution;
at least one of the first offer distribution, the second offer distribution, and the third offer distribution is designated as the merchant offer distribution.
8. The method of claim 1, wherein the convolving the input matrix to obtain a number of feature maps comprises:
and performing convolution processing on the input matrix by adopting a certain number of convolution kernels with different widths to obtain a certain number of characteristic graphs.
9. An apparatus for estimating a price interval, comprising:
the input characteristic acquisition module is used for splicing the historical price distribution of the user and the supply distribution of merchants within a preset range of the designated position into an input matrix;
the price interval estimation module is used for inputting the input matrix into a price interval estimation model, and outputting an estimated price interval by the price interval estimation model based on the input matrix;
wherein, the price interval estimation module comprises:
the convolution processing unit is used for carrying out convolution processing on the input matrix to obtain a certain number of characteristic graphs;
the pooling processing unit is used for respectively performing space pyramid pooling on the certain number of feature maps to obtain a certain number of first feature vectors, and the scale of each first feature vector is a preset scale;
and the price interval estimation unit is used for carrying out full-connection processing on the certain number of first characteristic vectors to obtain a starting value and a terminating value of the price interval.
10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of estimating a price interval according to any one of claims 1 to 8 when executing the computer program.
11. A computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, is adapted to carry out the steps of the method for estimating a price interval of any one of claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010131079.7A CN111461760A (en) | 2020-02-28 | 2020-02-28 | Price interval estimation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010131079.7A CN111461760A (en) | 2020-02-28 | 2020-02-28 | Price interval estimation method and device, electronic equipment and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111461760A true CN111461760A (en) | 2020-07-28 |
Family
ID=71678216
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010131079.7A Pending CN111461760A (en) | 2020-02-28 | 2020-02-28 | Price interval estimation method and device, electronic equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111461760A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113344628A (en) * | 2021-06-04 | 2021-09-03 | 网易(杭州)网络有限公司 | Information processing method and device, computer equipment and storage medium |
-
2020
- 2020-02-28 CN CN202010131079.7A patent/CN111461760A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113344628A (en) * | 2021-06-04 | 2021-09-03 | 网易(杭州)网络有限公司 | Information processing method and device, computer equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN113508378B (en) | Training method, recommendation method, device and computer readable medium for recommendation model | |
US20220198289A1 (en) | Recommendation model training method, selection probability prediction method, and apparatus | |
CN108898459B (en) | Commodity recommendation method and device | |
CN110647696B (en) | Business object sorting method and device | |
TWI648642B (en) | Data search processing method and system | |
CN105989004A (en) | Information releasing pretreatment method and device | |
CN110008397B (en) | Recommendation model training method and device | |
CN112258260A (en) | Page display method, device, medium and electronic equipment based on user characteristics | |
WO2024131762A1 (en) | Recommendation method and related device | |
CN114240555A (en) | Click rate prediction model training method and device and click rate prediction method and device | |
CN111680213B (en) | Information recommendation method, data processing method and device | |
CN111461760A (en) | Price interval estimation method and device, electronic equipment and storage medium | |
CN109885745A (en) | A kind of user draws a portrait method, apparatus, readable storage medium storing program for executing and terminal device | |
CN112348590A (en) | Method and device for determining value of article, electronic equipment and storage medium | |
CN111598638B (en) | Click rate determination method, device and equipment | |
CN109886299A (en) | A kind of user draws a portrait method, apparatus, readable storage medium storing program for executing and terminal device | |
CN111639989B (en) | Commodity recommendation method and readable storage medium | |
CN112650869B (en) | Image retrieval reordering method and device, electronic equipment and storage medium | |
CN105205098B (en) | Method and device for determining click arrival rate (CTR) | |
CN114154064A (en) | Commodity keyword optimization method and device | |
CN113887613A (en) | Deep learning method, device and equipment based on attention mechanism and storage medium | |
CN110837596A (en) | Intelligent recommendation method and device, computer equipment and storage medium | |
CN113763080A (en) | Method and device for determining recommended article, electronic equipment and storage medium | |
CN110069717B (en) | Searching method and device | |
CN112307298A (en) | Method and device for generating personal brand label |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200728 |
|
WD01 | Invention patent application deemed withdrawn after publication |